diff --git a/docs/.buildinfo b/docs/.buildinfo index 31e9e90d2..028bec665 100644 --- a/docs/.buildinfo +++ b/docs/.buildinfo @@ -1,4 +1,4 @@ # Sphinx build info version 1 # This file hashes the configuration used when building these files. When it is not found, a full rebuild will be done. -config: a81c8db7bc7a8bd4bd66e17446e21974 +config: d9a08019dec6882195fa0ef4f685a5cd tags: 645f666f9bcd5a90fca523b33c5a78b7 diff --git a/docs/_modules/brevitas/core/bit_width/const.html b/docs/_modules/brevitas/core/bit_width/const.html index 5ef63d0a4..63cecaf10 100644 --- a/docs/_modules/brevitas/core/bit_width/const.html +++ b/docs/_modules/brevitas/core/bit_width/const.html @@ -8,7 +8,7 @@ - brevitas.core.bit_width.const — Brevitas 0.10.1 documentation + brevitas.core.bit_width.const — Brevitas 0.10.2 documentation @@ -123,8 +123,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/_modules/brevitas/core/bit_width/parameter.html b/docs/_modules/brevitas/core/bit_width/parameter.html index 1bfa58469..7b8942cba 100644 --- a/docs/_modules/brevitas/core/bit_width/parameter.html +++ b/docs/_modules/brevitas/core/bit_width/parameter.html @@ -8,7 +8,7 @@ - brevitas.core.bit_width.parameter — Brevitas 0.10.1 documentation + brevitas.core.bit_width.parameter — Brevitas 0.10.2 documentation @@ -123,8 +123,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/_modules/brevitas/core/function_wrapper/clamp.html b/docs/_modules/brevitas/core/function_wrapper/clamp.html index 5659f20d8..76a755ef8 100644 --- a/docs/_modules/brevitas/core/function_wrapper/clamp.html +++ b/docs/_modules/brevitas/core/function_wrapper/clamp.html @@ -8,7 +8,7 @@ - brevitas.core.function_wrapper.clamp — Brevitas 0.10.1 documentation + brevitas.core.function_wrapper.clamp — Brevitas 0.10.2 documentation @@ -123,8 +123,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/_modules/brevitas/core/function_wrapper/misc.html b/docs/_modules/brevitas/core/function_wrapper/misc.html index 8f7ad7b6c..081a0b671 100644 --- a/docs/_modules/brevitas/core/function_wrapper/misc.html +++ b/docs/_modules/brevitas/core/function_wrapper/misc.html @@ -8,7 +8,7 @@ - brevitas.core.function_wrapper.misc — Brevitas 0.10.1 documentation + brevitas.core.function_wrapper.misc — Brevitas 0.10.2 documentation @@ -123,8 +123,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/_modules/brevitas/core/function_wrapper/ops_ste.html b/docs/_modules/brevitas/core/function_wrapper/ops_ste.html index d88d9b183..cd18f66aa 100644 --- a/docs/_modules/brevitas/core/function_wrapper/ops_ste.html +++ b/docs/_modules/brevitas/core/function_wrapper/ops_ste.html @@ -8,7 +8,7 @@ - brevitas.core.function_wrapper.ops_ste — Brevitas 0.10.1 documentation + brevitas.core.function_wrapper.ops_ste — Brevitas 0.10.2 documentation @@ -123,8 +123,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/_modules/brevitas/core/function_wrapper/shape.html b/docs/_modules/brevitas/core/function_wrapper/shape.html index f24ed913b..d4ca6da1c 100644 --- a/docs/_modules/brevitas/core/function_wrapper/shape.html +++ b/docs/_modules/brevitas/core/function_wrapper/shape.html @@ -8,7 +8,7 @@ - brevitas.core.function_wrapper.shape — Brevitas 0.10.1 documentation + brevitas.core.function_wrapper.shape — Brevitas 0.10.2 documentation @@ -123,8 +123,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/_modules/brevitas/core/quant/binary.html b/docs/_modules/brevitas/core/quant/binary.html index 1f027f3f3..0c1584966 100644 --- a/docs/_modules/brevitas/core/quant/binary.html +++ b/docs/_modules/brevitas/core/quant/binary.html @@ -8,7 +8,7 @@ - brevitas.core.quant.binary — Brevitas 0.10.1 documentation + brevitas.core.quant.binary — Brevitas 0.10.2 documentation @@ -123,8 +123,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/_modules/brevitas/core/quant/delay.html b/docs/_modules/brevitas/core/quant/delay.html index 0470e4cf6..347941eea 100644 --- a/docs/_modules/brevitas/core/quant/delay.html +++ b/docs/_modules/brevitas/core/quant/delay.html @@ -8,7 +8,7 @@ - brevitas.core.quant.delay — Brevitas 0.10.1 documentation + brevitas.core.quant.delay — Brevitas 0.10.2 documentation @@ -123,8 +123,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/_modules/brevitas/core/quant/int.html b/docs/_modules/brevitas/core/quant/int.html index fbb910603..1eb00d929 100644 --- a/docs/_modules/brevitas/core/quant/int.html +++ b/docs/_modules/brevitas/core/quant/int.html @@ -8,7 +8,7 @@ - brevitas.core.quant.int — Brevitas 0.10.1 documentation + brevitas.core.quant.int — Brevitas 0.10.2 documentation @@ -123,8 +123,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/_modules/brevitas/core/quant/int_base.html b/docs/_modules/brevitas/core/quant/int_base.html index c2af8ee18..d4223cf8b 100644 --- a/docs/_modules/brevitas/core/quant/int_base.html +++ b/docs/_modules/brevitas/core/quant/int_base.html @@ -8,7 +8,7 @@ - brevitas.core.quant.int_base — Brevitas 0.10.1 documentation + brevitas.core.quant.int_base — Brevitas 0.10.2 documentation @@ -123,8 +123,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/_modules/brevitas/core/quant/ternary.html b/docs/_modules/brevitas/core/quant/ternary.html index 63da34645..9ac3f7859 100644 --- a/docs/_modules/brevitas/core/quant/ternary.html +++ b/docs/_modules/brevitas/core/quant/ternary.html @@ -8,7 +8,7 @@ - brevitas.core.quant.ternary — Brevitas 0.10.1 documentation + brevitas.core.quant.ternary — Brevitas 0.10.2 documentation @@ -123,8 +123,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/_modules/brevitas/core/restrict_val.html b/docs/_modules/brevitas/core/restrict_val.html index 53dc812a4..cdf92c705 100644 --- a/docs/_modules/brevitas/core/restrict_val.html +++ b/docs/_modules/brevitas/core/restrict_val.html @@ -8,7 +8,7 @@ - brevitas.core.restrict_val — Brevitas 0.10.1 documentation + brevitas.core.restrict_val — Brevitas 0.10.2 documentation @@ -123,8 +123,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/_modules/brevitas/core/scaling/int_scaling.html b/docs/_modules/brevitas/core/scaling/int_scaling.html index 1a18a40ab..ab0663819 100644 --- a/docs/_modules/brevitas/core/scaling/int_scaling.html +++ b/docs/_modules/brevitas/core/scaling/int_scaling.html @@ -8,7 +8,7 @@ - brevitas.core.scaling.int_scaling — Brevitas 0.10.1 documentation + brevitas.core.scaling.int_scaling — Brevitas 0.10.2 documentation @@ -123,8 +123,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/_modules/brevitas/core/scaling/runtime.html b/docs/_modules/brevitas/core/scaling/runtime.html index fc35db615..566a80cc3 100644 --- a/docs/_modules/brevitas/core/scaling/runtime.html +++ b/docs/_modules/brevitas/core/scaling/runtime.html @@ -8,7 +8,7 @@ - brevitas.core.scaling.runtime — Brevitas 0.10.1 documentation + brevitas.core.scaling.runtime — Brevitas 0.10.2 documentation @@ -123,8 +123,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/_modules/brevitas/core/scaling/standalone.html b/docs/_modules/brevitas/core/scaling/standalone.html index 4b6ddb32b..54ec5390e 100644 --- a/docs/_modules/brevitas/core/scaling/standalone.html +++ b/docs/_modules/brevitas/core/scaling/standalone.html @@ -8,7 +8,7 @@ - brevitas.core.scaling.standalone — Brevitas 0.10.1 documentation + brevitas.core.scaling.standalone — Brevitas 0.10.2 documentation @@ -123,8 +123,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/_modules/brevitas/core/stats/stats_op.html b/docs/_modules/brevitas/core/stats/stats_op.html index 32dd0adae..5d95da826 100644 --- a/docs/_modules/brevitas/core/stats/stats_op.html +++ b/docs/_modules/brevitas/core/stats/stats_op.html @@ -8,7 +8,7 @@ - brevitas.core.stats.stats_op — Brevitas 0.10.1 documentation + brevitas.core.stats.stats_op — Brevitas 0.10.2 documentation @@ -123,8 +123,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/_modules/brevitas/core/utils.html b/docs/_modules/brevitas/core/utils.html index fd024bf1d..6c90c9602 100644 --- a/docs/_modules/brevitas/core/utils.html +++ b/docs/_modules/brevitas/core/utils.html @@ -8,7 +8,7 @@ - brevitas.core.utils — Brevitas 0.10.1 documentation + brevitas.core.utils — Brevitas 0.10.2 documentation @@ -123,8 +123,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/_modules/brevitas/core/zero_point.html b/docs/_modules/brevitas/core/zero_point.html index d6bd0229d..53919717c 100644 --- a/docs/_modules/brevitas/core/zero_point.html +++ b/docs/_modules/brevitas/core/zero_point.html @@ -8,7 +8,7 @@ - brevitas.core.zero_point — Brevitas 0.10.1 documentation + brevitas.core.zero_point — Brevitas 0.10.2 documentation @@ -123,8 +123,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/_modules/brevitas/function/ops.html b/docs/_modules/brevitas/function/ops.html index 71aa9ed96..c325b40bd 100644 --- a/docs/_modules/brevitas/function/ops.html +++ b/docs/_modules/brevitas/function/ops.html @@ -8,7 +8,7 @@ - brevitas.function.ops — Brevitas 0.10.1 documentation + brevitas.function.ops — Brevitas 0.10.2 documentation @@ -123,8 +123,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/_modules/brevitas/function/ops_ste.html b/docs/_modules/brevitas/function/ops_ste.html index 41d4a8659..14b00539e 100644 --- a/docs/_modules/brevitas/function/ops_ste.html +++ b/docs/_modules/brevitas/function/ops_ste.html @@ -8,7 +8,7 @@ - brevitas.function.ops_ste — Brevitas 0.10.1 documentation + brevitas.function.ops_ste — Brevitas 0.10.2 documentation @@ -123,8 +123,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/_modules/brevitas/function/shape.html b/docs/_modules/brevitas/function/shape.html index 1a99c0daf..9b6e94515 100644 --- a/docs/_modules/brevitas/function/shape.html +++ b/docs/_modules/brevitas/function/shape.html @@ -8,7 +8,7 @@ - brevitas.function.shape — Brevitas 0.10.1 documentation + brevitas.function.shape — Brevitas 0.10.2 documentation @@ -123,8 +123,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/_modules/brevitas/ops/autograd_ste_ops.html b/docs/_modules/brevitas/ops/autograd_ste_ops.html index 38a3a9f4b..21b701fe8 100644 --- a/docs/_modules/brevitas/ops/autograd_ste_ops.html +++ b/docs/_modules/brevitas/ops/autograd_ste_ops.html @@ -8,7 +8,7 @@ - brevitas.ops.autograd_ste_ops — Brevitas 0.10.1 documentation + brevitas.ops.autograd_ste_ops — Brevitas 0.10.2 documentation @@ -123,8 +123,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/_modules/index.html b/docs/_modules/index.html index 97d332ce9..5380f2cf8 100644 --- a/docs/_modules/index.html +++ b/docs/_modules/index.html @@ -8,7 +8,7 @@ - Overview: module code — Brevitas 0.10.1 documentation + Overview: module code — Brevitas 0.10.2 documentation @@ -123,8 +123,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/_static/documentation_options.js b/docs/_static/documentation_options.js index 1a724067d..db6d22fe9 100644 --- a/docs/_static/documentation_options.js +++ b/docs/_static/documentation_options.js @@ -1,6 +1,6 @@ var DOCUMENTATION_OPTIONS = { URL_ROOT: document.getElementById("documentation_options").getAttribute('data-url_root'), - VERSION: '0.10.1', + VERSION: '0.10.2', LANGUAGE: 'en', COLLAPSE_INDEX: false, BUILDER: 'html', diff --git a/docs/about.html b/docs/about.html index aec38de0b..8c96f9c07 100644 --- a/docs/about.html +++ b/docs/about.html @@ -9,7 +9,7 @@ - About — Brevitas 0.10.1 documentation + About — Brevitas 0.10.2 documentation @@ -125,8 +125,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/api_reference/brevitas.core.bit_width.html b/docs/api_reference/brevitas.core.bit_width.html index 8fab18dd0..c2a0a5090 100644 --- a/docs/api_reference/brevitas.core.bit_width.html +++ b/docs/api_reference/brevitas.core.bit_width.html @@ -9,7 +9,7 @@ - brevitas.core.bit_width package — Brevitas 0.10.1 documentation + brevitas.core.bit_width package — Brevitas 0.10.2 documentation @@ -126,8 +126,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/api_reference/brevitas.core.function_wrapper.html b/docs/api_reference/brevitas.core.function_wrapper.html index b68a6746d..458e945a3 100644 --- a/docs/api_reference/brevitas.core.function_wrapper.html +++ b/docs/api_reference/brevitas.core.function_wrapper.html @@ -9,7 +9,7 @@ - brevitas.core.function_wrapper package — Brevitas 0.10.1 documentation + brevitas.core.function_wrapper package — Brevitas 0.10.2 documentation @@ -126,8 +126,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/api_reference/brevitas.core.html b/docs/api_reference/brevitas.core.html index 1716fd4f4..a95aa4cd4 100644 --- a/docs/api_reference/brevitas.core.html +++ b/docs/api_reference/brevitas.core.html @@ -9,7 +9,7 @@ - brevitas.core package — Brevitas 0.10.1 documentation + brevitas.core package — Brevitas 0.10.2 documentation @@ -126,8 +126,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/api_reference/brevitas.core.quant.html b/docs/api_reference/brevitas.core.quant.html index 3fee9bcfd..da036f4e6 100644 --- a/docs/api_reference/brevitas.core.quant.html +++ b/docs/api_reference/brevitas.core.quant.html @@ -9,7 +9,7 @@ - brevitas.core.quant package — Brevitas 0.10.1 documentation + brevitas.core.quant package — Brevitas 0.10.2 documentation @@ -126,8 +126,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/api_reference/brevitas.core.scaling.html b/docs/api_reference/brevitas.core.scaling.html index 33d48e536..642ea6328 100644 --- a/docs/api_reference/brevitas.core.scaling.html +++ b/docs/api_reference/brevitas.core.scaling.html @@ -9,7 +9,7 @@ - brevitas.core.scaling package — Brevitas 0.10.1 documentation + brevitas.core.scaling package — Brevitas 0.10.2 documentation @@ -126,8 +126,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/api_reference/brevitas.core.stats.html b/docs/api_reference/brevitas.core.stats.html index 0d4983480..986d25373 100644 --- a/docs/api_reference/brevitas.core.stats.html +++ b/docs/api_reference/brevitas.core.stats.html @@ -9,7 +9,7 @@ - brevitas.core.stats package — Brevitas 0.10.1 documentation + brevitas.core.stats package — Brevitas 0.10.2 documentation @@ -126,8 +126,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/api_reference/brevitas.function.html b/docs/api_reference/brevitas.function.html index ce019e845..0a7b9e7ec 100644 --- a/docs/api_reference/brevitas.function.html +++ b/docs/api_reference/brevitas.function.html @@ -9,7 +9,7 @@ - brevitas.function package — Brevitas 0.10.1 documentation + brevitas.function package — Brevitas 0.10.2 documentation @@ -126,8 +126,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/api_reference/brevitas.ops.html b/docs/api_reference/brevitas.ops.html index 3044e3314..e7dbd7ca2 100644 --- a/docs/api_reference/brevitas.ops.html +++ b/docs/api_reference/brevitas.ops.html @@ -9,7 +9,7 @@ - brevitas.ops package — Brevitas 0.10.1 documentation + brevitas.ops package — Brevitas 0.10.2 documentation @@ -126,8 +126,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/api_reference/index.html b/docs/api_reference/index.html index bee6afc49..d7a42f0f1 100644 --- a/docs/api_reference/index.html +++ b/docs/api_reference/index.html @@ -9,7 +9,7 @@ - API reference — Brevitas 0.10.1 documentation + API reference — Brevitas 0.10.2 documentation @@ -126,8 +126,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/architecture.html b/docs/architecture.html index 0eb27d08a..452d6b035 100644 --- a/docs/architecture.html +++ b/docs/architecture.html @@ -9,7 +9,7 @@ - Architecture — Brevitas 0.10.1 documentation + Architecture — Brevitas 0.10.2 documentation @@ -126,8 +126,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/faq.html b/docs/faq.html index 113ceb1fb..afd6d724c 100644 --- a/docs/faq.html +++ b/docs/faq.html @@ -9,7 +9,7 @@ - F.A.Q. — Brevitas 0.10.1 documentation + F.A.Q. — Brevitas 0.10.2 documentation @@ -126,8 +126,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/genindex.html b/docs/genindex.html index dfa8d7550..ac066b5e1 100644 --- a/docs/genindex.html +++ b/docs/genindex.html @@ -8,7 +8,7 @@ - Index — Brevitas 0.10.1 documentation + Index — Brevitas 0.10.2 documentation @@ -123,8 +123,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/getting_started.html b/docs/getting_started.html index 3cb926d7f..038c85b06 100644 --- a/docs/getting_started.html +++ b/docs/getting_started.html @@ -9,7 +9,7 @@ - Getting started — Brevitas 0.10.1 documentation + Getting started — Brevitas 0.10.2 documentation @@ -126,8 +126,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/index.html b/docs/index.html index 30109204d..af2450986 100644 --- a/docs/index.html +++ b/docs/index.html @@ -9,7 +9,7 @@ - Brevitas — Brevitas 0.10.1 documentation + Brevitas — Brevitas 0.10.2 documentation @@ -125,8 +125,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/py-modindex.html b/docs/py-modindex.html index 551d357d9..b335c8f66 100644 --- a/docs/py-modindex.html +++ b/docs/py-modindex.html @@ -8,7 +8,7 @@ - Python Module Index — Brevitas 0.10.1 documentation + Python Module Index — Brevitas 0.10.2 documentation @@ -126,8 +126,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/search.html b/docs/search.html index 831bf0f77..5a23ace8d 100644 --- a/docs/search.html +++ b/docs/search.html @@ -7,7 +7,7 @@ - Search - Brevitas 0.10.1 documentation + Search - Brevitas 0.10.2 documentation @@ -125,8 +125,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/searchindex.js b/docs/searchindex.js index 96808628d..d1a264f26 100644 --- a/docs/searchindex.js +++ b/docs/searchindex.js @@ -1 +1 @@ -Search.setIndex({"docnames": ["about", "api_reference/brevitas.core", "api_reference/brevitas.core.bit_width", "api_reference/brevitas.core.function_wrapper", "api_reference/brevitas.core.quant", "api_reference/brevitas.core.scaling", "api_reference/brevitas.core.stats", "api_reference/brevitas.function", "api_reference/brevitas.ops", "api_reference/index", "architecture", "faq", "getting_started", "index", "settings", "setup", "tutorials/anatomy_quantizer", "tutorials/index", "tutorials/onnx_export", "tutorials/quant_activation_overview", "tutorials/quant_recurrent", "tutorials/quant_tensor_quant_conv2d_overview", "tutorials/tvmcon2021", "user_guide/datatypes", "user_guide/export", "user_guide/graph_transformations", "user_guide/index", "user_guide/precision", "user_guide/proxies", "user_guide/quantized_layers", "user_guide/quantized_tensor", "user_guide/quantizers", "user_guide/scaling", "user_guide/zero_point"], "filenames": ["about.rst", "api_reference/brevitas.core.rst", "api_reference/brevitas.core.bit_width.rst", "api_reference/brevitas.core.function_wrapper.rst", "api_reference/brevitas.core.quant.rst", "api_reference/brevitas.core.scaling.rst", "api_reference/brevitas.core.stats.rst", "api_reference/brevitas.function.rst", "api_reference/brevitas.ops.rst", "api_reference/index.rst", "architecture.rst", "faq.rst", "getting_started.rst", "index.rst", "settings.rst", "setup.rst", "tutorials/anatomy_quantizer.nblink", "tutorials/index.rst", "tutorials/onnx_export.nblink", "tutorials/quant_activation_overview.nblink", "tutorials/quant_recurrent.nblink", "tutorials/quant_tensor_quant_conv2d_overview.nblink", "tutorials/tvmcon2021.nblink", "user_guide/datatypes.rst", "user_guide/export.rst", "user_guide/graph_transformations.rst", "user_guide/index.rst", "user_guide/precision.rst", "user_guide/proxies.rst", "user_guide/quantized_layers.rst", "user_guide/quantized_tensor.rst", "user_guide/quantizers.rst", "user_guide/scaling.rst", "user_guide/zero_point.rst"], "titles": ["About", "brevitas.core package", "brevitas.core.bit_width package", "brevitas.core.function_wrapper package", "brevitas.core.quant package", "brevitas.core.scaling package", "brevitas.core.stats package", "brevitas.function package", "brevitas.ops package", "API reference", "Architecture", "F.A.Q.", "Getting started", "Brevitas", "Settings", "Setup", "Anatomy of a Quantizer", "Tutorials", "ONNX Export", "An Overview of Quantized Activations", "Quantized RNNs and LSTMs", "An overview of QuantTensor and QuantConv2d", "Brevitas TVMCon 2021 tutorial", "<no title>", "Export Compatibility", "<no title>", "User Guide", "<no title>", "<no title>", "<no title>", "<no title>", "<no title>", "<no title>", "<no title>"], "terms": {"alessandro": [0, 22], "pappalardo": [0, 7], "volcaciu": 0, "xilinx": [0, 13, 15, 22, 24], "research": [0, 13], "lab": 0, "For": [0, 10, 12, 13, 16, 18, 19, 20, 21, 22], "privat": 0, "commun": 0, "you": [0, 10, 11, 15, 18, 22], "can": [0, 3, 7, 10, 11, 12, 14, 15, 16, 18, 19, 20, 21, 22, 24], "reach": 0, "me": 0, "alessand": [0, 18, 20], "name_of_my_employ": 0, "dot": 0, "com": [0, 15, 22], "softwar": [0, 10], "brevita": [0, 9, 10, 11, 14, 15, 16, 18, 19, 20, 21, 24], "titl": 0, "publish": 0, "zenodo": 0, "doi": 0, "10": [0, 3, 12, 16, 18, 19, 20, 21, 22], "5281": 0, "3333552": 0, "url": 0, "http": [0, 15, 18, 20, 21, 22], "org": 0, "bit_width": [1, 4, 5, 7, 12, 16, 18, 19, 20, 21, 22], "const": [1, 5, 22], "bitwidthconst": [1, 2, 4, 16, 22], "forward": [1, 2, 3, 4, 5, 6, 12, 16, 18, 20, 21, 22], "bitwidthstatefulconst": [1, 2], "msbclampbitwidth": [1, 2], "paramet": [1, 4, 5, 7, 10, 14, 16, 19, 20, 22], "bitwidthparamet": [1, 2], "removebitwidthparamet": [1, 2], "function_wrapp": [1, 4, 22], "clamp": [1, 4, 7, 8, 16], "clampmin": [1, 3], "scalarclamp": [1, 3], "tensorclamp": [1, 3, 4, 16, 22], "misc": 1, "ident": [1, 3, 4, 5, 7, 16, 19, 22], "inplacelogtwo": [1, 3], "logtwo": [1, 3], "poweroftwo": [1, 3], "ops_st": [1, 10, 16], "ceilst": [1, 3], "dpuroundst": [1, 3], "floorst": [1, 3], "inplacetensorclampst": [1, 3], "roundst": [1, 2, 3, 4, 16, 22], "roundtozerost": [1, 3], "scalarclampminst": [1, 3, 16], "tensorclampst": [1, 3], "shape": [1, 5, 12, 16, 18, 20, 22], "overbatchoveroutputchannelview": [1, 3], "overbatchovertensorview": [1, 3, 5], "overoutputchannelview": [1, 3], "overtensorview": [1, 3, 16, 22], "permutedim": [1, 3], "statsinputviewshapeimpl": [1, 3], "over_batch_over_output_channel": [1, 3, 7], "over_batch_over_tensor": [1, 3, 7], "over_output_channel": [1, 3, 7], "over_tensor": [1, 3, 7], "quant": [1, 12, 16, 18, 19, 20, 21, 22], "binari": [1, 7, 8, 10, 15, 20], "binaryqu": [1, 4, 10, 16], "clampedbinaryqu": [1, 4, 10, 16], "delai": [1, 16, 22], "delaywrapp": [1, 4, 16, 22], "int": [1, 2, 5, 7, 16, 18, 20, 21, 22], "decoupledrescalingintqu": [1, 4], "decoupledrescalingintquantwithinput": [1, 4], "prescaledrestrictintqu": [1, 4], "prescaledrestrictintquantwithinputbitwidth": [1, 4], "rescalingintqu": [1, 4, 16, 22], "truncintqu": [1, 4], "int_bas": 1, "decoupledintqu": [1, 4], "max_int": [1, 4, 7], "min_int": [1, 4, 7], "to_int": [1, 4], "intquant": [1, 4, 16, 22], "ternari": 1, "ternaryqu": [1, 4], "scale": [1, 4, 10, 12, 13, 14, 16, 18, 19, 20, 21], "int_scal": 1, "intscal": [1, 4, 5, 16, 22], "poweroftwointsc": [1, 5], "runtim": [1, 10, 12, 14, 15, 21, 22], "runtimestatssc": [1, 5], "statsfromparametersc": [1, 5], "standalon": [1, 18, 22], "constscal": [1, 4, 5, 16], "parameterfromruntimestatssc": [1, 5, 16, 22], "state_dict": [1, 5, 10, 16, 20, 22], "training_forward": [1, 5], "parameterfromstatsfromparametersc": [1, 5], "parametersc": [1, 5, 16, 22], "stat": [1, 10, 16, 22], "stats_op": 1, "absav": [1, 6], "absmax": [1, 5, 6, 14, 22], "absmaxav": [1, 6], "absmaxl2": [1, 6], "absminmax": [1, 6], "abspercentil": [1, 6, 16, 22], "klminimizerthreshold": [1, 6], "smooth_normalize_distribut": [1, 6], "l1norm": [1, 6], "l2norm": [1, 6], "mse": [1, 5, 6], "evaluate_loss": [1, 6], "mse_fib_search": [1, 6], "mse_grid_search": [1, 6], "mse_loss_fn": [1, 6], "mse_search": [1, 6], "meanlearnedsigmastd": [1, 6], "meansigmastd": [1, 6], "negativeminorzero": [1, 6], "negativepercentileorzero": [1, 6], "percentileinterv": [1, 6], "stats_wrapp": 1, "view_wrapp": 1, "class": [1, 2, 3, 4, 5, 6, 8, 10, 12, 16, 18, 22], "floatrestrictvalu": [1, 16, 22], "sourc": [1, 2, 3, 4, 5, 6, 7, 8, 10, 16, 20, 21, 22], "base": [1, 2, 3, 4, 5, 6, 8, 10, 14, 15, 16, 18, 19, 20], "x": [1, 3, 4, 5, 6, 7, 8, 11, 12, 16, 21, 22], "defin": [1, 2, 3, 4, 5, 6, 10, 16, 18, 19, 20, 21], "comput": [1, 2, 3, 4, 5, 6, 7, 10, 11, 12, 16, 18, 20, 21, 22], "perform": [1, 2, 3, 4, 5, 6, 7, 10, 12, 15, 16, 18, 19, 20, 21, 22], "everi": [1, 2, 3, 4, 5, 6, 10, 12, 15, 22], "call": [1, 2, 3, 4, 5, 6, 10, 12, 15, 16, 19, 21], "should": [1, 2, 3, 4, 5, 6, 10, 11, 16, 21, 22], "overridden": [1, 2, 3, 4, 5, 6], "all": [1, 2, 3, 4, 5, 6, 7, 10, 12, 16, 18, 19, 20, 21, 22], "subclass": [1, 2, 3, 4, 5, 6, 16], "rtype": [1, 2, 3, 4, 5, 6], "tensor": [1, 2, 3, 4, 5, 6, 7, 10, 12, 14, 16, 18, 19, 20, 21], "although": [1, 2, 3, 4, 5, 6, 18, 19, 22], "recip": [1, 2, 3, 4, 5, 6], "pass": [1, 2, 3, 4, 5, 6, 10, 12, 18, 19, 20, 21, 22], "need": [1, 2, 3, 4, 5, 6, 7, 10, 12, 16, 18, 21, 22], "within": [1, 2, 3, 4, 5, 6, 10, 12, 14, 16, 18, 20, 21, 22], "thi": [1, 2, 3, 4, 5, 6, 7, 10, 12, 14, 15, 16, 18, 19, 20, 21, 22], "function": [1, 2, 3, 4, 5, 6, 8, 9, 12, 15, 16, 18, 19, 20, 21, 24], "one": [1, 2, 3, 4, 5, 6, 7, 10, 12, 16, 18, 19, 20, 21, 22], "instanc": [1, 2, 3, 4, 5, 6, 20, 21, 22], "afterward": [1, 2, 3, 4, 5, 6, 22], "instead": [1, 2, 3, 4, 5, 6, 10, 16, 18, 20, 21], "sinc": [1, 2, 3, 4, 5, 6, 7, 15, 16, 18, 19, 21, 22], "former": [1, 2, 3, 4, 5, 6, 16], "take": [1, 2, 3, 4, 5, 6, 10, 12, 16, 18, 19, 20, 21, 22], "care": [1, 2, 3, 4, 5, 6, 16, 22], "run": [1, 2, 3, 4, 5, 6, 10, 11, 13, 16, 18, 20, 22], "regist": [1, 2, 3, 4, 5, 6, 22], "hook": [1, 2, 3, 4, 5, 6], "while": [1, 2, 3, 4, 5, 6, 7, 8, 10, 12, 15, 16, 19, 20, 21, 22], "latter": [1, 2, 3, 4, 5, 6], "silent": [1, 2, 3, 4, 5, 6, 14], "ignor": [1, 2, 3, 4, 5, 6, 14, 22], "them": [1, 2, 3, 4, 5, 6, 10, 12, 16, 21, 22], "restrict_init_float": 1, "return": [1, 2, 4, 5, 7, 10, 12, 16, 18, 19, 20, 21, 22], "type": [1, 2, 4, 5, 7, 10, 12, 13, 16, 18, 21, 22], "float": [1, 2, 4, 5, 7, 10, 11, 14, 15, 16, 18, 19, 20, 21], "restrict_init_inplace_modul": 1, "restrict_init_modul": 1, "restrict_init_tensor": 1, "intrestrictvalu": [1, 2], "restrict_value_float_to_int_impl": 1, "logfloatrestrictvalu": 1, "poweroftworestrictvalu": [1, 5], "parameterwrapp": 1, "valu": [1, 2, 5, 7, 10, 12, 16, 18, 19, 20, 21, 22], "singleargstatelessbuff": 1, "placehold": [1, 5, 21], "slicetensor": 1, "eager_forward": 1, "statelessbuff": [1, 16, 22], "destin": [1, 5], "none": [1, 2, 3, 4, 5, 6, 7, 8, 10, 16, 18, 19, 20, 21, 22], "prefix": [1, 5, 10, 16, 20, 22], "keep_var": [1, 5], "fals": [1, 2, 5, 6, 7, 10, 14, 16, 18, 19, 20, 21, 22], "dictionari": [1, 5], "contain": [1, 5, 7, 10, 14, 16, 22], "refer": [1, 5, 13, 16], "whole": [1, 5, 16], "state": [1, 2, 5, 10, 12, 14, 16, 20, 22], "both": [1, 5, 10, 12, 13, 15, 16, 18, 19, 20, 21, 22], "persist": [1, 5], "buffer": [1, 5, 22], "e": [1, 5, 10, 11, 14, 16, 18, 19, 21, 22, 24], "g": [1, 5, 10, 11, 14, 16, 18, 21, 22, 24], "averag": [1, 5, 16, 19, 21, 22], "ar": [1, 4, 5, 8, 9, 10, 12, 15, 16, 18, 19, 20, 21, 22, 24], "includ": [1, 5, 10, 11, 16, 19, 22], "kei": [1, 5, 14, 16, 20, 22], "correspond": [1, 4, 5, 7, 14, 15, 22], "name": [1, 5, 10, 16, 18, 20, 22], "set": [1, 2, 4, 5, 10, 11, 12, 13, 15, 16, 18, 19, 20, 21, 22, 24], "The": [1, 2, 5, 7, 10, 11, 12, 13, 14, 15, 16, 19, 20, 21, 22], "object": [1, 3, 5, 10, 16], "i": [1, 2, 3, 4, 5, 7, 8, 10, 11, 12, 13, 14, 15, 16, 18, 19, 20, 21, 24], "shallow": [1, 5], "copi": [1, 5], "It": [1, 5, 10, 13, 16, 18, 21, 22], "": [1, 2, 5, 7, 10, 11, 12, 13, 14, 15, 19, 20, 21, 22, 24], "current": [1, 5, 10, 11, 13, 15, 16, 18, 19, 20, 21, 22, 24], "also": [1, 5, 10, 16, 18, 19, 20, 21, 22], "accept": [1, 5, 12, 16, 19, 20, 22], "posit": [1, 5, 19], "argument": [1, 5, 7, 10, 16, 18, 19, 20, 21], "order": [1, 5, 7, 8, 10, 12, 16, 21, 22, 24], "howev": [1, 5, 10, 16, 18, 19, 20, 21, 22], "being": [1, 5, 10, 11, 14, 16, 18, 20, 21, 22], "deprec": [1, 5, 20], "keyword": [1, 5, 10, 16, 19, 20, 21], "enforc": [1, 5, 10, 21, 22], "futur": [1, 5, 20], "releas": [1, 5, 15, 22, 24], "pleas": [1, 5, 20, 21, 22], "avoid": [1, 2, 5, 7, 15, 16, 22], "us": [1, 3, 4, 5, 7, 10, 11, 12, 14, 16, 18, 19, 20, 21, 22], "design": [1, 5, 10, 11, 15, 16, 22], "end": [1, 5, 7, 11, 12, 15, 16, 18, 19, 20, 22], "user": [1, 5, 10, 12, 15, 16, 18, 19, 20, 21, 22], "dict": [1, 2, 5, 10, 14, 16, 22], "option": [1, 5, 16, 18, 20, 21, 22], "If": [1, 5, 10, 16, 18, 21, 22], "provid": [1, 5, 7, 10, 11, 12, 13, 15, 16, 18, 22], "updat": [1, 5, 14, 16, 21, 22], "same": [1, 3, 5, 10, 11, 16, 18, 19, 20, 21, 22], "otherwis": [1, 5, 10, 18, 20, 22], "an": [1, 3, 4, 5, 7, 10, 11, 12, 14, 15, 18, 20, 22, 24], "ordereddict": [1, 5], "creat": [1, 5, 10, 16, 21, 22], "default": [1, 2, 4, 5, 10, 12, 14, 15, 16, 18, 19, 20, 21], "str": [1, 5, 20, 21], "ad": [1, 5, 18, 20, 21], "compos": [1, 5, 10], "bool": [1, 2, 4, 5, 7, 20, 21, 22], "detach": [1, 5, 21, 22], "from": [1, 2, 4, 5, 7, 10, 14, 16, 18, 19, 20, 21], "autograd": [1, 5, 7, 8, 10, 15], "true": [1, 3, 4, 5, 7, 10, 12, 14, 16, 18, 19, 20, 21, 22], "exampl": [1, 2, 3, 4, 5, 7, 8, 10, 12, 19, 20, 21, 22, 24], "xdoctest": [1, 5], "skip": [1, 5, 10, 12, 16, 19], "undefin": [1, 5], "var": [1, 5], "bia": [1, 5, 10, 12, 18, 20], "weight": [1, 4, 5, 6, 7, 10, 14, 18, 21], "inplace_momentum_upd": 1, "momentum": [1, 5], "counter": 1, "new_count": 1, "inplace_tensor_add": 1, "inplace_tensor_mul": 1, "parameterfromruntimezeropoint": 1, "collect_stats_step": [1, 5, 22], "int_quant": [1, 4, 16, 22], "quantize_zero_point": 1, "zero_point_stats_impl": 1, "zero_point_shap": 1, "zero_point_stats_input_view_shape_impl": 1, "zero_point_stats_momentum": 1, "0": [1, 2, 3, 4, 5, 6, 7, 8, 10, 12, 14, 16, 18, 19, 20, 21, 22], "1": [1, 2, 4, 5, 7, 8, 10, 12, 14, 15, 16, 18, 19, 20, 21, 22], "dtype": [1, 2, 5, 6, 12, 18, 22], "devic": [1, 2, 5, 6], "parameterfromstatsfromparameterzeropoint": 1, "zero_point_stats_input_concat_dim": 1, "tracked_parameter_list": [1, 5], "scriptmodul": [1, 2, 3, 4, 5, 6, 16, 22], "implement": [1, 3, 4, 5, 6, 7, 8, 10, 11, 13, 14, 15, 16, 19, 21, 22, 24], "learn": [1, 2, 5, 10, 14, 16, 19, 21], "factor": [1, 4, 5, 10, 16, 18, 19, 21, 22], "initi": [1, 2, 5, 10, 14, 18, 19, 20, 21, 22], "statist": [1, 5, 10, 19, 22], "parameterzeropoint": 1, "zero_point_init": 1, "statsfromparameterzeropoint": 1, "zerozeropoint": [1, 4, 16, 22], "constant": [2, 5, 18, 22], "bit": [2, 4, 7, 10, 12, 16, 18, 19, 20, 21, 24], "width": [2, 4, 7, 10, 18, 20, 21, 24], "wrap": [2, 4, 5, 8, 10, 16, 21], "torch": [2, 3, 4, 5, 7, 8, 10, 12, 14, 15, 16, 18, 19, 20, 21], "8": [2, 3, 5, 7, 12, 15, 16, 19, 20, 21, 22], "part": [2, 7, 10, 15, 16, 22], "mean": [2, 5, 10, 12, 16, 18, 19, 20, 21, 22], "won": [2, 18, 19], "t": [2, 4, 7, 8, 10, 11, 12, 16, 18, 19, 20, 21, 22], "save": [2, 18, 20], "checkpoint": 2, "map": [2, 4, 5, 16, 22], "bit_width_impl_typ": [2, 22], "bitwidthimpltyp": [2, 22], "higher": [2, 4, 5, 16, 18], "level": [2, 4, 5, 10, 13, 16, 20, 22], "api": [2, 4, 5, 10, 13, 18, 22], "retain": [2, 5], "counterpart": [2, 10, 18], "differ": [2, 3, 5, 10, 12, 13, 16, 18, 19, 20, 21, 22], "stateful_const": 2, "bit_width_to_remove_impl": 2, "min_overall_bit_width": 2, "max_overall_bit_width": 2, "input_bit_width": [2, 4, 7, 21, 22], "min_bit_width": 2, "2": [2, 3, 4, 5, 7, 12, 16, 18, 19, 20, 21, 22], "restrict_bit_width_impl": 2, "float_to_int_impl": [2, 4, 16, 22], "override_pretrained_bit_width": 2, "learnabl": 2, "output": [2, 4, 5, 7, 10, 12, 16, 18, 19, 20], "lower": [2, 5, 7, 20], "bound": [2, 5, 7], "restrict": [2, 4, 5, 7, 10, 12, 18, 21, 22, 24], "subset": [2, 16, 24], "pretrain": [2, 10, 12, 14, 16, 20, 22], "load": [2, 10, 14, 15, 16, 22], "backend": [2, 7, 10, 11, 14], "nn": [2, 5, 10, 14, 16, 18, 19, 20, 21, 22], "rais": [2, 5, 16, 18, 19, 20, 21, 22], "runtimeerror": [2, 5, 16, 19, 20, 21, 22], "bit_width_paramet": 2, "grad_fn": [2, 4, 5, 7, 16, 19, 20, 21, 22], "roundstefnbackward": [2, 7, 22], "env": [2, 4, 5, 7, 8, 15, 16, 20, 21, 22], "variabl": [2, 4, 5, 14, 15, 16, 20, 22], "brevitas_ignore_missing_kei": [2, 5, 14, 16, 22], "error": [2, 5, 14, 16, 18, 19, 20, 21, 22], "when": [2, 4, 5, 7, 8, 10, 14, 15, 16, 18, 19, 20, 21, 22], "retrain": [2, 5], "point": [1, 2, 4, 5, 10, 11, 12, 14, 15, 16, 18, 19, 20, 21], "bit_width_to_remov": 2, "non_zero_epsilon": 2, "1e": [2, 6, 20], "06": 2, "remove_zero_bit_width": 2, "wrapper": [3, 7, 10, 22], "variou": [3, 7, 8, 10, 12, 13, 16, 21, 22, 24], "variant": [3, 7, 10, 12, 16, 19, 20, 22], "min_val": [3, 7, 8, 18, 19], "clamp_min": [3, 7, 8], "3": [3, 4, 5, 7, 10, 12, 15, 16, 18, 19, 20, 21, 22], "max_val": [3, 7, 8, 18, 19], "scalar_clamp": 3, "tensor_clamp": [3, 7, 8], "A": [3, 7, 10, 12, 18, 21, 22], "collect": [3, 5, 6, 10, 19, 22], "miscellan": 3, "quantiz": [3, 4, 6, 7, 11, 13, 14, 15, 18, 24], "randn": [3, 7, 12, 16, 18, 19, 20, 21, 22], "size": [3, 10, 11, 20, 21, 22], "y": [3, 4, 7, 8, 11, 16], "log2_": 3, "inplace_log_two": 3, "note": [3, 7, 12, 16, 18, 20, 21, 22, 24], "inplac": 3, "oper": [3, 7, 11, 12, 13, 15, 18, 19, 20, 21, 24], "torchscript": [3, 4, 5, 7, 10, 14, 16, 20], "problemat": 3, "compil": [3, 4, 7, 10, 13, 14, 15, 16, 22], "disabl": [3, 10, 14, 19, 20, 21, 22], "log2": [3, 22], "log_two": 3, "power_of_two": 3, "5": [3, 4, 7, 12, 16, 18, 19, 20, 21, 22], "32": [3, 12, 16, 22], "ceil_st": [3, 7, 8], "dpu_round_st": [3, 7, 8], "floor_st": [3, 7, 8], "tensor_clamp_ste_": [3, 7], "round_st": [3, 7, 8], "round_to_zero_st": [3, 7, 8], "scalar_clamp_min_st": [3, 7, 8], "tensor_clamp_st": [3, 7], "view": [3, 5, 7, 16, 19], "accord": [3, 4, 5, 7, 10, 12, 16, 21, 22], "criteria": [3, 5, 12], "permute_dim": 3, "input": [3, 4, 5, 7, 10, 12, 16, 18, 19, 20], "view_modul": 3, "empti": [3, 5, 10], "25": [3, 16, 20, 22], "scriptmoodul": 3, "250": 3, "16": [3, 12, 16, 19, 20, 22], "200": [3, 18], "6": [3, 4, 5, 12, 16, 18, 19, 20, 21, 22], "2400": 3, "enum": 3, "like": [3, 4, 10, 12, 16, 19, 20, 21, 22], "pointer": [3, 16, 19, 22], "adher": [3, 7, 10, 11], "interfac": [3, 8, 10, 12, 18, 19, 20, 22], "alia": [3, 8, 22], "scaling_impl": [4, 5, 16, 22], "quant_delay_step": [4, 16], "uniform": [4, 10, 11, 13, 16, 20], "binary_sign_st": [4, 7, 8, 16], "number": [4, 5, 7, 10, 16, 19, 20, 22], "train": [4, 5, 10, 11, 12, 13, 16, 19, 20, 21], "step": [4, 16, 18, 19, 22], "de": [4, 10, 16], "format": [4, 10, 11, 12, 16, 18, 19, 21, 22], "zero": [1, 4, 7, 10, 16, 18, 19, 20, 21, 22], "tupl": [4, 5, 7, 10, 16, 18, 20, 21, 22], "import": [4, 12, 16, 18, 19, 20, 21, 22], "binary_qu": [4, 16], "inp": [4, 16, 18, 19, 20, 21, 22], "04": [4, 16, 21], "out": [4, 12, 13, 16, 18, 19, 21, 22], "zero_point": [4, 16, 19, 20, 21, 22], "1000": [4, 7, 16, 20, 22], "quant_typ": [4, 10, 16, 22], "quanttyp": [4, 10, 16, 22], "appli": [4, 5, 7, 8, 10, 13, 16, 19, 20, 21, 22], "brevitas_jit": [4, 7, 8, 10, 11, 14, 15, 16, 20, 22], "enabl": [4, 7, 10, 11, 12, 14, 15, 16, 19, 21, 22], "tensor_clamp_impl": [4, 16, 22], "befor": [4, 5, 16, 18, 20, 21, 22], "go": [4, 5, 10, 16, 18, 19, 20, 21, 22], "through": [4, 7, 8, 10, 12, 14, 15, 16, 18, 19, 20, 22, 24], "between": [4, 7, 10, 12, 16, 18, 19, 21, 22], "which": [4, 7, 10, 11, 14, 16, 18, 19, 20, 21, 22], "backward": [4, 7, 8], "gradient": [4, 7, 8, 14, 21], "outsid": [4, 11, 16, 22], "rang": [4, 18, 19, 22], "wise": [4, 10, 11], "requires_grad_": 4, "mulbackward0": [4, 16, 19, 21, 22], "grad": [4, 7], "0000": [4, 7, 19, 20, 22], "activ": [4, 5, 10, 14, 18, 20, 21], "decoupled_int_qu": 4, "pre_scaling_impl": 4, "int_scaling_impl": [4, 16, 22], "pre_zero_point_impl": 4, "zero_point_impl": [4, 16, 22], "bit_width_impl": [4, 6, 22], "input_is_sign": [4, 7], "around": [4, 10, 12, 16, 22], "integ": [4, 7, 11, 18, 19, 21, 22], "taken": [4, 22], "narrow_rang": [4, 5, 7, 18, 22], "sign": [4, 5, 6, 7, 8, 10, 16, 18, 19, 21, 22], "int_quant_wrapp": 4, "01": [4, 16], "4": [4, 5, 7, 12, 16, 18, 19, 20, 21, 22], "042": 4, "053": 4, "31": [4, 16, 22], "44": [4, 22], "0400": 4, "0500": 4, "0700": [4, 21], "0100": [4, 16], "respect": [4, 10, 12, 16, 21, 22], "here": [4, 16, 22], "interpret": [4, 18], "threshold": 4, "0429": 4, "0571": 4, "0143": 4, "shift": [4, 19], "pre": [1, 4, 10, 16, 22], "flag": [4, 14, 16, 20, 21, 22], "determin": 4, "whether": [4, 7, 10, 14, 16, 18, 21], "narrow": [4, 7, 18], "convers": [4, 19, 20], "represent": [4, 7, 10, 12, 18, 19, 21, 22], "pre_scal": 4, "pre_zero_point": 4, "02": [4, 21], "0200": 4, "0300": [4, 20], "ternary_sign_st": [4, 7, 8], "ternar": 4, "w": [4, 7, 8, 10, 11], "r": [4, 7, 8, 10], "ternary_qu": 4, "scaling_stats_impl": [5, 22], "scaling_stats_input_view_shape_impl": [5, 22], "restrict_scaling_impl": [5, 22], "scaling_shap": [5, 16, 22], "affine_resc": 5, "affine_shift_scal": 5, "scaling_stats_momentum": 5, "scaling_min_v": 5, "scaling_stats_input_concat_dim": 5, "scaling_init": [5, 16, 22], "union": [5, 21, 22], "some": [5, 10, 11, 12, 15, 18, 19, 20, 21, 22], "forc": [5, 10, 14, 20, 21, 22], "method": [5, 16, 20, 21], "singl": [5, 10, 13, 15, 16, 18, 20, 22], "requir": [5, 10, 14, 16, 19, 20, 21, 22, 24], "earli": 5, "version": [5, 10, 15, 16, 18, 20, 22], "consist": [5, 16], "across": [5, 10, 12, 13, 20, 22], "scaling_impl_typ": [5, 22], "scalingimpltyp": [5, 10, 22], "permute_impl": 5, "work": [5, 10, 16, 20, 22], "two": [5, 10, 12, 16, 18, 19, 20, 21, 22], "phase": [5, 19, 22], "dure": [5, 11, 16, 18, 19, 22], "first": [5, 8, 10, 12, 16, 18, 19, 20, 21, 22], "fashion": [5, 22], "batchnorm": 5, "mode": [5, 10, 19, 21, 22], "per": [5, 6, 16, 18, 20, 21], "batch": [5, 7, 11, 18, 19, 20, 22], "background": [5, 16], "infer": [5, 10, 11, 12, 13, 22, 24], "second": [5, 16, 19, 20, 21, 22], "accumul": [5, 7, 10, 19, 21, 22], "behaviour": [5, 10, 14, 16, 19, 20, 21, 22], "scalar_shap": 5, "move": [5, 19, 22], "default_momentum": 5, "scaling_stats_permute_dim": [5, 22], "arang": 5, "randn_lik": 5, "absbinarysigngradfnbackward": [5, 7, 16], "parameter_from_stat": [5, 10, 22], "stats_input": 5, "extend": [5, 10, 12, 16, 18, 22], "scalar": [5, 7, 18], "non": [5, 10, 13, 16, 19, 21], "powbackward1": 5, "stats_reduce_dim": [1, 6, 22], "keepdim": 6, "high_percentile_q": [6, 22], "percentile_q": 6, "num_bin": 6, "1001": 6, "smoothing_ep": 6, "0001": [6, 21], "apach": 6, "incub": 6, "mxnet": 6, "p": [6, 20], "ep": 6, "channel": [6, 7, 10, 11, 16, 18, 20], "l1": [6, 7], "normal": [6, 16, 18, 19, 22], "l2": 6, "proxy_modul": 6, "mse_init_op": 6, "inner_stats_input_view_shape_impl": 6, "mse_search_method": 6, "fibonacci": 6, "mse_it": 6, "20": [6, 16, 19, 20, 22], "candid": 6, "xl": 6, "xr": 6, "quant_valu": [6, 21], "sigma": 6, "stats_output_shap": 6, "std_dev_epsilon": 6, "08": 6, "low_percentile_q": 6, "core": [7, 9, 16, 22], "often": [7, 16], "impos": [7, 10, 16, 24], "pytorch": [7, 10, 11, 12, 13, 15, 16, 18, 20, 21, 22, 24], "binary_sign": [7, 8], "dpu_round": [7, 8], "dpu": [7, 24], "round": [7, 8, 21, 22], "get_upper_bound_on_l1_norm": 7, "accumulator_bit_width": 7, "calcul": 7, "upper": 7, "norm": [7, 18], "deriv": [7, 10, 12, 22], "neural": [12, 22], "network": [12, 20, 22], "low": [11, 12, 18, 21, 22, 24], "precis": [10, 11, 12, 13, 18, 22], "guarante": 7, "overflow": 7, "colbert": 7, "j": 7, "petri": 7, "koenig": 7, "THe": 7, "7": [7, 15, 16, 18, 19, 20, 21, 22], "max_float": 7, "exponent_bit_width": 7, "mantissa_bit_width": 7, "exponent_bia": 7, "maximum": [7, 16, 22], "given": [7, 18, 20, 21, 22], "indic": 7, "repres": [7, 10, 12, 18, 20, 21, 22], "unsign": [7, 12, 18, 19, 22], "avail": [7, 16, 22], "127": [7, 18, 21, 22], "254": 7, "255": 7, "minimum": [7, 18], "128": [7, 18, 22], "round_to_zero": [7, 8], "toward": [7, 10, 11], "gener": [7, 10, 11, 13, 14, 15, 16, 18, 19, 20, 21, 22, 24], "support": [7, 10, 11, 12, 13, 14, 15, 16, 18, 20, 21, 22, 24], "broadcast": 7, "differenti": [7, 22], "tensor_clamp_": [7, 8], "In": [7, 10, 12, 14, 16, 18, 19, 20, 21, 22, 24], "place": [7, 12, 16, 22], "Not": [7, 21], "wrt": 7, "ani": [7, 10, 11, 12, 14, 16, 18, 19, 20, 21, 22, 24], "straight": [7, 8, 10, 16], "estim": [7, 8, 10, 14], "dispatch": 7, "either": [7, 10, 12, 16, 22], "nativ": [7, 14], "just": [7, 10, 12, 14, 16, 19, 21, 22], "time": [7, 10, 11, 12, 13, 14, 15, 16, 18, 21, 22], "autograd_ste_op": [7, 10], "allow": [7, 10, 16, 18, 19, 20, 21, 22], "built": [7, 10, 14, 18, 22], "abs_binary_sign_grad": [7, 8], "ab": [7, 8, 16, 22], "have": [7, 8, 10, 11, 12, 16, 18, 19, 20, 21, 22], "subgradi": [7, 8], "compar": [7, 8, 12, 18, 19, 20], "abs_binary_sign_grad_impl": [7, 8], "its": [7, 10, 11, 12, 16, 18, 19, 20, 21, 22], "requires_grad": [7, 22], "item": [7, 16, 19, 21], "binary_sign_ste_impl": [7, 8], "binarysignstefnbackward": 7, "ceil": [7, 8], "ceil_ste_impl": [7, 8], "ceilstefnbackward": [7, 22], "dpu_round_ste_impl": [7, 8], "dpuroundstefnbackward": 7, "floor": [7, 8], "floor_ste_impl": [7, 8], "floorstefnbackward": 7, "round_ste_impl": [7, 8], "round_to_zero_ste_impl": [7, 8], "roundtozerostefnbackward": 7, "alwai": [7, 8, 10, 12, 18, 19, 20, 21, 22], "scalar_clamp_min_ste_impl": [7, 8], "c": [7, 10, 11, 14, 15, 16, 19, 20, 22], "5000": 7, "4000": 7, "scalarclampminstefnbackward": 7, "scalar_clamp_st": [7, 8], "scalar_clamp_ste_impl": [7, 8], "scalarclampstefnbackward": 7, "tensor_clamp_ste_impl": [7, 8], "tensorclampstefnbackward": 7, "tensor_clamp_ste_impl_": [7, 8], "inplacetensorclampstefnbackward": 7, "ternary_sign_ste_impl": [7, 8], "ternarysignstefnbackward": 7, "induc": 7, "flatten": 7, "along": [7, 10], "certain": [7, 10, 13, 16, 20, 22, 24], "dimens": 7, "dim": [7, 16, 18, 20, 22], "other": [7, 10, 12, 16, 18, 19, 20, 21, 22, 24], "featur": [7, 10, 16, 18, 20, 22, 24], "arg": [7, 8, 16, 18, 22], "flat": [7, 10, 16], "absbinarysigngradfn": 8, "kwarg": [8, 12, 16, 19, 20, 21, 22], "alias": 8, "see": [8, 11, 16, 18, 19, 21, 22], "detail": [8, 10, 12], "binarysignstefn": 8, "ceilstefn": 8, "dpuroundstefn": 8, "floorstefn": 8, "inplacetensorclampstefn": 8, "roundstefn": 8, "roundtozerostefn": 8, "scalarclampminstefn": 8, "invok": [8, 10, 22], "scalarclampstefn": 8, "tensorclampstefn": 8, "ternarysignstefn": 8, "wip": 9, "most": [9, 10, 16, 19, 21, 22], "packag": [9, 10, 14, 15, 16, 21, 22], "still": [9, 11, 16, 18, 20, 22], "miss": [9, 14, 16, 19, 22], "op": [9, 10, 20, 21], "organ": [10, 20], "few": [10, 14, 16, 18, 20, 21, 22], "concept": [10, 20], "list": [10, 16, 19], "below": [10, 22], "found": [10, 14, 16, 20, 21, 22], "under": [10, 13, 16, 21, 22], "python": [10, 14, 15, 16, 18, 20, 21, 22], "csrc": [10, 20], "becaus": [10, 15, 16, 18, 19, 20, 21, 22, 24], "date": 10, "cannot": [10, 14, 16, 22], "jit": [10, 14, 15, 16, 20, 21, 22], "extens": [10, 16, 18, 22], "simplifi": [10, 15], "distribut": [10, 15], "onli": [10, 16, 18, 19, 21, 22, 24], "cpp": [10, 15, 20], "file": [10, 15], "appropri": [10, 14, 15, 16, 22], "so": [10, 12, 14, 15, 16, 18, 19, 20, 21, 22], "fallback": 10, "long": 10, "switch": [10, 16, 22], "happen": [10, 12, 14, 16, 18, 19, 21, 22], "piec": [10, 16, 22], "commonli": [10, 12], "thei": [10, 16, 18, 20, 21, 22], "algorithm": [10, 11, 12, 16, 19, 22], "build": [10, 13, 19, 22], "block": [10, 13], "assembl": [10, 16], "affin": [10, 11, 13, 16, 21, 22], "old": 10, "style": [10, 13, 16, 18, 19, 22], "scripe": 10, "inherit": 10, "mani": [10, 16, 21, 22], "describ": 10, "section": [10, 11, 12, 22], "abov": [10, 18, 21, 22], "modul": [10, 12, 16, 18, 19, 20, 21, 22], "compat": [10, 15], "everyth": [10, 16, 20, 22], "memori": 10, "optim": [10, 16, 18, 20], "more": [10, 11, 12, 16, 18, 20, 22, 24], "complic": [10, 16], "pipelin": 10, "quit": [10, 16, 19, 20, 21], "signific": 10, "thu": [10, 18], "reduc": [10, 11, 13], "intrins": 10, "cost": [10, 11, 22], "awar": [7, 10, 11, 12, 13, 16, 20, 21, 22], "pose": [10, 16], "challang": 10, "term": [10, 11, 12, 15, 16, 21, 22], "how": [10, 11, 12, 16, 18, 20, 21], "achiev": [10, 11, 16], "flexibl": [10, 11, 16], "minim": [10, 19, 20, 22], "redund": 10, "flavour": [10, 22], "adopt": [10, 13, 15, 16, 19, 20, 21, 22], "doe": [10, 11, 12, 16, 18, 22], "highli": 10, "bias": [10, 11, 16, 21, 22], "leverag": [10, 12, 16, 18, 20, 22], "composit": [10, 22], "particular": [10, 11, 16, 18, 19, 20, 22], "favour": [10, 22], "invers": 10, "control": [10, 14, 16, 20, 22], "depend": [10, 19, 20, 21], "inject": 10, "di": [10, 16], "manual": [10, 19, 22], "librari": [10, 12, 16, 18, 20, 22], "explain": [10, 19, 21], "auto": [10, 21, 22], "wire": [10, 22], "machineri": 10, "heart": 10, "ever": 10, "fixtur": 10, "pytest": 10, "alreadi": [10, 16, 19, 20, 21, 22], "know": [10, 16], "high": [10, 21], "case": [10, 12, 14, 16, 18, 19, 20, 21, 22], "idea": [10, 16, 22], "instanti": [10, 12, 16, 20, 22], "togeth": [10, 15, 16, 19, 20, 21, 22], "declar": [10, 16, 22], "attribut": [10, 16, 18, 19, 22], "drive": [10, 22], "mechan": [10, 15, 22], "behind": [10, 16], "process": [10, 16, 22], "match": [10, 12, 16, 18, 20, 22], "throw": 10, "bunch": [10, 16, 20], "compon": [10, 16, 22], "chosen": 10, "automat": [10, 12, 16, 18, 21, 22], "tensor_qu": [10, 16, 19, 21, 22], "expect": [10, 16, 19, 20, 21, 22], "four": [10, 20, 21], "dequant": [10, 12, 18, 19, 21, 22], "power": [10, 16, 22], "wai": [10, 12, 16, 18, 19, 20, 21, 22, 24], "express": [10, 16, 22], "standard": [10, 11, 12, 15, 16, 18, 19, 20, 21, 22, 24], "lend": 10, "themselv": 10, "multipl": [10, 12, 16, 20, 22], "That": [10, 11, 16, 19, 21, 22], "new": [10, 13, 16, 22], "simpli": [10, 16, 21, 22], "exist": [10, 16, 19, 22], "overrid": [10, 16, 18, 21, 22], "smaller": [10, 22], "specif": [10, 11, 16, 22], "shifted_scaled_int": [10, 19], "hold": 10, "scaled_int": [10, 16, 18, 19, 21, 22], "fixed_point": 10, "fix": [10, 16], "older": [10, 16], "v2": 10, "plai": 10, "well": [10, 12, 13, 21, 22], "extendedinjector": [10, 22], "addit": [10, 18, 20, 21, 22, 24], "dynam": 10, "syntax": [10, 16, 19], "intermedi": [10, 12, 18, 22], "abstract": [10, 13, 16, 22], "try": [10, 11, 16, 19, 20, 21, 22], "navig": 10, "confus": 10, "lack": 10, "clear": [10, 22], "hierarchi": [10, 16], "obviou": [10, 21, 24], "addition": [10, 12, 19, 22], "might": [10, 16, 18, 19, 20, 21, 22], "fit": [10, 11], "pure": [10, 22], "perspect": 10, "machin": 10, "awai": [10, 22], "top": [10, 14, 15, 16, 20, 21, 22], "specifi": [10, 12, 18, 19], "kind": [10, 14, 16, 22], "sai": [10, 12, 16, 19, 21, 22], "v": [10, 16], "gonna": [10, 20, 22], "translat": [10, 22], "target": [10, 11, 13, 22, 24], "relationship": 10, "hyperparamet": [10, 16, 22], "realli": [10, 16, 19], "advantag": [10, 12, 16, 22], "scope": [10, 22], "solv": [10, 21, 22], "against": [10, 19], "enough": [10, 16], "weightquantsolv": [10, 22], "itself": [10, 11, 16, 22, 24], "individu": 10, "task": 10, "look": [10, 12, 16, 18, 19, 20, 21, 22], "seen": [10, 16, 18, 19, 21, 22], "actual": [10, 18, 21, 22], "rather": [10, 16, 18, 22], "than": [10, 11, 12, 15, 16, 18, 20, 22], "directli": [10, 15, 18, 20, 21, 22], "Then": [10, 19], "meant": 10, "blueprint": 10, "understand": [10, 16, 22], "similar": [10, 12, 16, 18, 19, 20, 21], "direct": [10, 20, 22], "scaling_per_output_channel": [10, 16, 22], "ha": [10, 12, 13, 16, 18, 20, 21, 22], "alloc": [10, 16, 20], "much": [10, 16, 21], "known": [10, 22], "capabl": [10, 16, 18], "whose": [10, 14, 22], "again": [10, 16, 19, 21], "thank": [10, 16, 22], "whenev": [10, 14, 16, 20, 22], "last": [7, 10, 16, 18, 19, 21, 22], "possibl": [10, 15, 16, 18, 20, 21, 22], "even": [10, 16, 18, 19, 21, 22], "advanc": [10, 22], "mix": [10, 12, 19, 21], "custom": [10, 13, 15, 18, 20], "final": [10, 16, 18, 19, 20, 21], "possibli": [10, 12, 16], "data": [10, 11, 12, 13, 16, 18, 20, 21, 22], "structur": [10, 12, 16, 20, 21, 22], "quant_tensor": [10, 21, 22], "valid": [10, 16, 21], "had": [10, 21], "been": [10, 11, 12, 13, 16, 20, 21, 22], "previous": [10, 21, 22], "now": [10, 12, 16, 19, 20, 21, 22], "mark": [10, 19, 21, 22], "doesn": [10, 12, 16, 18, 19, 21, 22], "carri": [10, 11, 16], "inform": [10, 12, 16, 21, 22], "back": [10, 14, 20], "wa": [10, 11, 16, 19, 20, 21, 22], "arithmet": [10, 11], "implment": 10, "main": [10, 18], "assumpt": [10, 11, 21, 22], "sum": [10, 20, 21], "constrain": 10, "involv": [10, 11, 20, 22], "deal": 10, "residu": 10, "topologi": 10, "connect": 10, "special": [10, 16, 19], "respons": 10, "make": [10, 11, 12, 15, 16, 18, 19, 20, 21, 22], "sure": [10, 15, 18, 20, 21], "wouldn": [10, 16, 21, 22], "re": [10, 14, 16, 18, 20, 21, 22], "necessari": [10, 15, 16, 20, 21], "priori": 10, "later": [10, 16, 19, 22], "model": [10, 11, 13, 14, 15, 16, 20, 22, 24], "definit": [10, 12, 16, 18, 20, 22], "logic": [10, 16], "recomput": [10, 14, 22], "chang": [10, 12, 16, 18, 21, 22], "do": [10, 11, 12, 16, 18, 19, 21, 22], "reconcil": 10, "inher": 10, "rigid": 10, "typic": [10, 11, 14, 16, 19, 21, 22], "execut": [10, 16, 18, 19, 20, 22], "complex": [10, 11, 16, 21], "scenario": [10, 12, 15, 16, 18, 19, 20, 21, 22], "share": [10, 18, 20], "situat": [10, 11, 16, 21, 22], "where": [10, 16, 18, 19, 20, 21, 22], "branch": 10, "without": [10, 12, 14, 16, 18, 21, 22], "requant": 10, "feed": 10, "accomplish": 10, "weightquantproxi": 10, "among": [10, 16, 18, 20], "what": [10, 11, 12, 18, 21, 22], "affine_stat": 10, "concaten": [10, 20], "start": [10, 13, 18, 20, 22], "track": 10, "underli": [10, 22], "quantlay": 10, "quantconv2d": [10, 12, 16, 18, 19, 22], "conv2d": [10, 16, 18, 21], "plu": [10, 20], "seri": [10, 20], "each": [10, 12, 16, 18, 20, 24], "responsibil": 10, "quantinputmixin": 10, "quantoutputmixin": 10, "_quantwbiol_": 10, "quantweightmixin": 10, "quantbiasmixin": 10, "quantwbiol": [10, 21, 22], "quant_weight": [10, 16, 20, 21, 22], "quant_bia": [10, 21, 22], "quant_input": [10, 20, 22], "quant_output": [10, 22], "suppos": 10, "act": [10, 16, 18, 19], "exactli": [10, 22], "weight_quant": [10, 16, 20, 21, 22], "bias_quant": [10, 12, 18, 20, 21, 22], "input_qu": [10, 16, 18, 19, 21, 22], "output_qu": [10, 16, 18, 19, 21, 22], "behav": [10, 19, 21, 22], "expos": [10, 16, 20, 21, 22], "To": [10, 11, 12, 15, 16, 19, 20, 21, 22], "ux": 10, "weight_": [10, 16, 22], "bias_": [10, 16, 22], "input_": [10, 16, 22], "output_": [10, 16, 22], "quantrelu": [10, 12, 16, 18, 19], "prioriti": [10, 22], "over": [10, 18, 19, 22], "real": 10, "life": 10, "want": [10, 12, 16, 21, 22], "exploratori": 10, "analysi": 10, "accuraci": 10, "correct": [10, 20, 21, 22], "hardwar": [10, 11, 13, 22, 24], "concern": [10, 11, 21], "friction": [10, 22], "remain": [10, 18], "partial": [10, 20], "With": [10, 12, 16, 18, 19, 20, 21, 22], "except": [7, 10, 15, 16, 19, 20, 21, 22], "truncavgpool2d": 10, "receiv": 10, "altough": [10, 21], "nor": [10, 20], "return_quant_tensor": [10, 12, 16, 18, 19, 20, 21, 22], "easili": [10, 16, 19, 21, 22], "todo": 10, "why": [11, 12, 16, 21, 22], "fbgemm": 11, "qnnpack": 11, "platform": [11, 13, 15, 24], "novel": [11, 22], "varieti": [11, 22], "loos": 11, "z": 11, "find": [11, 20], "document": [11, 18, 20, 24], "spars": 11, "until": [11, 22], "improv": 11, "feel": 11, "free": 11, "open": 11, "issu": [11, 16, 20, 21, 22], "ask": 11, "our": [11, 12, 16, 18, 22], "gitter": 11, "slow": [11, 15, 20], "lot": [11, 22], "element": [11, 18], "intens": 11, "contribut": [11, 16], "graph": [11, 16, 18, 22], "backproprag": 11, "As": [11, 16, 18, 19, 20, 21, 22, 24], "up": [11, 16, 19], "slower": [11, 15], "resourc": 11, "greater": 11, "effieci": 11, "principl": [11, 16, 21, 22], "trade": 11, "off": [11, 18], "effici": [11, 15], "mitig": [11, 16, 20], "somewhat": 11, "down": 11, "report": 11, "thought": 11, "my": 11, "faster": 11, "am": 11, "wrong": 11, "path": [11, 13, 18], "acceler": [11, 12, 13, 15, 18, 20, 22, 24], "own": [11, 12, 13, 14, 16, 20, 21, 22], "export": [11, 13, 15, 19], "your": [11, 14, 18, 20], "downstream": [11, 24], "toolchain": [11, 12, 15, 22, 24], "pu": 11, "float16": 11, "bfloat16": 11, "bfloat19": 11, "datatyp": [11, 22, 24], "float32": [11, 20], "test": 11, "tpu": 11, "xla": 11, "math": [11, 22], "reason": [11, 16, 20], "risk": [11, 14], "serv": [12, 16, 18, 20, 22], "goal": [12, 16, 18], "flow": [12, 22, 24], "By": [12, 15, 19, 21, 22], "write": 12, "modifi": [12, 16, 20], "origin": [12, 16, 21, 22], "onc": [12, 18, 21], "post": 12, "qat": 12, "scratch": [12, 16, 21], "finetun": [12, 16], "follow": [12, 16, 18, 19, 20, 21, 22], "combin": [12, 13, 22], "best": [12, 16], "approach": [12, 15, 18, 20, 22], "checkout": 12, "done": [12, 16, 18, 21], "imagenet": 12, "classif": 12, "torchvis": 12, "script": 12, "we": [12, 16, 18, 19, 20, 21, 22], "consid": [12, 16, 18, 19, 20, 21], "classic": 12, "lenet": 12, "let": [12, 16, 19, 21, 22], "interest": [12, 13, 16, 22], "assess": 12, "cifar10": 12, "purpos": [12, 20, 22], "tutori": [12, 13, 16, 18, 19, 20, 21], "tradit": [12, 16, 22], "quantlinear": [12, 18, 20, 21], "weight_bit_width": [12, 18, 20, 22], "relu": [12, 16, 18, 19, 22], "max": [12, 16, 18, 21, 22], "pool": [12, 21], "usual": [12, 21], "max_pool2d": [12, 21], "result": [12, 16, 18, 19, 20, 21, 22, 24], "f": [12, 18, 20, 21, 22], "qnn": [12, 18], "quantweightlenet": 12, "def": [12, 16, 18, 19, 20, 21, 22], "__init__": [12, 16, 18, 19, 20, 21, 22], "self": [12, 16, 18, 19, 20, 21, 22], "super": [12, 13, 16, 18, 19, 20, 24], "conv1": 12, "relu1": 12, "conv2": 12, "relu2": 12, "fc1": 12, "120": [12, 16, 19, 21], "relu3": 12, "fc2": 12, "84": 12, "relu4": 12, "fc3": 12, "reshap": 12, "quant_weight_lenet": 12, "storag": 12, "convert": [12, 18], "practic": [12, 19, 21, 22], "too": [12, 16, 19, 20], "keep": [12, 16, 19, 20, 22], "replac": [12, 18, 20, 22], "veri": [12, 15, 16, 19, 21, 22], "introduc": [12, 16, 18, 19, 20, 22], "quantident": [12, 18, 19, 20, 21], "begin": 12, "int8bia": [12, 21], "biasquant": 12, "quantweightactlenet": 12, "quant_inp": 12, "quant_weight_act_lenet": 12, "coupl": [12, 16, 19, 20, 21], "thing": [12, 16, 18, 19, 20, 21, 22], "fact": [12, 19, 22], "int4": 12, "15": [12, 16, 18, 19, 20, 22], "int32bia": [12, 18, 20], "quantweightactbiaslenet": 12, "lowprecisionlenet": 12, "quant_weight_act_bias_lenet": 12, "previou": [12, 16, 19, 20, 21, 22], "propag": [12, 21], "quanttensor": [12, 16, 18, 19, 20], "next": [12, 18, 19], "metadata": [12, 19, 20, 21], "about": [12, 13, 16, 19, 21, 22], "qint": 12, "friendli": [12, 22], "affect": [12, 18, 20], "bias_scal": 12, "input_scal": [12, 21, 22], "weight_scal": [12, 22], "read": 12, "invari": [12, 16, 21, 22], "extra": [12, 16, 18, 21, 22], "One": 12, "popular": [12, 16], "qdq": [12, 18, 22], "qcdq": [12, 20], "insert": [12, 18], "clip": [12, 20], "node": [12, 18, 20, 22], "export_onnx_qcdq": [12, 18, 20], "export_path": [12, 18, 20, 22], "4b_weight_lenet": 12, "4b_weight_act_lenet": 12, "4b_weight_act_bias_lenet": 12, "check": [12, 20], "recurr": [12, 20], "overview": 12, "tvmcon": 12, "setup": 13, "get": [13, 15, 16, 19, 20, 21, 22], "architectur": 13, "faq": 13, "techinqu": 13, "practition": 13, "techniqu": [13, 16], "scheme": 13, "framework": 13, "unifi": 13, "layer": [13, 14, 18, 19, 20, 21, 24], "finn": [13, 15, 21, 22, 24], "onnxruntim": [13, 15, 18, 20, 24], "successfulli": [13, 16, 20, 22], "project": 13, "larg": 13, "commerci": 13, "deploy": 13, "cpu": 13, "gpu": [13, 15], "fpga": [13, 15, 22, 24], "focu": [13, 21], "box": 13, "boolean": [14, 16, 21], "global": 14, "enviroment": 14, "config": [14, 16, 20, 22], "jit_en": 14, "written": 14, "warn": [14, 18, 20, 22], "fall": 14, "pytorch_jit": 14, "ignore_missing_kei": [14, 16, 20, 22], "intern": [14, 15, 16, 20, 21, 22], "better": 14, "altern": [14, 18, 19, 20, 22], "load_state_dict": [14, 16, 20, 22], "strict": [14, 16, 22], "would": [14, 15, 16, 18, 19, 20, 21, 22], "mismatch": 14, "brevitas_verbos": 14, "verbos": [14, 16, 22], "brevitas_native_ste_backend": 14, "native_ste_backend_en": 14, "brevitas_reinit_on_state_dict_load": 14, "reinit_on_state_dict_load": 14, "trigger": [14, 16, 20, 22], "9": [15, 16, 18, 19, 20, 21, 22], "window": 15, "linux": 15, "maco": 15, "recommend": 15, "latest": 15, "pip": [15, 18, 20, 22], "git": [15, 20, 21], "small": [15, 22], "lt": [15, 16, 18, 19, 20, 21, 22], "benefit": 15, "almost": [15, 22], "never": 15, "util": [15, 20], "cpp_extens": 15, "mantain": 15, "precompil": 15, "put": [15, 22], "burden": 15, "present": [15, 16, 18, 22], "primit": 15, "fake": [15, 18, 22], "dataflow": [15, 22, 24], "integr": [15, 22], "onnx": [15, 19, 20, 21, 24], "broad": 16, "sens": [16, 22], "anyth": [16, 19, 22], "terminologi": 16, "prefer": 16, "said": [16, 19, 20], "int8actpertensorfloat": [16, 18, 19, 20, 21, 22], "issubclass": 16, "delay_wrapp": [16, 22], "delay_impl": [16, 22], "_nodelai": [16, 22], "stats_input_view_shape_impl": [16, 22], "_stat": [16, 22], "stats_impl": [16, 22], "restrict_sc": 16, "_restrictvalu": 16, "restrict_value_impl": [16, 22], "clamp_scal": 16, "_clampvalu": 16, "clamp_min_st": [16, 22], "restrict_inplace_preprocess": [16, 22], "restrict_preprocess": [16, 22], "msb_clamp_bit_width_impl": [16, 22], "explod": 16, "straightforward": 16, "numpi": [16, 18, 20, 22], "come": [16, 18, 19, 22], "program": 16, "problem": [16, 18, 21], "decis": 16, "spot": 16, "line": [16, 22], "code": 16, "yet": [16, 20], "after": [16, 19, 21, 22], "loop": 16, "react": 16, "world": [16, 22], "hard": 16, "impact": [16, 20], "recompil": 16, "lossi": 16, "arbitrarli": 16, "injector": 16, "excel": 16, "interwin": 16, "turn": [16, 22], "assembli": [16, 22], "goe": [16, 18, 22], "context": [16, 21, 22], "gather": 16, "demand": [16, 22], "simpl": 16, "mention": [16, 18, 22], "heavi": 16, "binar": 16, "inspect": [16, 20, 21, 22], "ipython": [16, 18, 19, 20, 21, 22], "displai": [16, 18, 20, 21, 22], "markdown": [16, 20, 21, 22], "pretty_print_sourc": [16, 20, 21, 22], "n": [16, 20, 21, 22], "getsourc": [16, 20, 21, 22], "func": [16, 22], "script_method": 16, "apart": 16, "pick": [16, 19, 21, 22], "equal": [16, 18, 21, 22], "random": [16, 19, 20, 21], "manual_tensor_qu": 16, "gt": [16, 18, 19, 20, 21, 22], "noth": 16, "surpris": 16, "limit": [16, 22], "mybinaryquant": 16, "inj_tensor_qu": 16, "retriev": [16, 22], "mychildbinaryquant": 16, "child_inj_tensor_qu": 16, "Or": [16, 21], "mybinaryimpl": 16, "myscalingimpl": 16, "mycomposedbinaryquant": 16, "comp_inj_tensor_qu": 16, "live": 16, "mostli": 16, "land": 16, "afford": 16, "conveni": [16, 22], "proxy_class": [16, 19, 22], "weightquantproxyfrominjector": [16, 22], "mybinaryweightquant": 16, "11": [16, 18, 19, 20, 21, 22], "binary_weight_quant_conv": 16, "signed_t": [16, 19, 20, 21, 22], "training_t": [16, 19, 20, 21, 22], "properli": 16, "form": [16, 19, 22], "12": [16, 19, 20, 21, 22], "assert": [16, 18, 19, 20, 21], "is_valid": [16, 19, 21], "those": [16, 19, 21, 22], "explicitli": [16, 19, 21, 22], "13": [16, 18, 19, 20, 22], "mysignedbinaryweightquant": 16, "signed_quant_weight": 16, "And": [16, 22], "add": [16, 18, 19, 20, 21], "There": [16, 18, 21], "simpler": 16, "did": [16, 18, 22], "could": [16, 21, 22], "small_scale_quant_conv": 16, "weight_sign": 16, "arbitrari": [16, 21, 22], "17": [16, 19, 20, 22], "actquantproxyfrominjector": [16, 19, 22], "mysignedbinaryactquant": 16, "binary_relu": 16, "act_quant": [16, 19, 22], "isn": 16, "keywork": 16, "18": [16, 19, 20, 22], "small_scale_binary_ident": 16, "001": [16, 18], "0010": [16, 21], "far": [16, 19, 20, 22], "show": [16, 18, 22], "sort": [16, 19, 21, 22], "shine": 16, "19": [16, 19, 20, 22], "paramfrommaxweightquant": 16, "decor": 16, "spirit": 16, "properti": [16, 19, 22], "param_from_max_quant_conv": 16, "1876": [16, 21], "inde": [16, 21], "verifi": [16, 21], "quant_weight_scal": [16, 21], "21": [16, 20, 21, 22], "simuat": 16, "separ": [16, 21, 22], "22": [16, 20, 21, 22], "float_conv": [16, 21], "1897": 16, "maxbackward1": 16, "23": [16, 18, 22], "traceback": [16, 19, 21, 22], "recent": [16, 19, 21, 22], "5b3646241211": 16, "programdata": 16, "miniconda3": [16, 22], "lib": [16, 21, 22], "site": [16, 21, 22], "py": [16, 18, 19, 20, 21, 22], "1405": [16, 22], "len": [16, 22], "error_msg": [16, 22], "1406": [16, 22], "39": [16, 18, 19, 20, 22], "1407": [16, 22], "__class__": [16, 22], "__name__": [16, 22], "34": [16, 18, 19, 21, 22], "join": [16, 22], "1408": [16, 22], "_incompatiblekei": [16, 22], "missing_kei": [16, 22], "unexpected_kei": [16, 22], "1409": [16, 22], "ouch": 16, "forgiv": 16, "strategi": [16, 22], "26": [16, 20, 22], "quant_conv1": 16, "quant_conv2": 16, "27": [16, 20, 22], "assertionerror": 16, "tmp": [16, 21], "ipykernel_58415": 16, "1066539094": 16, "access": 16, "parent": [16, 19], "28": [16, 22], "sharedparamfrommeanweightquant": 16, "isinst": [16, 19], "cat": [16, 20, 22], "els": [16, 19, 20, 21, 22], "old_quant_conv1_scal": 16, "new_quant_conv1_scal": 16, "eager": [16, 22], "don": [16, 19, 20, 22], "semant": [16, 20], "correctli": [16, 19, 21], "easier": 16, "someth": [16, 19, 22], "mind": [16, 19], "linear": [16, 18, 19, 22], "identityqu": 16, "count": 16, "though": [16, 18, 19, 22, 24], "belong": 16, "quant_conv_w_init": 16, "init": [16, 21], "uniform_": 16, "anymor": 16, "33": [16, 22], "init_tensor_qu": [16, 19, 22], "plan": 16, "distinguish": 16, "illustr": [16, 20, 24], "abl": [16, 22], "leav": 16, "commonquant": 16, "is_clamp": 16, "per_channel_broadcastable_shap": 16, "advancedweightquant": 16, "num_ch": 16, "advancedactquant": 16, "unpack": [16, 22], "chain": 16, "35": [16, 22], "per_channel_quant_conv": 16, "weight_is_clamp": 16, "weight_scaling_per_output_channel": [16, 18, 20, 22], "1842": 16, "1838": 16, "vector": 16, "observ": [16, 18, 19, 20, 21], "36": [16, 22], "1875": [16, 22], "similarli": [16, 18, 20, 22], "37": [16, 22], "quant_ident": [16, 19, 20, 21, 22], "dependencyerror": [16, 19], "b3479e90d1a9": 16, "brevitas_fx": [16, 19], "src": [16, 18, 19, 20, 21, 22], "quant_activ": [16, 19], "134": [16, 19], "135": 16, "136": 16, "137": 16, "quant_lay": [16, 19, 21, 22], "act_impl": [16, 19], "passthrough_act": [16, 19], "77": [16, 19, 20], "78": [16, 19], "79": [16, 19], "80": [3, 16, 19], "81": [16, 19], "mixin": [16, 19, 20], "157": [16, 19], "proxy_prefix": [16, 19], "act_": [16, 19], "158": [16, 19], "kwargs_prefix": [16, 19], "159": [16, 19], "160": [16, 19, 21], "161": [16, 19, 21], "proxy_protocol": [16, 19], "none_quant_injector": [16, 19], "98": [16, 19, 22], "quant_injector": [16, 19], "99": [16, 19, 22], "filter_kwarg": [16, 19], "100": [16, 18, 19, 20, 22], "101": [16, 19], "102": [16, 19], "runtime_qu": [16, 19], "108": [16, 19, 21], "109": [16, 19, 21], "110": [16, 19, 21], "111": [16, 19, 21], "is_passthrough_act": [16, 19], "_is_passthrough_act": [16, 19], "112": [16, 19, 20, 21], "quant_proxi": [16, 19], "export_mod": [16, 19, 21, 22], "export_handl": [16, 18, 19, 21, 22], "74": [16, 19, 22], "modulelist": [16, 19, 20], "75": [16, 19], "tracked_module_list": [16, 19], "76": [16, 19], "add_tracked_modul": [16, 19], "130": [16, 19], "append": [16, 19], "131": [16, 19], "update_tracked_modul": [16, 19], "132": [16, 19], "133": [16, 19], "121": [16, 19, 21], "122": [16, 19, 21, 22], "123": [16, 19, 21], "124": [16, 19, 21], "is_act_en": [16, 19], "_is_act_en": [16, 19], "hidden": [16, 19, 20], "frame": [16, 19], "_depend": 16, "__call__": 16, "__self__": 16, "49": 16, "50": 16, "51": 16, "getattr": 16, "symbol": [16, 22], "52": 16, "53": 16, "messag": 16, "resolv": [16, 19], "complet": [16, 22], "instal": [18, 20, 22], "pypi": [18, 22], "notebook": [18, 22], "onnxoptim": [18, 22], "netron": [18, 20, 22], "visual": [18, 20, 22], "qlinearconv": [18, 22], "qlinearmatmul": [18, 22], "cover": [18, 19, 22], "qonnx": [18, 20, 22], "q": 18, "dq": 18, "three": 18, "fp": [18, 22], "u": 18, "int8": [18, 21], "ntenger": 18, "min": 18, "equival": [18, 19, 22], "sever": 18, "implic": [18, 20], "associ": [18, 20, 22], "int32": [18, 22], "deduantizelinear": 18, "desir": 18, "opset": [18, 20, 21, 22], "consider": 18, "quantconv1d": [18, 21, 22], "quantconvtranspose1d": [18, 21, 22], "quantconvtranspose2d": [18, 21, 22], "offer": 18, "ifram": [18, 20, 22], "show_netron": [18, 20, 22], "model_path": [18, 20, 22], "port": [18, 20, 22], "sleep": [18, 20, 22], "address": [18, 20, 22], "localhost": [18, 20, 22], "brows": [18, 20, 22], "height": [18, 20, 22], "400": [18, 20, 22], "in_ch": 18, "out_ch": 18, "batch_siz": [18, 20], "quant_linear_qcdq": 18, "exported_model": [18, 20], "opset_vers": [18, 20], "8082": [18, 20, 22], "moreov": 18, "interv": [18, 21], "perfectli": [18, 19], "symmetr": [18, 22], "absorb": 18, "fusion": 18, "entir": [18, 22], "quantmodel": 18, "quant_model_qcdq": 18, "8083": [18, 20, 22], "furthermor": 18, "uint8": 18, "expand": 18, "imposs": 18, "less": 18, "etc": 18, "eval": [18, 19, 22], "quant_model_3b_4b_qcdq": 18, "8084": [18, 22], "anoth": [18, 21, 22], "preceed": 18, "opposit": 18, "contraint": 18, "export_onnx_qop": [18, 22], "img_siz": 18, "kernel_s": [18, 19, 21, 22], "quant_model_qop": 18, "manag": 18, "userwarn": [18, 20, 22], "default_opset": 18, "ka": 18, "ir_vers": 18, "producer_nam": 18, "producer_vers": 18, "constant_output_0": 18, "op_typ": 18, "data_typ": 18, "raw_data": 18, "000": 18, "constant_1_output_0": 18, "constant_1": 18, "quantizelinear_output_0": 18, "374": 18, "372": 18, "376": 18, "005": 18, "375": 18, "004": 18, "373": 18, "007": 18, "377": 18, "371": 18, "003": 18, "006": 18, "002": 18, "263": 18, "341": 18, "constant_2_output_0": 18, "constant_2": 18, "271": 18, "032": 18, "0009": [18, 21], "302": 18, "031": 18, "024": 18, "000d": 18, "327": 18, "363": 18, "377u": 18, "000t": 18, "321": 18, "236": 18, "241": 18, "237": 18, "010": 18, "350": 18, "267": 18, "355": [18, 22], "000n": 18, "346": 18, "317": 18, "207": 18, "000y": 18, "326": 18, "334": [18, 21], "362": 18, "304": 18, "340": 18, "275": 18, "324": 18, "332": [18, 21], "026": 18, "333": [18, 21], "335": 18, "226": 18, "0002": [18, 21], "000f": 18, "310": 18, "344": 18, "177": 18, "033": 18, "000i": 18, "315": 18, "035": 18, "354": [18, 20], "377z": 18, "377i": 18, "030": 18, "000w": 18, "303": 18, "022": 18, "336": 18, "377p": 18, "351": 18, "000x": 18, "247": 18, "000h": 18, "225": 18, "3776": 18, "301": 18, "210": 18, "307": 18, "320": 18, "377f": 18, "352": 18, "3770": 18, "313": 18, "261": 18, "037": 18, "220": 18, "202": 18, "013": 18, "266": 18, "347": 18, "377o": 18, "244": 18, "214": 18, "014": 18, "qlinearconv_output_0": 18, "dilat": [18, 21, 22], "group": [18, 21, 22], "kernel_shap": 18, "pad": [18, 21, 22], "stride": [18, 21, 22], "torch_jit": 18, "tensor_typ": 18, "elem_typ": 18, "dim_valu": 18, "126": [18, 21], "opset_import": 18, "domain": [18, 22], "8085": [18, 20, 22], "shown": [18, 22], "prevent": 18, "captur": [18, 20, 22], "examin": 18, "ort": [18, 20], "sess_opt": 18, "sessionopt": 18, "sess": [18, 20], "inferencesess": [18, 20], "input_nam": [18, 20], "get_input": [18, 20], "pred_onx": 18, "out_brevita": 18, "out_ort": 18, "allclos": [18, 20], "kernel": [18, 22], "seem": 18, "must": 18, "behavior": 18, "quantconvnd": 18, "qgemm_ort": 18, "unfortun": 18, "log": [18, 22], "unoptim": 18, "conv": 18, "output_bit_width": [18, 20, 21, 22], "quant_model_qops_4b_4b": 18, "atol": [18, 20], "vecaus": 18, "At": 18, "due": 18, "slightli": 18, "closer": 18, "toler": 18, "deeper": 19, "were": [19, 22], "obtain": [19, 21], "manual_se": [19, 20, 21, 22], "output_quant_conv": [19, 21], "in_channel": [19, 21], "out_channel": [19, 21], "default_quant_conv": [19, 21], "output_identity_qu": 19, "out_tensor1": 19, "out_tensor2": 19, "isclos": [19, 21], "input_output_quant_conv": 19, "input_identity_qu": 19, "becom": [19, 21], "clearer": 19, "meantim": 19, "disabled_quant_ident": 19, "return_quant_ident": 19, "out_tensor": [19, 21], "4566": [19, 20], "5707": 19, "5517": [19, 20], "5897": 19, "5409": 19, "5136": 19, "1902": 19, "0761": 19, "4946": 19, "5029": 19, "4376": 19, "3317": 19, "6361": 19, "0736": [19, 20], "7122": 19, "3780": 19, "1224": 19, "3234": 19, "0844": 19, "0951": 19, "7610": 19, "5980": 19, "0190": 19, "7419": 19, "6278": 19, "6468": 19, "2473": 19, "5327": 19, "1605": 19, "7990": 19, "2936": 19, "3127": 19, "2283": 19, "4351": [19, 20], "3615": 19, "2175": 19, "9214": 19, "divbackward0": [19, 20, 21, 22], "whather": 19, "strip": 19, "implicitli": [19, 21, 22], "out_torch_tensor": 19, "return_disabled_quant_ident": 19, "identity_out_tensor": 19, "uint8actpertensorfloat": [19, 20, 22], "return_quant_relu": 19, "5974": 19, "5402": 19, "5041": 19, "1867": 19, "4481": 19, "3255": 19, "0817": 19, "7083": 19, "3804": [19, 20], "0187": 19, "6254": 19, "6348": 19, "1668": 19, "4387": 19, "2334": 19, "7935": 19, "9230": 19, "0093": [19, 21], "preserv": [19, 21], "return_disabled_quant_relu": 19, "relu_out_tensor": 19, "quantsigmoid": 19, "return_disabled_quant_sigmoid": 19, "sigmoid_out_tensor": 19, "3878": 19, "3611": 19, "3655": [19, 20, 21], "6433": [19, 20], "8236": 19, "6257": 19, "3567": [19, 20], "5474": 19, "4810": 19, "3788": 19, "1820": 19, "4526": 19, "6077": 19, "7911": 19, "1630": 19, "8883": 19, "8471": 19, "9151": 19, "2456": 19, "4198": 19, "2527": [19, 20], "4762": 19, "3184": 19, "1683": 19, "5048": 19, "3226": 19, "6520": 19, "6563": 19, "4385": [19, 21], "3699": 19, "7614": 19, "3102": 19, "2152": [19, 20, 21], "2120": 19, "4432": 19, "0805": 19, "5568": 19, "6898": 19, "4106": 19, "2284": 19, "3480": 19, "8723": 19, "sigmoidbackward0": 19, "shifteduint8actpertensorfloat": [19, 22], "neg": 19, "effect": [19, 20, 22], "shifted_quant_ident": 19, "5854": 19, "5485": 19, "5099": 19, "1888": 19, "4532": 19, "3219": 19, "0772": 19, "6996": 19, "3794": 19, "0189": [19, 20], "6232": 19, "6421": 19, "1708": 19, "4343": [19, 22], "2266": 19, "7931": 19, "9262": 19, "relubackward0": 19, "129": 19, "swherebackward0": 19, "common": [19, 21], "3134": 19, "2557": 19, "0392": 19, "4186": 19, "7361": 19, "5340": 19, "8516": 19, "2887": 19, "3175": 19, "8949": 19, "6743": 19, "0722": 19, "0289": 19, "2021": 19, "4907": 19, "14": [19, 20, 22], "default_quant_relu": 19, "3078": [19, 20], "2555": 19, "0397": 19, "4185": 19, "7454": 19, "5427": 19, "8566": 19, "2943": 19, "3269": 19, "8893": 19, "6674": 19, "0785": 19, "0065": [19, 20], "0262": 19, "1962": 19, "4839": 19, "close": 19, "half": 19, "numer": 19, "lost": [19, 22], "treat": 19, "wast": 19, "regard": 19, "premad": 19, "word": 19, "caution": 19, "anticip": 19, "theme": 19, "interact": [19, 22], "30": [19, 22], "basic": [19, 21, 22], "calibr": 19, "These": [19, 22], "exponenti": [19, 22], "inp1": 19, "inp2": 19, "ones": 19, "out1_train": 19, "out2_train": 19, "out1_ev": 19, "out2_ev": 19, "quanthardtanh": 19, "hardtanh": 19, "8145d2f87fcb": 19, "117": [19, 21], "118": [19, 21], "119": [19, 21], "proxi": [19, 21, 22], "int8actpertensorfloatminmaxinit": 19, "scaling_init_impl": 19, "concer": 19, "quant_hard_tanh": 19, "remind": [19, 20], "legal": [19, 21], "encourag": 19, "drop": 20, "further": 20, "upstream": 20, "input_s": 20, "hidden_s": 20, "num_lay": 20, "nonlinear": 20, "tanh": [20, 21], "batch_first": 20, "bidirect": 20, "int8weightpertensorfloat": [20, 21, 22], "io_quant": 20, "gate_acc_qu": 20, "shared_input_hidden_weight": 20, "layer_impl": 20, "_quantrnnlay": 20, "gate": 20, "fuse": [20, 22], "fed": 20, "stack": 20, "potenti": 20, "nest": [20, 22], "rnn_sublay": 20, "sublayer_numb": 20, "right_to_left_direct": 20, "quant_rnn": 20, "quant_rnn_0_left_to_right": 20, "quant_rnn_0_right_to_left": 20, "quant_rnn_1_left_to_right": 20, "quant_rnn_1_right_to_left": 20, "gate_param": 20, "input_weight": 20, "cell": 20, "bitwidth": 20, "deepspeech": 20, "dai": 20, "qualiti": 20, "count_weight": 20, "numel": 20, "named_paramet": 20, "quant_rnn_single_direct": 20, "quant_rnn_bidirect": 20, "quant_rnn_bidirectional_shared_input_hidden": 20, "print": [20, 21, 22], "600": 20, "1200": 20, "4b": [20, 22], "6b": 20, "io": [20, 21], "quant_rnn_4b": 20, "io_bit_width": 20, "quant_rnn_4b_0_left_to_right": 20, "input_hidden_weight": 20, "hidden_hidden_weight": 20, "hidden_weight": 20, "o": 20, "0316": 20, "0317": 20, "0319": 20, "0318": 20, "0314": 20, "0298": 20, "0285": 20, "0306": 20, "0312": [20, 21], "0315": 20, "0293": 20, "0310": 20, "0309": 20, "pack": 20, "length": 20, "unbatch": 20, "sequenc": 20, "hidden_st": 20, "num_direct": 20, "343": 20, "__torch_function__": [20, 21, 22], "plain": 20, "classmethod": 20, "cb": 20, "pytorch_1000000000000": 20, "python_arg_pars": 20, "seq_dim": 20, "4458": [20, 21], "1651": 20, "7045": 20, "5889": 20, "2532": 20, "0330": 20, "1706": 20, "1376": [20, 21], "4348": 20, "5834": 20, "3577": 20, "2807": 20, "1046": [20, 21], "4293": 20, "1486": 20, "1569": 20, "3530": 20, "6995": 20, "0458": 20, "5295": 20, "3007": [20, 21], "7257": 20, "2877": 20, "1308": 20, "6603": 20, "0196": 20, "8237": 20, "4380": 20, "2615": 20, "3138": 20, "0850": 20, "1961": 20, "1929": 20, "5981": 20, "2508": 20, "2251": 20, "5917": 20, "0257": 20, "3023": 20, "2830": 20, "3344": 20, "4309": 20, "0836": 20, "2701": 20, "3666": 20, "1351": 20, "1736": 20, "1286": 20, "6174": 20, "4682": [20, 21], "1804": 20, "2780": 20, "4974": 20, "4389": 20, "0585": 20, "6242": 20, "0098": 20, "2341": 20, "3511": 20, "2926": 20, "4925": 20, "1414": 20, "4633": 20, "0683": 20, "2633": 20, "3024": 20, "1951": 20, "1707": 20, "0852": 20, "0965": 20, "4656": 20, "3180": 20, "3464": 20, "2782": 20, "1931": 20, "6360": 20, "3293": 20, "7211": 20, "4316": 20, "4145": 20, "3066": 20, "5224": [20, 22], "5849": 20, "1420": 20, "5669": 20, "2367": 20, "3027": 20, "3137": 20, "3632": 20, "5999": 20, "2036": 20, "2201": 20, "2862": 20, "3908": 20, "2091": 20, "2697": 20, "0055": [20, 21], "1761": 20, "1242": 20, "4184": 20, "6472": [20, 22], "4707": 20, "5034": [20, 21], "8368": 20, "1504": 20, "0654": 20, "7714": 20, "4903": 20, "6015": 20, "3596": 20, "2484": 20, "2942": 20, "3409": [20, 21], "8168": 20, "7396": 20, "2958": 20, "7782": 20, "1994": 20, "7846": 20, "3087": 20, "1029": 20, "1479": 20, "3216": 20, "2315": 20, "5209": 20, "0878": 20, "0390": 20, "1365": [20, 21], "2243": 20, "2390": 20, "3706": 20, "1609": 20, "5511": 20, "4096": 20, "5121": 20, "5901": 20, "3609": 20, "3755": 20, "0780": 20, "2829": 20, "1987": 20, "0057": 20, "1306": 20, "3861": 20, "2839": 20, "5962": 20, "1647": 20, "0227": 20, "4372": 20, "3748": 20, "1703": 20, "0738": [20, 21], "catbackward0": 20, "0058": 20, "unsqueezebackward0": 20, "1760": 20, "2670": 20, "1214": 20, "3702": 20, "3884": 20, "4127": [20, 21], "0243": 20, "0425": 20, "2246": 20, "0910": 20, "4734": 20, "0971": 20, "3824": 20, "1396": 20, "6858": 20, "0061": [20, 22], "1275": 20, "5037": 20, "2831": 20, "0566": 20, "2661": [20, 21], "0793": 20, "4926": 20, "0510": 20, "6455": 20, "7191": 20, "1812": 20, "6172": 20, "1529": 20, "4077": 20, "7078": 20, "0453": 20, "0963": 20, "4983": 20, "3977": 20, "0947": 20, "1894": 20, "3725": 20, "2589": 20, "3914": 20, "0063": [20, 21], "2652": 20, "5177": 20, "4230": 20, "0821": 20, "0631": 20, "0505": 20, "0253": [20, 22], "1578": 20, "4988": 20, "5556": 20, "4809": 20, "8144": 20, "6925": 20, "4360": 20, "0256": 20, "5130": 20, "2501": 20, "1347": 20, "7631": 20, "5386": 20, "2437": 20, "4296": 20, "1988": 20, "7246": 20, "1154": 20, "0641": 20, "3142": 20, "0706": 20, "0192": 20, "7185": 20, "8211": 20, "5709": 20, "1155": 20, "4683": 20, "3400": 20, "3015": 20, "3528": 20, "3143": 20, "1411": 20, "2309": 20, "5132": 20, "3721": 20, "5196": 20, "5453": 20, "4066": 20, "7768": 20, "6008": 20, "0546": 20, "0182": [20, 22], "1821": 20, "3763": 20, "3520": 20, "0486": 20, "2124": 20, "3641": 20, "4248": 20, "0789": 20, "2321": 20, "1982": 20, "1302": 20, "0283": 20, "4869": 20, "5379": 20, "6964": 20, "0340": 20, "2944": 20, "4643": 20, "3454": 20, "3284": 20, "3341": 20, "5945": 20, "2020": 20, "0379": 20, "8081": 20, "7260": 20, "0694": 20, "5430": 20, "8018": 20, "2273": 20, "3472": 20, "4924": 20, "4735": 20, "5745": 20, "5619": 20, "6313": 20, "1768": 20, "6541": [20, 22], "0385": [20, 21], "5835": 20, "0449": [20, 21], "3270": 20, "7951": 20, "3591": 20, "2757": 20, "7567": 20, "5194": 20, "7438": 20, "7695": 20, "5451": 20, "2630": 20, "4747": 20, "2245": 20, "3336": 20, "4490": 20, "4619": 20, "1796": 20, "3913": 20, "2053": 20, "2823": [20, 22], "6992": 20, "6607": 20, "1989": 20, "6928": 20, "5581": 20, "5966": 20, "0062": 20, "0064": 20, "1984": 20, "2499": [20, 21], "1102": [20, 21], "0955": 20, "4630": 20, "8672": 20, "1911": 20, "4851": 20, "6982": 20, "5806": 20, "4189": 20, "7423": 20, "9260": 20, "0147": 20, "0514": 20, "2167": 20, "5092": 20, "3846": 20, "0650": [20, 21], "6717": 20, "2492": 20, "0867": 20, "3900": 20, "3521": 20, "4767": 20, "1137": 20, "6879": [20, 21], "1733": 20, "0596": 20, "4279": 20, "5471": 20, "2762": 20, "5904": 20, "3737": 20, "1335": 20, "0140": [20, 22], "2810": 20, "5339": 20, "0562": 20, "7236": 20, "1264": 20, "0211": 20, "3021": 20, "1124": 20, "4777": 20, "3793": [20, 22], "2388": [20, 21], "0702": 20, "4847": 20, "3340": 20, "5225": 20, "1499": 20, "3083": 20, "1756": 20, "1713": 20, "3512": 20, "3041": 20, "3126": 20, "5482": 20, "4882": 20, "1028": 20, "4796": 20, "3640": 20, "0471": 20, "4438": 20, "2686": 20, "3095": 20, "2978": 20, "0993": 20, "0584": 20, "4846": 20, "0526": 20, "4496": 20, "1109": 20, "7416": 20, "3445": 20, "4963": 20, "2803": 20, "1927": 20, "6131": 20, "0661": 20, "3013": 20, "2646": 20, "6321": [20, 21], "4557": 20, "0294": 20, "9407": 20, "7350": 20, "6027": 20, "4116": 20, "6835": 20, "1787": 20, "0271": 20, "1354": 20, "3033": 20, "6229": 20, "3250": 20, "0812": [20, 21], "5633": 20, "0325": 20, "2383": 20, "5850": 20, "6771": 20, "3196": 20, "3934": 20, "3231": 20, "0492": 20, "5128": 20, "8149": 20, "7517": 20, "8711": 20, "4004": 20, "8992": 20, "2178": 20, "8851": 20, "5760": 20, "1054": 20, "5198": 20, "2612": 20, "2570": 20, "1542": 20, "1071": 20, "3854": 20, "0685": 20, "0728": 20, "4240": 20, "1627": 20, "3383": 20, "0428": 20, "1199": 20, "3683": 20, "3298": 20, "4204": 20, "2452": 20, "0934": [20, 21], "2336": 20, "1285": [20, 22], "2044": 20, "0701": 20, "3971": 20, "0175": 20, "1810": 20, "5547": 20, "0467": [20, 21], "0060": [20, 21, 22], "simul": [20, 22], "untrain": 20, "float_rnn": 20, "recov": 20, "123456": 20, "prebuilt": 20, "signedbinaryweightpertensorconst": [20, 22], "binary_rnn": 20, "3684": 20, "0946": 20, "4480": 20, "0050": [20, 21], "1543": 20, "6322": 20, "1643": 20, "1693": 20, "2937": 20, "5227": 20, "2290": 20, "3534": 20, "3883": 20, "4331": 20, "3634": 20, "1941": 20, "2240": 20, "0199": 20, "3485": 20, "1145": 20, "4082": 20, "2987": 20, "0647": 20, "0498": 20, "1493": 20, "0299": 20, "1195": 20, "0776": 20, "5670": 20, "4178": 20, "0239": 20, "4476": 20, "2029": 20, "7042": 20, "6326": 20, "4058": 20, "4118": 20, "0477": 20, "2387": 20, "0179": 20, "4416": 20, "4237": 20, "3282": 20, "1074": 20, "2626": 20, "3581": [20, 21], "2328": 20, "2268": 20, "3103": 20, "4536": 20, "3461": 20, "3163": 20, "7639": 20, "5252": 20, "1790": 20, "2984": 20, "5411": 20, "3147": 20, "6184": 20, "3037": 20, "1877": 20, "1767": 20, "1491": 20, "1049": [20, 22], "2871": 20, "0552": 20, "0883": 20, "0331": 20, "4749": 20, "7013": 20, "2264": 20, "0773": 20, "4583": 20, "0166": 20, "5743": 20, "1160": 20, "0442": 20, "1325": 20, "1657": 20, "0718": 20, "1215": 20, "6240": 20, "3092": 20, "0627": 20, "1882": 20, "4642": 20, "1443": 20, "4705": 20, "2447": 20, "1129": 20, "3011": 20, "2572": 20, "2384": 20, "0376": 20, "1380": 20, "0251": 20, "6399": 20, "5771": 20, "2133": 20, "7967": 20, "1631": 20, "4078": 20, "3199": 20, "0753": 20, "6524": [20, 22], "0690": 20, "1819": 20, "2258": 20, "3889": 20, "3764": 20, "5458": 20, "5704": 20, "6139": 20, "1209": 20, "5173": 20, "4447": 20, "0048": [20, 21], "3481": 20, "5946": 20, "5221": 20, "1644": 20, "2949": 20, "1789": 20, "2707": 20, "2900": 20, "5124": 20, "4399": 20, "0725": 20, "6091": 20, "0435": 20, "2030": 20, "2659": 20, "1547": 20, "0580": 20, "4254": 20, "5559": 20, "1740": 20, "4592": 20, "2369": 20, "3046": 20, "3626": 20, "2079": 20, "4641": 20, "sigmoid_qu": 20, "tanh_quant": 20, "cell_state_qu": 20, "coupled_input_forget_g": 20, "cat_output_cell_st": 20, "shared_intra_layer_weight_qu": 20, "shared_intra_layer_gate_acc_qu": 20, "shared_cell_state_qu": 20, "_quantlstmlay": 20, "cifg": 20, "ourselv": [20, 21, 22], "concen": 20, "illeg": 20, "sigmoid": 20, "forget": 20, "forget_g": 20, "input_g": 20, "orthogon": 20, "wors": 20, "calibration_mod": [20, 22], "bias_correction_mod": [20, 22], "progress": 20, "proof": 20, "quantizelinear": [20, 22], "dequantizelinear": [20, 22], "qeight": 20, "quant_lstm_weight_onli": 20, "quant_lstm_weight_only_4b": 20, "8080": 20, "np": 20, "np_input": 20, "astyp": 20, "seq_len": 20, "pred_onnx": 20, "quant_lstm_weight_only_cifg": 20, "quant_lstm_weight_only_cifg_4b": 20, "24": [7, 20, 22], "quant_lstm_weight_only_bidirectional_2_lay": 20, "home": [20, 21], "giusepp": 20, "quant_lstm_weight_only_bidirectional_2_layers_shar": 20, "quant_lstm_weight_only_bidirectional_2_layers_shared_ih": 20, "zp": 20, "quant_lstm_weight_only_bidirectional_2_layers_shared_q": 20, "8086": 20, "side": 20, "quant_lstm_weight_only_bidirectional_2_layers_shared_q_ih": 20, "8087": 20, "quantlstmcel": 20, "export_qonnx": [20, 22], "quant_lstm": 20, "8088": 20, "quant_hidden_st": 20, "quant_cell_st": 20, "quant_weight_ii": 20, "quant_weight_if": 20, "quant_weight_": 20, "quant_weight_io": 20, "quant_weight_hi": 20, "quant_weight_hf": 20, "quant_weight_hc": 20, "quant_weight_ho": 20, "quant_bias_input": 20, "quant_bias_forget": 20, "quant_bias_cel": 20, "quant_bias_output": 20, "output_scal": [20, 21, 22], "output_zero_point": 20, "cell_state_scal": 20, "cell_state_zero_point": 20, "cell_state_bit_width": 20, "input_acc_scal": 20, "input_acc_zero_point": 20, "input_acc_bit_width": 20, "forget_acc_scal": 20, "forget_acc_zero_point": 20, "forget_acc_bit_width": 20, "cell_acc_scal": 20, "cell_acc_zero_point": 20, "cell_acc_bit_width": 20, "output_acc_scal": 20, "output_acc_zero_point": 20, "output_acc_bit_width": 20, "input_sigmoid_scal": 20, "input_sigmoid_zero_point": 20, "input_sigmoid_bit_width": 20, "forget_sigmoid_scal": 20, "forget_sigmoid_zero_point": 20, "forget_sigmoid_bit_width": 20, "cell_tanh_scal": 20, "cell_tanh_zero_point": 20, "cell_tanh_bit_width": 20, "output_sigmoid_scal": 20, "output_sigmoid_zero_point": 20, "output_sigmoid_bit_width": 20, "hidden_state_tanh_scal": 20, "hidden_state_tanh_zero_point": 20, "hidden_state_tanh_bit_width": 20, "quantweightbiasinputoutputlay": [21, 22], "notic": [21, 22], "local": [21, 22], "python3": 21, "tqdm": 21, "tqdmwarn": 21, "iprogress": 21, "jupyt": 21, "ipywidget": 21, "readthedoc": 21, "en": 21, "stabl": [21, 22], "user_instal": 21, "html": 21, "autonotebook": 21, "notebook_tqdm": 21, "padding_typ": 21, "weightquanttyp": [21, 22], "biasquanttyp": [21, 22], "actquanttyp": [21, 22], "padding_mod": 21, "Its": 21, "intercept": [21, 22], "dedic": 21, "is_weight_quant_en": 21, "is_bias_quant_en": 21, "is_input_quant_en": 21, "is_output_quant_en": 21, "convolut": [21, 22], "2594": 21, "5392": 21, "5916": 21, "3493": 21, "6813": 21, "3732": 21, "1229": 21, "0084": [21, 22], "0031": 21, "1702": 21, "1069": 21, "8181": 21, "8056": 21, "4738": 21, "0589": 21, "1278": 21, "1718": 21, "1162": 21, "1526": 21, "9903": 21, "3541": 21, "1645": 21, "0557": 21, "2080": 21, "thnnconv2dbackward0": 21, "unquant": [21, 22], "everywher": 21, "seed": 21, "reproduc": 21, "disabled_quant_conv": 21, "freedom": 21, "experi": [21, 22], "littl": 21, "formula": 21, "integer_valu": 21, "quant_conv": 21, "short": 21, "0790": 21, "0503": 21, "1149": 21, "1903": 21, "1329": 21, "1813": 21, "0108": 21, "0593": 21, "0970": 21, "0215": 21, "0144": 21, "2280": 21, "1239": 21, "0090": 21, "1957": 21, "2011": 21, "0018": 21, "1993": 21, "0359": 21, "1778": 21, "1400": 21, "0916": 21, "1059": 21, "2173": 21, "1670": 21, "1939": 21, "2191": 21, "1688": 21, "1383": 21, "1185": 21, "1742": 21, "0808": 21, "1652": 21, "0233": 21, "0485": 21, "1418": 21, "1077": 21, "0036": 21, "1508": 21, "1616": 21, "0287": 21, "int_weight": 21, "quant_weight_zero_point": 21, "quant_weight_manu": 21, "popul": 21, "field": 21, "account": 21, "rel": 21, "expens": 21, "sparingli": 21, "quant_act": 21, "out_tensor_0": 21, "out_tensor_1": 21, "0173": 21, "0307": 21, "evalu": [21, 22], "9489": 21, "9111": 21, "0536": 21, "5788": 21, "3645": 21, "3401": 21, "4325": 21, "6498": 21, "6411": 21, "4390": 21, "9029": 21, "7012": [21, 22], "1591": 21, "9235": 21, "5883": 21, "7258": 21, "5330": 21, "9165": 21, "0820": 21, "4148": 21, "3651": 21, "0164": 21, "9567": 21, "2758": 21, "2414": 21, "2111": 21, "9124": 21, "3814": 21, "8805": 21, "3191": 21, "8965": 21, "2048": 21, "8113": 21, "1142": 21, "3381": 21, "2238": 21, "0068": 21, "2567": 21, "0731": 21, "4280": 21, "0909": 21, "0875": 21, "6851": 21, "7744": 21, "8143": 21, "3557": 21, "2802": 21, "addbackward0": 21, "0240": 21, "consequ": 21, "longer": 21, "handl": 21, "5800": 21, "0157": 21, "4445": 21, "8577": 21, "5643": 21, "0383": 21, "9028": 21, "5191": 21, "6546": 21, "1442": 21, "5868": 21, "maxpool2dwithindicesbackward0": 21, "0226": 21, "decai": [21, 22], "4943": 21, "9938": 21, "9073": 21, "7681": 21, "3262": 21, "9186": 21, "1786": 21, "3659": 21, "7489": 21, "8946": 21, "0451": 21, "5594": 21, "1346": 21, "4770": 21, "6951": 21, "0676": 21, "5111": 21, "8459": 21, "8990": 21, "9426": 21, "7945": 21, "9220": 21, "7772": 21, "7177": 21, "4414": 21, "2220": 21, "5747": 21, "6710": 21, "4594": 21, "3462": 21, "9729": 21, "5896": 21, "5276": 21, "0900": 21, "8852": 21, "tanhbackward0": 21, "input_quant_conv": 21, "9693": 21, "9431": 21, "2459": 21, "5416": 21, "9037": 21, "5278": 21, "6207": 21, "3578": 21, "4815": 21, "4551": 21, "4065": 21, "8889": 21, "3393": 21, "0803": 21, "1748": 21, "0977": 21, "6284": 21, "7193": [21, 22], "7626": 21, "2634": 21, "3453": 21, "3349": 21, "1923": 21, "5993": 21, "9579": 21, "2208e": 21, "05": [21, 22], "mimpli": 21, "reflect": [21, 22], "worst": 21, "largest": 21, "int_valu": 21, "randint": 21, "quant_tensor_input": 21, "7000e": 21, "03": 21, "5000e": 21, "2400e": 21, "2000e": 21, "3000e": 21, "0000e": 21, "2700e": 21, "9000e": 21, "6000e": 21, "4000e": 21, "1100e": 21, "1500e": 21, "8000e": 21, "0600e": 21, "1000e": 21, "1300e": 21, "0100e": 21, "1900e": 21, "113": 21, "return_quant_conv": 21, "114": [21, 22], "0085": 21, "0066": 21, "0038": 21, "0115": 21, "0037": 21, "0015": [21, 22], "0027": 21, "0079": 21, "0034": 21, "0043": 21, "0008": 21, "0052": 21, "0033": 21, "0082": 21, "0021": [21, 22], "0004": 21, "0054": 21, "0013": [21, 22], "8448e": 21, "07": 21, "115": 21, "116": 21, "0035": 21, "0051": 21, "0047": 21, "0017": [21, 22], "0028": 21, "0044": [21, 22], "0024": 21, "0011": 21, "0039": 21, "7410e": 21, "4060": 21, "3654": 21, "7876": 21, "8119": 21, "9825": 21, "5115": 21, "3979": 21, "3248": 21, "3816": 21, "0568": 21, "8038": 21, "3491": 21, "4141": 21, "5846": 21, "4222": 21, "7389": 21, "2517": 21, "1624": 21, "7308": 21, "0081": 21, "got": 21, "saw": [21, 22], "assum": [7, 21, 22], "somehow": [21, 22], "predefin": [21, 22], "tri": [21, 22], "bias_quant_conv": 21, "ipykernel_48365": 21, "2280634207": 21, "opt": 21, "conda": 21, "torch_1": 21, "_call_impl": [21, 22], "1100": 21, "_backward_hook": [21, 22], "_forward_hook": [21, 22], "_forward_pre_hook": [21, 22], "_global_backward_hook": [21, 22], "1101": 21, "_global_forward_hook": [21, 22], "_global_forward_pre_hook": [21, 22], "forward_cal": [21, 22], "1103": 21, "1104": 21, "full_backward_hook": [21, 22], "non_full_backward_hook": [21, 22], "workspac": 21, "fork_brevita": 21, "190": 21, "191": 21, "192": 21, "forward_impl": [21, 22], "193": 21, "194": [21, 22], "inner_forward_impl": [21, 22], "330": 21, "331": 21, "cache_inference_quant_bia": [21, 22], "_cached_bia": [21, 22], "_cachedio": [21, 22], "metadata_onli": [21, 22], "parameter_qu": [21, 22], "impl": [21, 22], "requires_input_scal": [21, 22], "162": 21, "163": 21, "requires_input_bit_width": [21, 22], "164": 21, "0005": 21, "0106": 21, "0012": 21, "0007": 21, "0067": 21, "0059": 21, "0071": 21, "8108e": 21, "input_bias_quant_conv": 21, "3825": 21, "1371": 21, "9135": 21, "2016": 21, "7495": 21, "4071": 21, "0755": 21, "5283": 21, "0788": 21, "3802": 21, "2234": 21, "8678": 21, "5546": 21, "4408": 21, "6788": 21, "4422": 21, "4412": 21, "3205": 21, "0083": 21, "3295": 21, "2076": 21, "4417": 21, "8610e": 21, "0080": 21, "0014": 21, "0029": 21, "0003": 21, "0019": 21, "0056": 21, "0091": 21, "0095": 21, "8384e": 21, "give": 21, "output_bias_quant_conv": 21, "2990591641": 21, "int8biaspertensorfloatinternalsc": 21, "bias_internal_scale_quant_conv": 21, "8346": 21, "0746": 21, "5212": 21, "1019": 21, "6004": 21, "1500": 21, "1453": 21, "1551": 21, "3458": 21, "1312": 21, "2502": 21, "5267": 21, "2412": 21, "3556": 21, "3289": 21, "2276": 21, "4599": 21, "6094": 21, "5064": 21, "6768": 21, "6638": 21, "2359": 21, "lead": 21, "125": 21, "unquant_bias_input_quant_conv": 21, "6632": 21, "2411": 21, "2064": 21, "7371": 21, "3910": 21, "9533": 21, "2994": 21, "4684": 21, "4495": 21, "5021": 21, "5738": 21, "4199": 21, "3380": 21, "6218": 21, "0408": 21, "8483": 21, "5625": [21, 22], "1837": 21, "0575": 21, "2816": [21, 22], "4993": 21, "4556": 21, "4269": 21, "5369": 21, "0975e": 21, "1276": 21, "0774": 21, "3152": 21, "4585": 21, "7320": 21, "2324": 21, "subject": [21, 22], "retur": 21, "bias_input_quant_conv": 21, "8357": 21, "0733": 21, "9527": 21, "1803": 21, "2154": 21, "7598": 21, "1121": 21, "8728": 21, "7917": 21, "6516": 21, "1852": 21, "7263": 21, "0956": 21, "2747": 21, "1617": 21, "8299": 21, "9934": 21, "3821": 21, "4865": 21, "9309": 21, "7924": 21, "4201": 21, "2343": 21, "1532": 21, "github": 22, "tree": 22, "master": 22, "in_featur": 22, "out_featur": 22, "quant_linear": 22, "0053": 22, "5820": 22, "5204": 22, "2723": 22, "1896": 22, "5607": 22, "0046": 22, "3803": 22, "2704": 22, "1879": 22, "0137": 22, "5591": 22, "absolut": 22, "full": 22, "corrispond": 22, "83": 22, "59": 22, "41": 22, "float_input": 22, "float_output": 22, "5410": 22, "2934": 22, "1788": 22, "5684": 22, "0845": 22, "3986": 22, "9036": 22, "4586": 22, "3096": 22, "2058": 22, "6525": 22, "3723": 22, "8677": 22, "3873": 22, "2801": 22, "9009": 22, "9507": 22, "mmbackward": 22, "made": 22, "int8weightpertensorfixedpoint": 22, "0078": 22, "3828": 22, "5781": 22, "5234": 22, "2734": 22, "0156": 22, "handi": 22, "overal": 22, "quant_linear1": 22, "quant_linear1_scale_before_shar": 22, "quant_linear2": 22, "4f": 22, "9109": 22, "4609": 22, "3135": 22, "6523": 22, "2089": 22, "3752": 22, "8697": 22, "3893": 22, "9011": 22, "9521": 22, "0542e": 22, "5490": 22, "2894": 22, "5617": 22, "0894": 22, "3958": 22, "0170": 22, "quant_relu": 22, "5681": 22, "signed_quant_output": 22, "unsigned_quant_output": 22, "5588": 22, "999": 22, "percentil": 22, "300": 22, "proper": 22, "int16bia": 22, "alessa": 22, "appdata": 22, "temp": 22, "ipykernel_18920": 22, "2660651517": 22, "pt190": 22, "1050": 22, "1051": 22, "1052": 22, "1053": 22, "documenti": 22, "brevitas_tvmcon": 22, "96": 22, "97": 22, "356": 22, "357": 22, "358": 22, "359": 22, "195": 22, "196": 22, "197": 22, "198": 22, "1263": 22, "1680": 22, "1231": 22, "4658": 22, "2395": 22, "5207": 22, "3989": 22, "6461": 22, "8687": 22, "0466": 22, "4813": 22, "addmmbackward": 22, "constraint": 22, "operand": 22, "allign": 22, "easi": 22, "float_inp1": 22, "float_inp2": 22, "train_quant_inp1": 22, "train_quant_inp2": 22, "train_mode_add": 22, "ema": 22, "eval_quant_inp1": 22, "eval_quant_inp2": 22, "eval_mode_add": 22, "5335": 22, "2875": 22, "0447": 22, "5751": 22, "0863": 22, "4057": 22, "0160": 22, "3994": 22, "8307": 22, "7188": 22, "5910": 22, "1757": 22, "9329": 22, "5431": 22, "7636": 22, "6773": 22, "2300": 22, "input_dequant_valu": 22, "input_integer_valu": 22, "output_integer_valu": 22, "max_pool1d": 22, "float_inp": 22, "1218": 22, "1580": 22, "2533": 22, "8504": 22, "6876": 22, "3076": 22, "1170": 22, "4704": 22, "1628": 22, "4475": 22, "2714": 22, "8685": 22, "1448": 22, "1086": 22, "9228": 22, "2666": 22, "0543": 22, "6152": 22, "4162": 22, "8323": 22, "3160": 22, "0181": 22, "squeezebackward1": 22, "652": 22, "experiment": [1, 22, 24], "c10": 22, "tensorimpl": 22, "h": 22, "1156": 22, "ceil_mod": 22, "8204": 22, "2480": 22, "4089": 22, "6913": 22, "5964": 22, "2983": 22, "9714": 22, "4386": 22, "1614": 22, "8952": 22, "2649": 22, "7006": 22, "1438": 22, "1081": 22, "7272": 22, "8529": 22, "9646": 22, "0542": 22, "5478": 22, "3937": 22, "6817": 22, "9807": 22, "tanhbackward": 22, "simiarli": 22, "train_mode_cat": 22, "eval_mode_cat": 22, "concat": 22, "easiest": 22, "3880": 22, "5044": 22, "2716": 22, "1940": 22, "5432": 22, "0388": 22, "1816": 22, "0374": 22, "6341": 22, "5447": 22, "uint8actpertensorfloatmaxinit": 22, "5294": 22, "5647": 22, "0235": 22, "static": 22, "depthwis": 22, "per_channel_depthwise_quant_conv": 22, "input_scaling_per_output_channel": 22, "input_scaling_stats_permute_dim": 22, "permut": 22, "input_per_channel_broadcastable_shap": 22, "4033": 22, "8380": 22, "8616": 22, "4503": 22, "4937": 22, "1901": 22, "compact": 22, "usabl": 22, "perchannel3bactqu": 22, "driven": 22, "solver": 22, "actquantsolv": 22, "float_to_int_impl_typ": 22, "floattointimpltyp": 22, "nearest": 22, "scaling_stats_op": 22, "statsop": 22, "restrict_scaling_typ": 22, "restrictvaluetyp": 22, "overriden": 22, "tweak": 22, "8b": 22, "learnedintweightperchannelfloat": 22, "log_fp": 22, "1887": 22, "0132": 22, "0030": 22, "backpropag": 22, "loss": 22, "regular": 22, "push": 22, "larger": 22, "learnedintactpertensorfloat": 22, "4588": 22, "3119": 22, "6530": 22, "6493": 22, "3731": 22, "8706": 22, "8979": 22, "9543": 22, "9068e": 22, "6866e": 22, "4251e": 22, "didn": 22, "29": 22, "float_linear": 22, "1653109852": 22, "fused_activation_quant_proxi": 22, "bit_width_offset": 22, "supress": 22, "accomod": 22, "basi": 22, "0023": 22, "question": 22, "underneath": 22, "immedi": 22, "_zero_hw_sentinel": 22, "fusedactivationquantproxi": 22, "activation_impl": 22, "restrict_clamp_sc": 22, "_restrictclampvalu": 22, "deped": 22, "procedur": 22, "restrict_v": 22, "modular": 22, "degre": 22, "overhead": 22, "scatter": 22, "int8actpertensorfloatfromscratch": 22, "dequant_valu": 22, "convini": 22, "engin": 22, "\u00e8xtendedinjector": 22, "repeat": 22, "recurs": 22, "manner": 22, "indipend": 22, "fine": 22, "grain": 22, "int8actpertensorfloatparameterfromscratch": 22, "quant_identity_bit_width": 22, "quant_linear_bit_width": 22, "perfect": 22, "indipedent": 22, "satisfi": 22, "protobuf": 22, "six": 22, "38": 22, "32b": 22, "16b": 22, "quant_conv_4b8b": 22, "output_path": 22, "qop_onnx_conv_4b8b": 22, "input_t": 22, "relax": 22, "dialect": 22, "brevitas_onnx_conv4b8b": 22, "40": 22, "quant_conv_4b_weight": 22, "brevitas_onnx_conv_4b_weight": 22, "tvm": [22, 24], "7b": 22, "export_torch_qop": 22, "quant_conv_8b7b": 22, "pytorch_qf_conv_8b7b": 22, "pt": 22, "42": 22, "tracerwarn": 22, "trace": 22, "safe": 22, "caus": 22, "incorrect": 22, "incorpor": 22, "programmat": 22, "pattern": 22, "subsystem": 22, "transform": 22, "beyond": 22, "worth": 22, "embrac": 22, "backport": 22, "tracer": 22, "value_trac": 22, "condit": 22, "caveat": 22, "temporar": 22, "quant_model": 22, "calibrate_model": 22, "calibration_load": 22, "no_grad": 22, "imag": 22, "_": 22, "enumer": 22, "iter": 22, "pyxir": 24, "compliant": 24, "necessar": 24, "overoutputfeaturesview": [1, 3], "prezerocenterzeropoint": 1, "pre_zero_point_stats_input_view_shape_impl": 1, "pre_zero_point_shap": 1, "center": 1, "incom": 1, "intend": 1, "get_zero_cent": 1, "over_output_featur": [1, 3, 7], "a2q": 7}, "objects": {"brevitas": [[1, 0, 0, "-", "core"]], "brevitas.core": [[2, 0, 0, "-", "bit_width"], [3, 0, 0, "-", "function_wrapper"], [4, 0, 0, "-", "quant"], [1, 0, 0, "-", "restrict_val"], [5, 0, 0, "-", "scaling"], [6, 0, 0, "-", "stats"], [1, 0, 0, "-", "utils"], [1, 0, 0, "-", "zero_point"]], "brevitas.core.bit_width": [[2, 0, 0, "-", "const"], [2, 0, 0, "-", "parameter"]], "brevitas.core.bit_width.const": [[2, 1, 1, "", "BitWidthConst"], [2, 1, 1, "", "BitWidthStatefulConst"], [2, 1, 1, "", "MsbClampBitWidth"]], "brevitas.core.bit_width.const.BitWidthConst": [[2, 2, 1, "", "forward"]], "brevitas.core.bit_width.const.BitWidthStatefulConst": [[2, 2, 1, "", "forward"]], "brevitas.core.bit_width.const.MsbClampBitWidth": [[2, 2, 1, "", "forward"]], "brevitas.core.bit_width.parameter": [[2, 1, 1, "", "BitWidthParameter"], [2, 1, 1, "", "RemoveBitwidthParameter"]], "brevitas.core.bit_width.parameter.BitWidthParameter": [[2, 2, 1, "", "forward"]], "brevitas.core.bit_width.parameter.RemoveBitwidthParameter": [[2, 2, 1, "", "forward"]], "brevitas.core.function_wrapper": [[3, 0, 0, "-", "clamp"], [3, 0, 0, "-", "misc"], [3, 0, 0, "-", "ops_ste"], [3, 0, 0, "-", "shape"]], "brevitas.core.function_wrapper.clamp": [[3, 1, 1, "", "ClampMin"], [3, 1, 1, "", "ScalarClamp"], [3, 1, 1, "", "TensorClamp"]], "brevitas.core.function_wrapper.clamp.ClampMin": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.clamp.ScalarClamp": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.clamp.TensorClamp": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.misc": [[3, 1, 1, "", "Identity"], [3, 1, 1, "", "InplaceLogTwo"], [3, 1, 1, "", "LogTwo"], [3, 1, 1, "", "PowerOfTwo"]], "brevitas.core.function_wrapper.misc.Identity": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.misc.InplaceLogTwo": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.misc.LogTwo": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.misc.PowerOfTwo": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.ops_ste": [[3, 1, 1, "", "CeilSte"], [3, 1, 1, "", "DPURoundSte"], [3, 1, 1, "", "FloorSte"], [3, 1, 1, "", "InplaceTensorClampSte"], [3, 1, 1, "", "RoundSte"], [3, 1, 1, "", "RoundToZeroSte"], [3, 1, 1, "", "ScalarClampMinSte"], [3, 1, 1, "", "TensorClampSte"]], "brevitas.core.function_wrapper.ops_ste.CeilSte": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.ops_ste.DPURoundSte": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.ops_ste.FloorSte": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.ops_ste.InplaceTensorClampSte": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.ops_ste.RoundSte": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.ops_ste.RoundToZeroSte": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.ops_ste.ScalarClampMinSte": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.ops_ste.TensorClampSte": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.shape": [[3, 1, 1, "", "OverBatchOverOutputChannelView"], [3, 1, 1, "", "OverBatchOverTensorView"], [3, 1, 1, "", "OverOutputChannelView"], [3, 1, 1, "", "OverOutputFeaturesView"], [3, 1, 1, "", "OverTensorView"], [3, 1, 1, "", "PermuteDims"], [3, 1, 1, "", "StatsInputViewShapeImpl"]], "brevitas.core.function_wrapper.shape.OverBatchOverOutputChannelView": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.shape.OverBatchOverTensorView": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.shape.OverOutputChannelView": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.shape.OverOutputFeaturesView": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.shape.OverTensorView": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.shape.PermuteDims": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.shape.StatsInputViewShapeImpl": [[3, 3, 1, "", "OVER_BATCH_OVER_OUTPUT_CHANNELS"], [3, 3, 1, "", "OVER_BATCH_OVER_TENSOR"], [3, 3, 1, "", "OVER_OUTPUT_CHANNELS"], [3, 3, 1, "", "OVER_OUTPUT_FEATURES"], [3, 3, 1, "", "OVER_TENSOR"]], "brevitas.core.quant": [[4, 0, 0, "-", "binary"], [4, 0, 0, "-", "delay"], [4, 0, 0, "-", "int"], [4, 0, 0, "-", "int_base"], [4, 0, 0, "-", "ternary"]], "brevitas.core.quant.binary": [[4, 1, 1, "", "BinaryQuant"], [4, 1, 1, "", "ClampedBinaryQuant"]], "brevitas.core.quant.binary.BinaryQuant": [[4, 2, 1, "", "forward"]], "brevitas.core.quant.binary.ClampedBinaryQuant": [[4, 2, 1, "", "forward"]], "brevitas.core.quant.delay": [[4, 1, 1, "", "DelayWrapper"]], "brevitas.core.quant.delay.DelayWrapper": [[4, 2, 1, "", "forward"]], "brevitas.core.quant.int": [[4, 1, 1, "", "DecoupledRescalingIntQuant"], [4, 1, 1, "", "DecoupledRescalingIntQuantWithInput"], [4, 1, 1, "", "PrescaledRestrictIntQuant"], [4, 1, 1, "", "PrescaledRestrictIntQuantWithInputBitWidth"], [4, 1, 1, "", "RescalingIntQuant"], [4, 1, 1, "", "TruncIntQuant"]], "brevitas.core.quant.int.DecoupledRescalingIntQuant": [[4, 2, 1, "", "forward"]], "brevitas.core.quant.int.DecoupledRescalingIntQuantWithInput": [[4, 2, 1, "", "forward"]], "brevitas.core.quant.int.PrescaledRestrictIntQuant": [[4, 2, 1, "", "forward"]], "brevitas.core.quant.int.PrescaledRestrictIntQuantWithInputBitWidth": [[4, 2, 1, "", "forward"]], "brevitas.core.quant.int.RescalingIntQuant": [[4, 2, 1, "", "forward"]], "brevitas.core.quant.int.TruncIntQuant": [[4, 2, 1, "", "forward"]], "brevitas.core.quant.int_base": [[4, 1, 1, "", "DecoupledIntQuant"], [4, 1, 1, "", "IntQuant"]], "brevitas.core.quant.int_base.DecoupledIntQuant": [[4, 2, 1, "", "forward"], [4, 2, 1, "", "max_int"], [4, 2, 1, "", "min_int"], [4, 2, 1, "", "to_int"]], "brevitas.core.quant.int_base.IntQuant": [[4, 2, 1, "", "forward"], [4, 2, 1, "", "max_int"], [4, 2, 1, "", "min_int"], [4, 2, 1, "", "to_int"]], "brevitas.core.quant.ternary": [[4, 1, 1, "", "TernaryQuant"]], "brevitas.core.quant.ternary.TernaryQuant": [[4, 2, 1, "", "forward"]], "brevitas.core.restrict_val": [[1, 1, 1, "", "FloatRestrictValue"], [1, 1, 1, "", "IntRestrictValue"], [1, 1, 1, "", "LogFloatRestrictValue"], [1, 1, 1, "", "PowerOfTwoRestrictValue"]], "brevitas.core.restrict_val.FloatRestrictValue": [[1, 2, 1, "", "forward"], [1, 2, 1, "", "restrict_init_float"], [1, 2, 1, "", "restrict_init_inplace_module"], [1, 2, 1, "", "restrict_init_module"], [1, 2, 1, "", "restrict_init_tensor"]], "brevitas.core.restrict_val.IntRestrictValue": [[1, 2, 1, "", "forward"], [1, 2, 1, "", "restrict_init_float"], [1, 2, 1, "", "restrict_init_inplace_module"], [1, 2, 1, "", "restrict_init_module"], [1, 2, 1, "", "restrict_init_tensor"]], "brevitas.core.restrict_val.LogFloatRestrictValue": [[1, 2, 1, "", "forward"], [1, 2, 1, "", "restrict_init_float"], [1, 2, 1, "", "restrict_init_inplace_module"], [1, 2, 1, "", "restrict_init_module"], [1, 2, 1, "", "restrict_init_tensor"]], "brevitas.core.restrict_val.PowerOfTwoRestrictValue": [[1, 2, 1, "", "forward"], [1, 2, 1, "", "restrict_init_float"], [1, 2, 1, "", "restrict_init_inplace_module"], [1, 2, 1, "", "restrict_init_module"], [1, 2, 1, "", "restrict_init_tensor"]], "brevitas.core.scaling": [[5, 0, 0, "-", "int_scaling"], [5, 0, 0, "-", "runtime"], [5, 0, 0, "-", "standalone"]], "brevitas.core.scaling.int_scaling": [[5, 1, 1, "", "IntScaling"], [5, 1, 1, "", "PowerOfTwoIntScaling"]], "brevitas.core.scaling.int_scaling.IntScaling": [[5, 2, 1, "", "forward"]], "brevitas.core.scaling.int_scaling.PowerOfTwoIntScaling": [[5, 2, 1, "", "forward"]], "brevitas.core.scaling.runtime": [[5, 1, 1, "", "RuntimeStatsScaling"], [5, 1, 1, "", "StatsFromParameterScaling"]], "brevitas.core.scaling.runtime.RuntimeStatsScaling": [[5, 2, 1, "", "forward"]], "brevitas.core.scaling.runtime.StatsFromParameterScaling": [[5, 2, 1, "", "forward"]], "brevitas.core.scaling.standalone": [[5, 1, 1, "", "ConstScaling"], [5, 1, 1, "", "ParameterFromRuntimeStatsScaling"], [5, 1, 1, "", "ParameterFromStatsFromParameterScaling"], [5, 1, 1, "", "ParameterScaling"]], "brevitas.core.scaling.standalone.ConstScaling": [[5, 2, 1, "", "forward"]], "brevitas.core.scaling.standalone.ParameterFromRuntimeStatsScaling": [[5, 2, 1, "", "forward"], [5, 2, 1, "", "state_dict"], [5, 2, 1, "", "training_forward"]], "brevitas.core.scaling.standalone.ParameterFromStatsFromParameterScaling": [[5, 2, 1, "", "forward"], [5, 2, 1, "", "state_dict"]], "brevitas.core.scaling.standalone.ParameterScaling": [[5, 2, 1, "", "forward"]], "brevitas.core.stats": [[6, 0, 0, "-", "stats_op"], [6, 0, 0, "-", "stats_wrapper"], [6, 0, 0, "-", "view_wrapper"]], "brevitas.core.stats.stats_op": [[6, 1, 1, "", "AbsAve"], [6, 1, 1, "", "AbsMax"], [6, 1, 1, "", "AbsMaxAve"], [6, 1, 1, "", "AbsMaxL2"], [6, 1, 1, "", "AbsMinMax"], [6, 1, 1, "", "AbsPercentile"], [6, 1, 1, "", "KLMinimizerThreshold"], [6, 1, 1, "", "L1Norm"], [6, 1, 1, "", "L2Norm"], [6, 1, 1, "", "MSE"], [6, 1, 1, "", "MeanLearnedSigmaStd"], [6, 1, 1, "", "MeanSigmaStd"], [6, 1, 1, "", "NegativeMinOrZero"], [6, 1, 1, "", "NegativePercentileOrZero"], [6, 1, 1, "", "PercentileInterval"]], "brevitas.core.stats.stats_op.AbsAve": [[6, 2, 1, "", "forward"]], "brevitas.core.stats.stats_op.AbsMax": [[6, 2, 1, "", "forward"]], "brevitas.core.stats.stats_op.AbsMaxAve": [[6, 2, 1, "", "forward"]], "brevitas.core.stats.stats_op.AbsMaxL2": [[6, 2, 1, "", "forward"]], "brevitas.core.stats.stats_op.AbsMinMax": [[6, 2, 1, "", "forward"]], "brevitas.core.stats.stats_op.AbsPercentile": [[6, 2, 1, "", "forward"]], "brevitas.core.stats.stats_op.KLMinimizerThreshold": [[6, 2, 1, "", "forward"], [6, 2, 1, "", "smooth_normalize_distribution"]], "brevitas.core.stats.stats_op.L1Norm": [[6, 2, 1, "", "forward"]], "brevitas.core.stats.stats_op.L2Norm": [[6, 2, 1, "", "forward"]], "brevitas.core.stats.stats_op.MSE": [[6, 2, 1, "", "evaluate_loss"], [6, 2, 1, "", "forward"], [6, 2, 1, "", "mse_fib_search"], [6, 2, 1, "", "mse_grid_search"], [6, 2, 1, "", "mse_loss_fn"], [6, 2, 1, "", "mse_search"]], "brevitas.core.stats.stats_op.MeanLearnedSigmaStd": [[6, 2, 1, "", "forward"]], "brevitas.core.stats.stats_op.MeanSigmaStd": [[6, 2, 1, "", "forward"]], "brevitas.core.stats.stats_op.NegativeMinOrZero": [[6, 2, 1, "", "forward"]], "brevitas.core.stats.stats_op.NegativePercentileOrZero": [[6, 2, 1, "", "forward"]], "brevitas.core.stats.stats_op.PercentileInterval": [[6, 2, 1, "", "forward"]], "brevitas.core.utils": [[1, 1, 1, "", "ParameterWrapper"], [1, 1, 1, "", "SingleArgStatelessBuffer"], [1, 1, 1, "", "SliceTensor"], [1, 1, 1, "", "StatelessBuffer"], [1, 4, 1, "", "inplace_momentum_update"], [1, 4, 1, "", "inplace_tensor_add"], [1, 4, 1, "", "inplace_tensor_mul"]], "brevitas.core.utils.ParameterWrapper": [[1, 2, 1, "", "forward"]], "brevitas.core.utils.SingleArgStatelessBuffer": [[1, 2, 1, "", "forward"]], "brevitas.core.utils.SliceTensor": [[1, 2, 1, "", "eager_forward"], [1, 2, 1, "", "forward"]], "brevitas.core.utils.StatelessBuffer": [[1, 2, 1, "", "forward"], [1, 2, 1, "", "state_dict"]], "brevitas.core.zero_point": [[1, 1, 1, "", "ParameterFromRuntimeZeroPoint"], [1, 1, 1, "", "ParameterFromStatsFromParameterZeroPoint"], [1, 1, 1, "", "ParameterZeroPoint"], [1, 1, 1, "", "PreZeroCenterZeroPoint"], [1, 1, 1, "", "StatsFromParameterZeroPoint"], [1, 1, 1, "", "ZeroZeroPoint"]], "brevitas.core.zero_point.ParameterFromRuntimeZeroPoint": [[1, 2, 1, "", "forward"], [1, 2, 1, "", "state_dict"], [1, 2, 1, "", "training_forward"]], "brevitas.core.zero_point.ParameterFromStatsFromParameterZeroPoint": [[1, 2, 1, "", "forward"], [1, 2, 1, "", "state_dict"]], "brevitas.core.zero_point.ParameterZeroPoint": [[1, 2, 1, "", "forward"]], "brevitas.core.zero_point.PreZeroCenterZeroPoint": [[1, 2, 1, "", "forward"], [1, 2, 1, "", "get_zero_center"]], "brevitas.core.zero_point.StatsFromParameterZeroPoint": [[1, 2, 1, "", "forward"]], "brevitas.core.zero_point.ZeroZeroPoint": [[1, 2, 1, "", "forward"]], "brevitas.function": [[7, 0, 0, "-", "ops"], [7, 0, 0, "-", "ops_ste"], [7, 0, 0, "-", "shape"]], "brevitas.function.ops": [[7, 4, 1, "", "binary_sign"], [7, 4, 1, "", "dpu_round"], [7, 4, 1, "", "get_upper_bound_on_l1_norm"], [7, 4, 1, "", "identity"], [7, 4, 1, "", "max_float"], [7, 4, 1, "", "max_int"], [7, 4, 1, "", "min_int"], [7, 4, 1, "", "round_to_zero"], [7, 4, 1, "", "tensor_clamp"], [7, 4, 1, "", "tensor_clamp_"]], "brevitas.function.ops_ste": [[7, 4, 1, "", "abs_binary_sign_grad"], [7, 4, 1, "", "binary_sign_ste"], [7, 4, 1, "", "ceil_ste"], [7, 4, 1, "", "dpu_round_ste"], [7, 4, 1, "", "floor_ste"], [7, 4, 1, "", "round_ste"], [7, 4, 1, "", "round_to_zero_ste"], [7, 4, 1, "", "scalar_clamp_min_ste"], [7, 4, 1, "", "scalar_clamp_ste"], [7, 4, 1, "", "tensor_clamp_ste"], [7, 4, 1, "", "tensor_clamp_ste_"], [7, 4, 1, "", "ternary_sign_ste"]], "brevitas.function.shape": [[7, 4, 1, "", "over_batch_over_output_channels"], [7, 4, 1, "", "over_batch_over_tensor"], [7, 4, 1, "", "over_output_channels"], [7, 4, 1, "", "over_output_features"], [7, 4, 1, "", "over_tensor"]], "brevitas.ops": [[8, 0, 0, "-", "autograd_ste_ops"]], "brevitas.ops.autograd_ste_ops": [[8, 1, 1, "", "AbsBinarySignGradFn"], [8, 1, 1, "", "BinarySignSteFn"], [8, 1, 1, "", "CeilSteFn"], [8, 1, 1, "", "DPURoundSteFn"], [8, 1, 1, "", "FloorSteFn"], [8, 1, 1, "", "InplaceTensorClampSteFn"], [8, 1, 1, "", "RoundSteFn"], [8, 1, 1, "", "RoundToZeroSteFn"], [8, 1, 1, "", "ScalarClampMinSteFn"], [8, 1, 1, "", "ScalarClampSteFn"], [8, 1, 1, "", "TensorClampSteFn"], [8, 1, 1, "", "TernarySignSteFn"], [8, 4, 1, "", "abs_binary_sign_grad_impl"], [8, 4, 1, "", "binary_sign_ste_impl"], [8, 4, 1, "", "ceil_ste_impl"], [8, 4, 1, "", "dpu_round_ste_impl"], [8, 4, 1, "", "floor_ste_impl"], [8, 4, 1, "", "round_ste_impl"], [8, 4, 1, "", "round_to_zero_ste_impl"], [8, 4, 1, "", "scalar_clamp_min_ste_impl"], [8, 4, 1, "", "scalar_clamp_ste_impl"], [8, 4, 1, "", "tensor_clamp_ste_impl"], [8, 4, 1, "", "ternary_sign_ste_impl"]]}, "objtypes": {"0": "py:module", "1": "py:class", "2": "py:method", "3": "py:attribute", "4": "py:function"}, "objnames": {"0": ["py", "module", "Python module"], "1": ["py", "class", "Python class"], "2": ["py", "method", "Python method"], "3": ["py", "attribute", "Python attribute"], "4": ["py", "function", "Python function"]}, "titleterms": {"about": 0, "author": 0, "cite": 0, "brevita": [1, 2, 3, 4, 5, 6, 7, 8, 12, 13, 22], "core": [1, 2, 3, 4, 5, 6, 10], "packag": [1, 2, 3, 4, 5, 6, 7, 8], "subpackag": [1, 9], "submodul": [1, 2, 3, 4, 5, 6, 7, 8], "restrict_v": 1, "modul": [1, 2, 3, 4, 5, 6, 7, 8], "util": 1, "zero_point": 1, "content": [1, 2, 3, 4, 5, 6], "bit_width": 2, "const": 2, "paramet": 2, "function_wrapp": 3, "clamp": 3, "misc": 3, "ops_st": [3, 7], "shape": [3, 7, 10], "quant": [4, 10], "binari": [4, 16, 22], "delai": 4, "int": 4, "int_bas": 4, "ternari": 4, "scale": [5, 22], "int_scal": 5, "runtim": [5, 18], "standalon": 5, "stat": 6, "stats_op": 6, "stats_wrapp": 6, "view_wrapp": 6, "function": [7, 10, 22], "op": [7, 8], "autograd_ste_op": 8, "api": [9, 16], "refer": 9, "architectur": 10, "scriptmodul": 10, "injector": 10, "quantiz": [10, 12, 16, 19, 20, 21, 22], "enum": [10, 22], "solver": 10, "quanttensor": [10, 21, 22], "proxi": [10, 16], "layer": [10, 12, 16, 22], "mixin": 10, "export": [10, 12, 18, 20, 22, 24], "fx": [10, 22], "graph": 10, "trace": 10, "transform": 10, "loss": 10, "f": 11, "A": [11, 16], "q": 11, "get": 12, "start": 12, "ptq": 12, "over": 12, "hand": 12, "programmat": 12, "defin": [12, 22], "model": [12, 18], "nn": 12, "weight": [12, 16, 20, 22], "onli": [12, 20], "float": [12, 22], "activ": [12, 16, 19, 22], "bias": 12, "onnx": [12, 18, 22], "where": 12, "go": 12, "from": [12, 15, 22], "here": 12, "set": 14, "setup": 15, "requir": [15, 18], "instal": 15, "pypi": 15, "github": 15, "option": 15, "train": [15, 22], "infer": 15, "anatomi": 16, "what": 16, "": 16, "auto": 16, "wire": 16, "depend": [16, 22], "inject": [16, 22], "practic": 16, "exampl": [16, 18], "manual": 16, "an": [16, 19, 21], "extendedinjector": 16, "inherit": [16, 22], "composit": 16, "interfac": 16, "pass": 16, "custom": [16, 22], "quantident": [16, 22], "initi": 16, "statist": 16, "share": [16, 22], "instanc": 16, "deal": 16, "build": 16, "tutori": [17, 22], "introduct": 18, "quantizelinear": 18, "clip": 18, "dequantizelinear": 18, "qcdq": 18, "basic": 18, "complet": 18, "The": 18, "c": 18, "bitwidth": 18, "8": 18, "qop": [18, 22], "qgemm": 18, "v": 18, "gemm": 18, "overview": [19, 21], "rnn": 20, "lstm": 20, "quantrnn": 20, "quantlstm": 20, "just": 20, "time": 20, "compil": 20, "calibr": [20, 22], "full": 20, "quantconv2d": 21, "input": [21, 22], "output": [21, 22], "bia": [21, 22], "tvmcon": 22, "2021": 22, "fundament": 22, "quantlinear": 22, "default": 22, "mix": 22, "point": 22, "fix": 22, "quantrelu": 22, "requant": 22, "tensor": 22, "how": 22, "i": 22, "determin": 22, "oper": 22, "element": 22, "wise": 22, "arithmet": 22, "add": 22, "call": 22, "torch": 22, "max_pool": 22, "tanh": 22, "concaten": 22, "common": 22, "keyword": 22, "argument": 22, "bit": 22, "width": 22, "per": 22, "channel": 22, "max_val": 22, "init": 22, "scratch": 22, "learn": 22, "retrain": 22, "among": 22, "torchscript": 22, "backend": 22, "base": 22, "post": 22, "compat": 24, "user": 26, "guid": 26}, "envversion": {"sphinx.domains.c": 2, "sphinx.domains.changeset": 1, "sphinx.domains.citation": 1, "sphinx.domains.cpp": 8, "sphinx.domains.index": 1, "sphinx.domains.javascript": 2, "sphinx.domains.math": 2, "sphinx.domains.python": 3, "sphinx.domains.rst": 2, "sphinx.domains.std": 2, "sphinx.ext.intersphinx": 1, "sphinx.ext.viewcode": 1, "nbsphinx": 4, "sphinx": 57}, "alltitles": {"About": [[0, "about"]], "Author": [[0, "author"]], "Cite as": [[0, "cite-as"]], "brevitas.core package": [[1, "brevitas-core-package"]], "Subpackages": [[1, "subpackages"], [9, "subpackages"]], "Submodules": [[1, "submodules"], [2, "submodules"], [3, "submodules"], [4, "submodules"], [5, "submodules"], [6, "submodules"], [7, "submodules"], [8, "submodules"]], "brevitas.core.restrict_val module": [[1, "module-brevitas.core.restrict_val"]], "brevitas.core.utils module": [[1, "module-brevitas.core.utils"]], "brevitas.core.zero_point module": [[1, "module-brevitas.core.zero_point"]], "Module contents": [[1, "module-brevitas.core"], [2, "module-brevitas.core.bit_width"], [3, "module-brevitas.core.function_wrapper"], [4, "module-brevitas.core.quant"], [5, "module-brevitas.core.scaling"], [6, "module-brevitas.core.stats"]], "brevitas.core.bit_width package": [[2, "brevitas-core-bit-width-package"]], "brevitas.core.bit_width.const module": [[2, "module-brevitas.core.bit_width.const"]], "brevitas.core.bit_width.parameter module": [[2, "module-brevitas.core.bit_width.parameter"]], "brevitas.core.function_wrapper package": [[3, "brevitas-core-function-wrapper-package"]], "brevitas.core.function_wrapper.clamp module": [[3, "module-brevitas.core.function_wrapper.clamp"]], "brevitas.core.function_wrapper.misc module": [[3, "module-brevitas.core.function_wrapper.misc"]], "brevitas.core.function_wrapper.ops_ste module": [[3, "module-brevitas.core.function_wrapper.ops_ste"]], "brevitas.core.function_wrapper.shape module": [[3, "module-brevitas.core.function_wrapper.shape"]], "brevitas.core.quant package": [[4, "brevitas-core-quant-package"]], "brevitas.core.quant.binary module": [[4, "module-brevitas.core.quant.binary"]], "brevitas.core.quant.delay module": [[4, "module-brevitas.core.quant.delay"]], "brevitas.core.quant.int module": [[4, "module-brevitas.core.quant.int"]], "brevitas.core.quant.int_base module": [[4, "module-brevitas.core.quant.int_base"]], "brevitas.core.quant.ternary module": [[4, "module-brevitas.core.quant.ternary"]], "brevitas.core.scaling package": [[5, "brevitas-core-scaling-package"]], "brevitas.core.scaling.int_scaling module": [[5, "module-brevitas.core.scaling.int_scaling"]], "brevitas.core.scaling.runtime module": [[5, "module-brevitas.core.scaling.runtime"]], "brevitas.core.scaling.standalone module": [[5, "module-brevitas.core.scaling.standalone"]], "brevitas.core.stats package": [[6, "brevitas-core-stats-package"]], "brevitas.core.stats.stats_op module": [[6, "module-brevitas.core.stats.stats_op"]], "brevitas.core.stats.stats_wrapper module": [[6, "module-brevitas.core.stats.stats_wrapper"]], "brevitas.core.stats.view_wrapper module": [[6, "module-brevitas.core.stats.view_wrapper"]], "brevitas.function package": [[7, "brevitas-function-package"]], "brevitas.function.ops module": [[7, "module-brevitas.function.ops"]], "brevitas.function.ops_ste module": [[7, "module-brevitas.function.ops_ste"]], "brevitas.function.shape module": [[7, "module-brevitas.function.shape"]], "brevitas.ops package": [[8, "brevitas-ops-package"]], "brevitas.ops.autograd_ste_ops module": [[8, "module-brevitas.ops.autograd_ste_ops"]], "API reference": [[9, "api-reference"]], "Architecture": [[10, "architecture"]], "Functions": [[10, "functions"]], "Core ScriptModules": [[10, "core-scriptmodules"]], "Injectors and Quantizers": [[10, "injectors-and-quantizers"]], "Enums, Shapes and Solvers": [[10, "enums-shapes-and-solvers"]], "QuantTensor": [[10, "quanttensor"], [21, "QuantTensor"]], "Proxies": [[10, "proxies"]], "Quant Layers and Mixins": [[10, "quant-layers-and-mixins"]], "Export": [[10, "export"], [20, "Export"], [22, "Export"]], "FX graph tracing and transformations": [[10, "fx-graph-tracing-and-transformations"]], "Losses": [[10, "losses"]], "F.A.Q.": [[11, "f-a-q"]], "Getting started": [[12, "getting-started"]], "PTQ over hand or programmatically defined quantized models": [[12, "ptq-over-hand-or-programmatically-defined-quantized-models"]], "Defining a quantized model with brevitas.nn layers": [[12, "defining-a-quantized-model-with-brevitas-nn-layers"]], "Weights-only quantization, float activations and biases": [[12, "weights-only-quantization-float-activations-and-biases"]], "Weights and activations quantization, float biases": [[12, "weights-and-activations-quantization-float-biases"]], "Weights, activations, biases quantization": [[12, "weights-activations-biases-quantization"]], "Export to ONNX": [[12, "export-to-onnx"]], "Where to go from here": [[12, "where-to-go-from-here"]], "Brevitas": [[13, "brevitas"]], "Settings": [[14, "settings"]], "Setup": [[15, "setup"]], "Requirements": [[15, "requirements"], [18, "Requirements"]], "Installation Requirements": [[15, "installation-requirements"]], "Installation": [[15, "installation"]], "Installing from PyPI": [[15, "installing-from-pypi"]], "Installing from Github": [[15, "installing-from-github"]], "Optional Training Requirements": [[15, "optional-training-requirements"]], "Optional Inference Requirements": [[15, "optional-inference-requirements"]], "Anatomy of a Quantizer": [[16, "Anatomy-of-a-Quantizer"]], "What\u2019s in a Quantizer?": [[16, "What's-in-a-Quantizer?"]], "Quantization with auto-wiring Dependency Injection": [[16, "Quantization-with-auto-wiring-Dependency-Injection"]], "A Practical Example: Binary Quantization": [[16, "A-Practical-Example:-Binary-Quantization"]], "Manual Binary Quantization": [[16, "Manual-Binary-Quantization"]], "Binary Quantization with an ExtendedInjector": [[16, "Binary-Quantization-with-an-ExtendedInjector"]], "Inheritance and Composition of Quantizers": [[16, "Inheritance-and-Composition-of-Quantizers"]], "Interfacing a Quantizer with a Quantized Layer": [[16, "Interfacing-a-Quantizer-with-a-Quantized-Layer"]], "Passing a custom quantizer to QuantIdentity": [[16, "Passing-a-custom-quantizer-to-QuantIdentity"]], "A Custom Quantizer initialized with Weight Statistics": [[16, "A-Custom-Quantizer-initialized-with-Weight-Statistics"]], "Sharing a Quantizer": [[16, "Sharing-a-Quantizer"]], "Sharing a proxy": [[16, "Sharing-a-proxy"]], "Sharing an instance of Activation Quantization": [[16, "Sharing-an-instance-of-Activation-Quantization"]], "Dealing with Weight Initialization": [[16, "Dealing-with-Weight-Initialization"]], "Building a Custom Quantization API": [[16, "Building-a-Custom-Quantization-API"]], "Tutorials": [[17, "tutorials"]], "Tutorials:": [[17, null]], "ONNX Export": [[18, "ONNX-Export"]], "Introduction": [[18, "Introduction"]], "QuantizeLinear-Clip-DeQuantizeLinear (QCDQ)": [[18, "QuantizeLinear-Clip-DeQuantizeLinear-(QCDQ)"]], "Basic Example": [[18, "Basic-Example"]], "Complete Model": [[18, "Complete-Model"]], "The C in QCDQ (Bitwidth <= 8)": [[18, "The-C-in-QCDQ-(Bitwidth-<=-8)"]], "QOps Export": [[18, "QOps-Export"]], "Clipping in QOps": [[18, "Clipping-in-QOps"]], "ONNX Runtime": [[18, "ONNX-Runtime"]], "QCDQ": [[18, "QCDQ"]], "QGEMM vs GEMM": [[18, "QGEMM-vs-GEMM"]], "QOps": [[18, "QOps"]], "An Overview of Quantized Activations": [[19, "An-Overview-of-Quantized-Activations"]], "Quantized RNNs and LSTMs": [[20, "Quantized-RNNs-and-LSTMs"]], "QuantRNN": [[20, "QuantRNN"]], "QuantLSTM": [[20, "QuantLSTM"]], "Just-in-time compilation": [[20, "Just-in-time-compilation"]], "Calibration": [[20, "Calibration"]], "QuantLSTM weight-only quantization export": [[20, "QuantLSTM-weight-only-quantization-export"]], "QuantLSTM full quantization export": [[20, "QuantLSTM-full-quantization-export"]], "An overview of QuantTensor and QuantConv2d": [[21, "An-overview-of-QuantTensor-and-QuantConv2d"]], "Input Quantization": [[21, "Input-Quantization"]], "Output Quantization": [[21, "Output-Quantization"]], "Bias Quantization": [[21, "Bias-Quantization"], [22, "Bias-Quantization"]], "Brevitas TVMCon 2021 tutorial": [[22, "Brevitas-TVMCon-2021-tutorial"]], "Fundamentals": [[22, "Fundamentals"]], "QuantLinear layer": [[22, "QuantLinear-layer"]], "Weight quantization": [[22, "Weight-quantization"]], "Default weight quantization": [[22, "Default-weight-quantization"]], "Mixing quantized weights and floating-point inputs": [[22, "Mixing-quantized-weights-and-floating-point-inputs"]], "Fixed-point weight quantization": [[22, "Fixed-point-weight-quantization"]], "Binary weight quantization": [[22, "Binary-weight-quantization"]], "Sharing a weight quantizer": [[22, "Sharing-a-weight-quantizer"]], "Inputs/Outputs/Activations quantization:": [[22, "Inputs/Outputs/Activations-quantization:"]], "QuantIdentity layer": [[22, "QuantIdentity-layer"]], "QuantReLU layer": [[22, "QuantReLU-layer"]], "Requantizing a tensor": [[22, "Requantizing-a-tensor"]], "How is the activation scale determined by default?": [[22, "How-is-the-activation-scale-determined-by-default?"]], "Operations on QuantTensor": [[22, "Operations-on-QuantTensor"]], "Element-wise Arithmetic": [[22, "Element-wise-Arithmetic"]], "Element-wise adds": [[22, "Element-wise-adds"]], "Calling torch functions": [[22, "Calling-torch-functions"]], "max_pool on QuantTensor": [[22, "max_pool-on-QuantTensor"]], "tanh on QuantTensor": [[22, "tanh-on-QuantTensor"]], "QuantTensor concatenation": [[22, "QuantTensor-concatenation"]], "Customizing Quantizers": [[22, "Customizing-Quantizers"]], "Common keyword arguments": [[22, "Common-keyword-arguments"]], "Weight bit-width": [[22, "Weight-bit-width"]], "Per-channel weight quantization": [[22, "Per-channel-weight-quantization"]], "Activation bit-width": [[22, "Activation-bit-width"]], "Activation quantization with max_val init": [[22, "Activation-quantization-with-max_val-init"]], "Per-channel activation quantization": [[22, "Per-channel-activation-quantization"]], "Inheriting from a quantizer": [[22, "Inheriting-from-a-quantizer"]], "Defining a quantizer from scratch with enums": [[22, "Defining-a-quantizer-from-scratch-with-enums"]], "Weight quantizer": [[22, "Weight-quantizer"]], "Activation quantizer": [[22, "Activation-quantizer"]], "Learned scale and bit-width quantizer": [[22, "Learned-scale-and-bit-width-quantizer"]], "Retraining from floating-point": [[22, "Retraining-from-floating-point"]], "Defining a quantizer from scratch with dependency-injection": [[22, "Defining-a-quantizer-from-scratch-with-dependency-injection"]], "Activation quantization from scratch": [[22, "Activation-quantization-from-scratch"]], "Weight quantization with learned scale from scratch": [[22, "Weight-quantization-with-learned-scale-from-scratch"]], "Sharing learned bit-width among layers": [[22, "Sharing-learned-bit-width-among-layers"]], "Export to ONNX QOps": [[22, "Export-to-ONNX-QOps"]], "Export to custom Quantized ONNX": [[22, "Export-to-custom-Quantized-ONNX"]], "Export to TorchScript quantization backend": [[22, "Export-to-TorchScript-quantization-backend"]], "Brevitas and FX": [[22, "Brevitas-and-FX"]], "Calibration-based post-training quantization": [[22, "Calibration-based-post-training-quantization"]], "Export Compatibility": [[24, "export-compatibility"]], "User Guide": [[26, "user-guide"]]}, "indexentries": {"floatrestrictvalue (class in brevitas.core.restrict_val)": [[1, "brevitas.core.restrict_val.FloatRestrictValue"]], "intrestrictvalue (class in brevitas.core.restrict_val)": [[1, "brevitas.core.restrict_val.IntRestrictValue"]], "logfloatrestrictvalue (class in brevitas.core.restrict_val)": [[1, "brevitas.core.restrict_val.LogFloatRestrictValue"]], "parameterfromruntimezeropoint (class in brevitas.core.zero_point)": [[1, "brevitas.core.zero_point.ParameterFromRuntimeZeroPoint"]], "parameterfromstatsfromparameterzeropoint (class in brevitas.core.zero_point)": [[1, "brevitas.core.zero_point.ParameterFromStatsFromParameterZeroPoint"]], "parameterwrapper (class in brevitas.core.utils)": [[1, "brevitas.core.utils.ParameterWrapper"]], "parameterzeropoint (class in brevitas.core.zero_point)": [[1, "brevitas.core.zero_point.ParameterZeroPoint"]], "poweroftworestrictvalue (class in brevitas.core.restrict_val)": [[1, "brevitas.core.restrict_val.PowerOfTwoRestrictValue"]], "prezerocenterzeropoint (class in brevitas.core.zero_point)": [[1, "brevitas.core.zero_point.PreZeroCenterZeroPoint"]], "singleargstatelessbuffer (class in brevitas.core.utils)": [[1, "brevitas.core.utils.SingleArgStatelessBuffer"]], "slicetensor (class in brevitas.core.utils)": [[1, "brevitas.core.utils.SliceTensor"]], "statelessbuffer (class in brevitas.core.utils)": [[1, "brevitas.core.utils.StatelessBuffer"]], "statsfromparameterzeropoint (class in brevitas.core.zero_point)": [[1, "brevitas.core.zero_point.StatsFromParameterZeroPoint"]], "zerozeropoint (class in brevitas.core.zero_point)": [[1, "brevitas.core.zero_point.ZeroZeroPoint"]], "brevitas.core": [[1, "module-brevitas.core"]], "brevitas.core.restrict_val": [[1, "module-brevitas.core.restrict_val"]], "brevitas.core.utils": [[1, "module-brevitas.core.utils"]], "brevitas.core.zero_point": [[1, "module-brevitas.core.zero_point"]], "eager_forward() (brevitas.core.utils.slicetensor method)": [[1, "brevitas.core.utils.SliceTensor.eager_forward"]], "forward() (brevitas.core.restrict_val.floatrestrictvalue method)": [[1, "brevitas.core.restrict_val.FloatRestrictValue.forward"]], "forward() (brevitas.core.restrict_val.intrestrictvalue method)": [[1, "brevitas.core.restrict_val.IntRestrictValue.forward"]], "forward() (brevitas.core.restrict_val.logfloatrestrictvalue method)": [[1, "brevitas.core.restrict_val.LogFloatRestrictValue.forward"]], "forward() (brevitas.core.restrict_val.poweroftworestrictvalue method)": [[1, "brevitas.core.restrict_val.PowerOfTwoRestrictValue.forward"]], "forward() (brevitas.core.utils.parameterwrapper method)": [[1, "brevitas.core.utils.ParameterWrapper.forward"]], "forward() (brevitas.core.utils.singleargstatelessbuffer method)": [[1, "brevitas.core.utils.SingleArgStatelessBuffer.forward"]], "forward() (brevitas.core.utils.slicetensor method)": [[1, "brevitas.core.utils.SliceTensor.forward"]], "forward() (brevitas.core.utils.statelessbuffer method)": [[1, "brevitas.core.utils.StatelessBuffer.forward"]], "forward() (brevitas.core.zero_point.parameterfromruntimezeropoint method)": [[1, "brevitas.core.zero_point.ParameterFromRuntimeZeroPoint.forward"]], "forward() (brevitas.core.zero_point.parameterfromstatsfromparameterzeropoint method)": [[1, "brevitas.core.zero_point.ParameterFromStatsFromParameterZeroPoint.forward"]], "forward() (brevitas.core.zero_point.parameterzeropoint method)": [[1, "brevitas.core.zero_point.ParameterZeroPoint.forward"]], "forward() (brevitas.core.zero_point.prezerocenterzeropoint method)": [[1, "brevitas.core.zero_point.PreZeroCenterZeroPoint.forward"]], "forward() (brevitas.core.zero_point.statsfromparameterzeropoint method)": [[1, "brevitas.core.zero_point.StatsFromParameterZeroPoint.forward"]], "forward() (brevitas.core.zero_point.zerozeropoint method)": [[1, "brevitas.core.zero_point.ZeroZeroPoint.forward"]], "get_zero_center() (brevitas.core.zero_point.prezerocenterzeropoint method)": [[1, "brevitas.core.zero_point.PreZeroCenterZeroPoint.get_zero_center"]], "inplace_momentum_update() (in module brevitas.core.utils)": [[1, "brevitas.core.utils.inplace_momentum_update"]], "inplace_tensor_add() (in module brevitas.core.utils)": [[1, "brevitas.core.utils.inplace_tensor_add"]], "inplace_tensor_mul() (in module brevitas.core.utils)": [[1, "brevitas.core.utils.inplace_tensor_mul"]], "module": [[1, "module-brevitas.core"], [1, "module-brevitas.core.restrict_val"], [1, "module-brevitas.core.utils"], [1, "module-brevitas.core.zero_point"], [2, "module-brevitas.core.bit_width"], [2, "module-brevitas.core.bit_width.const"], [2, "module-brevitas.core.bit_width.parameter"], [3, "module-brevitas.core.function_wrapper"], [3, "module-brevitas.core.function_wrapper.clamp"], [3, "module-brevitas.core.function_wrapper.misc"], [3, "module-brevitas.core.function_wrapper.ops_ste"], [3, "module-brevitas.core.function_wrapper.shape"], [4, "module-brevitas.core.quant"], [4, "module-brevitas.core.quant.binary"], [4, "module-brevitas.core.quant.delay"], [4, "module-brevitas.core.quant.int"], [4, "module-brevitas.core.quant.int_base"], [4, "module-brevitas.core.quant.ternary"], [5, "module-brevitas.core.scaling"], [5, "module-brevitas.core.scaling.int_scaling"], [5, "module-brevitas.core.scaling.runtime"], [5, "module-brevitas.core.scaling.standalone"], [6, "module-brevitas.core.stats"], [6, "module-brevitas.core.stats.stats_op"], [6, "module-brevitas.core.stats.stats_wrapper"], [6, "module-brevitas.core.stats.view_wrapper"], [7, "module-brevitas.function.ops"], [7, "module-brevitas.function.ops_ste"], [7, "module-brevitas.function.shape"], [8, "module-brevitas.ops.autograd_ste_ops"]], "restrict_init_float() (brevitas.core.restrict_val.floatrestrictvalue method)": [[1, "brevitas.core.restrict_val.FloatRestrictValue.restrict_init_float"]], "restrict_init_float() (brevitas.core.restrict_val.intrestrictvalue method)": [[1, "brevitas.core.restrict_val.IntRestrictValue.restrict_init_float"]], "restrict_init_float() (brevitas.core.restrict_val.logfloatrestrictvalue method)": [[1, "brevitas.core.restrict_val.LogFloatRestrictValue.restrict_init_float"]], "restrict_init_float() (brevitas.core.restrict_val.poweroftworestrictvalue method)": [[1, "brevitas.core.restrict_val.PowerOfTwoRestrictValue.restrict_init_float"]], "restrict_init_inplace_module() (brevitas.core.restrict_val.floatrestrictvalue method)": [[1, "brevitas.core.restrict_val.FloatRestrictValue.restrict_init_inplace_module"]], "restrict_init_inplace_module() (brevitas.core.restrict_val.intrestrictvalue method)": [[1, "brevitas.core.restrict_val.IntRestrictValue.restrict_init_inplace_module"]], "restrict_init_inplace_module() (brevitas.core.restrict_val.logfloatrestrictvalue method)": [[1, "brevitas.core.restrict_val.LogFloatRestrictValue.restrict_init_inplace_module"]], "restrict_init_inplace_module() (brevitas.core.restrict_val.poweroftworestrictvalue method)": [[1, "brevitas.core.restrict_val.PowerOfTwoRestrictValue.restrict_init_inplace_module"]], "restrict_init_module() (brevitas.core.restrict_val.floatrestrictvalue method)": [[1, "brevitas.core.restrict_val.FloatRestrictValue.restrict_init_module"]], "restrict_init_module() (brevitas.core.restrict_val.intrestrictvalue method)": [[1, "brevitas.core.restrict_val.IntRestrictValue.restrict_init_module"]], "restrict_init_module() (brevitas.core.restrict_val.logfloatrestrictvalue method)": [[1, "brevitas.core.restrict_val.LogFloatRestrictValue.restrict_init_module"]], "restrict_init_module() (brevitas.core.restrict_val.poweroftworestrictvalue method)": [[1, "brevitas.core.restrict_val.PowerOfTwoRestrictValue.restrict_init_module"]], "restrict_init_tensor() (brevitas.core.restrict_val.floatrestrictvalue method)": [[1, "brevitas.core.restrict_val.FloatRestrictValue.restrict_init_tensor"]], "restrict_init_tensor() (brevitas.core.restrict_val.intrestrictvalue method)": [[1, "brevitas.core.restrict_val.IntRestrictValue.restrict_init_tensor"]], "restrict_init_tensor() (brevitas.core.restrict_val.logfloatrestrictvalue method)": [[1, "brevitas.core.restrict_val.LogFloatRestrictValue.restrict_init_tensor"]], "restrict_init_tensor() (brevitas.core.restrict_val.poweroftworestrictvalue method)": [[1, "brevitas.core.restrict_val.PowerOfTwoRestrictValue.restrict_init_tensor"]], "state_dict() (brevitas.core.utils.statelessbuffer method)": [[1, "brevitas.core.utils.StatelessBuffer.state_dict"]], "state_dict() (brevitas.core.zero_point.parameterfromruntimezeropoint method)": [[1, "brevitas.core.zero_point.ParameterFromRuntimeZeroPoint.state_dict"]], "state_dict() (brevitas.core.zero_point.parameterfromstatsfromparameterzeropoint method)": [[1, "brevitas.core.zero_point.ParameterFromStatsFromParameterZeroPoint.state_dict"]], "training_forward() (brevitas.core.zero_point.parameterfromruntimezeropoint method)": [[1, "brevitas.core.zero_point.ParameterFromRuntimeZeroPoint.training_forward"]], "bitwidthconst (class in brevitas.core.bit_width.const)": [[2, "brevitas.core.bit_width.const.BitWidthConst"]], "bitwidthparameter (class in brevitas.core.bit_width.parameter)": [[2, "brevitas.core.bit_width.parameter.BitWidthParameter"]], "bitwidthstatefulconst (class in brevitas.core.bit_width.const)": [[2, "brevitas.core.bit_width.const.BitWidthStatefulConst"]], "msbclampbitwidth (class in brevitas.core.bit_width.const)": [[2, "brevitas.core.bit_width.const.MsbClampBitWidth"]], "removebitwidthparameter (class in brevitas.core.bit_width.parameter)": [[2, "brevitas.core.bit_width.parameter.RemoveBitwidthParameter"]], "brevitas.core.bit_width": [[2, "module-brevitas.core.bit_width"]], "brevitas.core.bit_width.const": [[2, "module-brevitas.core.bit_width.const"]], "brevitas.core.bit_width.parameter": [[2, "module-brevitas.core.bit_width.parameter"]], "forward() (brevitas.core.bit_width.const.bitwidthconst method)": [[2, "brevitas.core.bit_width.const.BitWidthConst.forward"]], "forward() (brevitas.core.bit_width.const.bitwidthstatefulconst method)": [[2, "brevitas.core.bit_width.const.BitWidthStatefulConst.forward"]], "forward() (brevitas.core.bit_width.const.msbclampbitwidth method)": [[2, "brevitas.core.bit_width.const.MsbClampBitWidth.forward"]], "forward() (brevitas.core.bit_width.parameter.bitwidthparameter method)": [[2, "brevitas.core.bit_width.parameter.BitWidthParameter.forward"]], "forward() (brevitas.core.bit_width.parameter.removebitwidthparameter method)": [[2, "brevitas.core.bit_width.parameter.RemoveBitwidthParameter.forward"]], "ceilste (class in brevitas.core.function_wrapper.ops_ste)": [[3, "brevitas.core.function_wrapper.ops_ste.CeilSte"]], "clampmin (class in brevitas.core.function_wrapper.clamp)": [[3, "brevitas.core.function_wrapper.clamp.ClampMin"]], "dpuroundste (class in brevitas.core.function_wrapper.ops_ste)": [[3, "brevitas.core.function_wrapper.ops_ste.DPURoundSte"]], "floorste (class in brevitas.core.function_wrapper.ops_ste)": [[3, "brevitas.core.function_wrapper.ops_ste.FloorSte"]], "identity (class in brevitas.core.function_wrapper.misc)": [[3, "brevitas.core.function_wrapper.misc.Identity"]], "inplacelogtwo (class in brevitas.core.function_wrapper.misc)": [[3, "brevitas.core.function_wrapper.misc.InplaceLogTwo"]], "inplacetensorclampste (class in brevitas.core.function_wrapper.ops_ste)": [[3, "brevitas.core.function_wrapper.ops_ste.InplaceTensorClampSte"]], "logtwo (class in brevitas.core.function_wrapper.misc)": [[3, "brevitas.core.function_wrapper.misc.LogTwo"]], "over_batch_over_output_channels (brevitas.core.function_wrapper.shape.statsinputviewshapeimpl attribute)": [[3, "brevitas.core.function_wrapper.shape.StatsInputViewShapeImpl.OVER_BATCH_OVER_OUTPUT_CHANNELS"]], "over_batch_over_tensor (brevitas.core.function_wrapper.shape.statsinputviewshapeimpl attribute)": [[3, "brevitas.core.function_wrapper.shape.StatsInputViewShapeImpl.OVER_BATCH_OVER_TENSOR"]], "over_output_channels (brevitas.core.function_wrapper.shape.statsinputviewshapeimpl attribute)": [[3, "brevitas.core.function_wrapper.shape.StatsInputViewShapeImpl.OVER_OUTPUT_CHANNELS"]], "over_output_features (brevitas.core.function_wrapper.shape.statsinputviewshapeimpl attribute)": [[3, "brevitas.core.function_wrapper.shape.StatsInputViewShapeImpl.OVER_OUTPUT_FEATURES"]], "over_tensor (brevitas.core.function_wrapper.shape.statsinputviewshapeimpl attribute)": [[3, "brevitas.core.function_wrapper.shape.StatsInputViewShapeImpl.OVER_TENSOR"]], "overbatchoveroutputchannelview (class in brevitas.core.function_wrapper.shape)": [[3, "brevitas.core.function_wrapper.shape.OverBatchOverOutputChannelView"]], "overbatchovertensorview (class in brevitas.core.function_wrapper.shape)": [[3, "brevitas.core.function_wrapper.shape.OverBatchOverTensorView"]], "overoutputchannelview (class in brevitas.core.function_wrapper.shape)": [[3, "brevitas.core.function_wrapper.shape.OverOutputChannelView"]], "overoutputfeaturesview (class in brevitas.core.function_wrapper.shape)": [[3, "brevitas.core.function_wrapper.shape.OverOutputFeaturesView"]], "overtensorview (class in brevitas.core.function_wrapper.shape)": [[3, "brevitas.core.function_wrapper.shape.OverTensorView"]], "permutedims (class in brevitas.core.function_wrapper.shape)": [[3, "brevitas.core.function_wrapper.shape.PermuteDims"]], "poweroftwo (class in brevitas.core.function_wrapper.misc)": [[3, "brevitas.core.function_wrapper.misc.PowerOfTwo"]], "roundste (class in brevitas.core.function_wrapper.ops_ste)": [[3, "brevitas.core.function_wrapper.ops_ste.RoundSte"]], "roundtozeroste (class in brevitas.core.function_wrapper.ops_ste)": [[3, "brevitas.core.function_wrapper.ops_ste.RoundToZeroSte"]], "scalarclamp (class in brevitas.core.function_wrapper.clamp)": [[3, "brevitas.core.function_wrapper.clamp.ScalarClamp"]], "scalarclampminste (class in brevitas.core.function_wrapper.ops_ste)": [[3, "brevitas.core.function_wrapper.ops_ste.ScalarClampMinSte"]], "statsinputviewshapeimpl (class in brevitas.core.function_wrapper.shape)": [[3, "brevitas.core.function_wrapper.shape.StatsInputViewShapeImpl"]], "tensorclamp (class in brevitas.core.function_wrapper.clamp)": [[3, "brevitas.core.function_wrapper.clamp.TensorClamp"]], "tensorclampste (class in brevitas.core.function_wrapper.ops_ste)": [[3, "brevitas.core.function_wrapper.ops_ste.TensorClampSte"]], "brevitas.core.function_wrapper": [[3, "module-brevitas.core.function_wrapper"]], "brevitas.core.function_wrapper.clamp": [[3, "module-brevitas.core.function_wrapper.clamp"]], "brevitas.core.function_wrapper.misc": [[3, "module-brevitas.core.function_wrapper.misc"]], "brevitas.core.function_wrapper.ops_ste": [[3, "module-brevitas.core.function_wrapper.ops_ste"]], "brevitas.core.function_wrapper.shape": [[3, "module-brevitas.core.function_wrapper.shape"]], "forward() (brevitas.core.function_wrapper.clamp.clampmin method)": [[3, "brevitas.core.function_wrapper.clamp.ClampMin.forward"]], "forward() (brevitas.core.function_wrapper.clamp.scalarclamp method)": [[3, "brevitas.core.function_wrapper.clamp.ScalarClamp.forward"]], "forward() (brevitas.core.function_wrapper.clamp.tensorclamp method)": [[3, "brevitas.core.function_wrapper.clamp.TensorClamp.forward"]], "forward() (brevitas.core.function_wrapper.misc.identity method)": [[3, "brevitas.core.function_wrapper.misc.Identity.forward"]], "forward() (brevitas.core.function_wrapper.misc.inplacelogtwo method)": [[3, "brevitas.core.function_wrapper.misc.InplaceLogTwo.forward"]], "forward() (brevitas.core.function_wrapper.misc.logtwo method)": [[3, "brevitas.core.function_wrapper.misc.LogTwo.forward"]], "forward() (brevitas.core.function_wrapper.misc.poweroftwo method)": [[3, "brevitas.core.function_wrapper.misc.PowerOfTwo.forward"]], "forward() (brevitas.core.function_wrapper.ops_ste.ceilste method)": [[3, "brevitas.core.function_wrapper.ops_ste.CeilSte.forward"]], "forward() (brevitas.core.function_wrapper.ops_ste.dpuroundste method)": [[3, "brevitas.core.function_wrapper.ops_ste.DPURoundSte.forward"]], "forward() (brevitas.core.function_wrapper.ops_ste.floorste method)": [[3, "brevitas.core.function_wrapper.ops_ste.FloorSte.forward"]], "forward() (brevitas.core.function_wrapper.ops_ste.inplacetensorclampste method)": [[3, "brevitas.core.function_wrapper.ops_ste.InplaceTensorClampSte.forward"]], "forward() (brevitas.core.function_wrapper.ops_ste.roundste method)": [[3, "brevitas.core.function_wrapper.ops_ste.RoundSte.forward"]], "forward() (brevitas.core.function_wrapper.ops_ste.roundtozeroste method)": [[3, "brevitas.core.function_wrapper.ops_ste.RoundToZeroSte.forward"]], "forward() (brevitas.core.function_wrapper.ops_ste.scalarclampminste method)": [[3, "brevitas.core.function_wrapper.ops_ste.ScalarClampMinSte.forward"]], "forward() (brevitas.core.function_wrapper.ops_ste.tensorclampste method)": [[3, "brevitas.core.function_wrapper.ops_ste.TensorClampSte.forward"]], "forward() (brevitas.core.function_wrapper.shape.overbatchoveroutputchannelview method)": [[3, "brevitas.core.function_wrapper.shape.OverBatchOverOutputChannelView.forward"]], "forward() (brevitas.core.function_wrapper.shape.overbatchovertensorview method)": [[3, "brevitas.core.function_wrapper.shape.OverBatchOverTensorView.forward"]], "forward() (brevitas.core.function_wrapper.shape.overoutputchannelview method)": [[3, "brevitas.core.function_wrapper.shape.OverOutputChannelView.forward"]], "forward() (brevitas.core.function_wrapper.shape.overoutputfeaturesview method)": [[3, "brevitas.core.function_wrapper.shape.OverOutputFeaturesView.forward"]], "forward() (brevitas.core.function_wrapper.shape.overtensorview method)": [[3, "brevitas.core.function_wrapper.shape.OverTensorView.forward"]], "forward() (brevitas.core.function_wrapper.shape.permutedims method)": [[3, "brevitas.core.function_wrapper.shape.PermuteDims.forward"]], "binaryquant (class in brevitas.core.quant.binary)": [[4, "brevitas.core.quant.binary.BinaryQuant"]], "clampedbinaryquant (class in brevitas.core.quant.binary)": [[4, "brevitas.core.quant.binary.ClampedBinaryQuant"]], "decoupledintquant (class in brevitas.core.quant.int_base)": [[4, "brevitas.core.quant.int_base.DecoupledIntQuant"]], "decoupledrescalingintquant (class in brevitas.core.quant.int)": [[4, "brevitas.core.quant.int.DecoupledRescalingIntQuant"]], "decoupledrescalingintquantwithinput (class in brevitas.core.quant.int)": [[4, "brevitas.core.quant.int.DecoupledRescalingIntQuantWithInput"]], "delaywrapper (class in brevitas.core.quant.delay)": [[4, "brevitas.core.quant.delay.DelayWrapper"]], "intquant (class in brevitas.core.quant.int_base)": [[4, "brevitas.core.quant.int_base.IntQuant"]], "prescaledrestrictintquant (class in brevitas.core.quant.int)": [[4, "brevitas.core.quant.int.PrescaledRestrictIntQuant"]], "prescaledrestrictintquantwithinputbitwidth (class in brevitas.core.quant.int)": [[4, "brevitas.core.quant.int.PrescaledRestrictIntQuantWithInputBitWidth"]], "rescalingintquant (class in brevitas.core.quant.int)": [[4, "brevitas.core.quant.int.RescalingIntQuant"]], "ternaryquant (class in brevitas.core.quant.ternary)": [[4, "brevitas.core.quant.ternary.TernaryQuant"]], "truncintquant (class in brevitas.core.quant.int)": [[4, "brevitas.core.quant.int.TruncIntQuant"]], "brevitas.core.quant": [[4, "module-brevitas.core.quant"]], "brevitas.core.quant.binary": [[4, "module-brevitas.core.quant.binary"]], "brevitas.core.quant.delay": [[4, "module-brevitas.core.quant.delay"]], "brevitas.core.quant.int": [[4, "module-brevitas.core.quant.int"]], "brevitas.core.quant.int_base": [[4, "module-brevitas.core.quant.int_base"]], "brevitas.core.quant.ternary": [[4, "module-brevitas.core.quant.ternary"]], "forward() (brevitas.core.quant.binary.binaryquant method)": [[4, "brevitas.core.quant.binary.BinaryQuant.forward"]], "forward() (brevitas.core.quant.binary.clampedbinaryquant method)": [[4, "brevitas.core.quant.binary.ClampedBinaryQuant.forward"]], "forward() (brevitas.core.quant.delay.delaywrapper method)": [[4, "brevitas.core.quant.delay.DelayWrapper.forward"]], "forward() (brevitas.core.quant.int.decoupledrescalingintquant method)": [[4, "brevitas.core.quant.int.DecoupledRescalingIntQuant.forward"]], "forward() (brevitas.core.quant.int.decoupledrescalingintquantwithinput method)": [[4, "brevitas.core.quant.int.DecoupledRescalingIntQuantWithInput.forward"]], "forward() (brevitas.core.quant.int.prescaledrestrictintquant method)": [[4, "brevitas.core.quant.int.PrescaledRestrictIntQuant.forward"]], "forward() (brevitas.core.quant.int.prescaledrestrictintquantwithinputbitwidth method)": [[4, "brevitas.core.quant.int.PrescaledRestrictIntQuantWithInputBitWidth.forward"]], "forward() (brevitas.core.quant.int.rescalingintquant method)": [[4, "brevitas.core.quant.int.RescalingIntQuant.forward"]], "forward() (brevitas.core.quant.int.truncintquant method)": [[4, "brevitas.core.quant.int.TruncIntQuant.forward"]], "forward() (brevitas.core.quant.int_base.decoupledintquant method)": [[4, "brevitas.core.quant.int_base.DecoupledIntQuant.forward"]], "forward() (brevitas.core.quant.int_base.intquant method)": [[4, "brevitas.core.quant.int_base.IntQuant.forward"]], "forward() (brevitas.core.quant.ternary.ternaryquant method)": [[4, "brevitas.core.quant.ternary.TernaryQuant.forward"]], "max_int() (brevitas.core.quant.int_base.decoupledintquant method)": [[4, "brevitas.core.quant.int_base.DecoupledIntQuant.max_int"]], "max_int() (brevitas.core.quant.int_base.intquant method)": [[4, "brevitas.core.quant.int_base.IntQuant.max_int"]], "min_int() (brevitas.core.quant.int_base.decoupledintquant method)": [[4, "brevitas.core.quant.int_base.DecoupledIntQuant.min_int"]], "min_int() (brevitas.core.quant.int_base.intquant method)": [[4, "brevitas.core.quant.int_base.IntQuant.min_int"]], "to_int() (brevitas.core.quant.int_base.decoupledintquant method)": [[4, "brevitas.core.quant.int_base.DecoupledIntQuant.to_int"]], "to_int() (brevitas.core.quant.int_base.intquant method)": [[4, "brevitas.core.quant.int_base.IntQuant.to_int"]], "constscaling (class in brevitas.core.scaling.standalone)": [[5, "brevitas.core.scaling.standalone.ConstScaling"]], "intscaling (class in brevitas.core.scaling.int_scaling)": [[5, "brevitas.core.scaling.int_scaling.IntScaling"]], "parameterfromruntimestatsscaling (class in brevitas.core.scaling.standalone)": [[5, "brevitas.core.scaling.standalone.ParameterFromRuntimeStatsScaling"]], "parameterfromstatsfromparameterscaling (class in brevitas.core.scaling.standalone)": [[5, "brevitas.core.scaling.standalone.ParameterFromStatsFromParameterScaling"]], "parameterscaling (class in brevitas.core.scaling.standalone)": [[5, "brevitas.core.scaling.standalone.ParameterScaling"]], "poweroftwointscaling (class in brevitas.core.scaling.int_scaling)": [[5, "brevitas.core.scaling.int_scaling.PowerOfTwoIntScaling"]], "runtimestatsscaling (class in brevitas.core.scaling.runtime)": [[5, "brevitas.core.scaling.runtime.RuntimeStatsScaling"]], "statsfromparameterscaling (class in brevitas.core.scaling.runtime)": [[5, "brevitas.core.scaling.runtime.StatsFromParameterScaling"]], "brevitas.core.scaling": [[5, "module-brevitas.core.scaling"]], "brevitas.core.scaling.int_scaling": [[5, "module-brevitas.core.scaling.int_scaling"]], "brevitas.core.scaling.runtime": [[5, "module-brevitas.core.scaling.runtime"]], "brevitas.core.scaling.standalone": [[5, "module-brevitas.core.scaling.standalone"]], "forward() (brevitas.core.scaling.int_scaling.intscaling method)": [[5, "brevitas.core.scaling.int_scaling.IntScaling.forward"]], "forward() (brevitas.core.scaling.int_scaling.poweroftwointscaling method)": [[5, "brevitas.core.scaling.int_scaling.PowerOfTwoIntScaling.forward"]], "forward() (brevitas.core.scaling.runtime.runtimestatsscaling method)": [[5, "brevitas.core.scaling.runtime.RuntimeStatsScaling.forward"]], "forward() (brevitas.core.scaling.runtime.statsfromparameterscaling method)": [[5, "brevitas.core.scaling.runtime.StatsFromParameterScaling.forward"]], "forward() (brevitas.core.scaling.standalone.constscaling method)": [[5, "brevitas.core.scaling.standalone.ConstScaling.forward"]], "forward() (brevitas.core.scaling.standalone.parameterfromruntimestatsscaling method)": [[5, "brevitas.core.scaling.standalone.ParameterFromRuntimeStatsScaling.forward"]], "forward() (brevitas.core.scaling.standalone.parameterfromstatsfromparameterscaling method)": [[5, "brevitas.core.scaling.standalone.ParameterFromStatsFromParameterScaling.forward"]], "forward() (brevitas.core.scaling.standalone.parameterscaling method)": [[5, "brevitas.core.scaling.standalone.ParameterScaling.forward"]], "state_dict() (brevitas.core.scaling.standalone.parameterfromruntimestatsscaling method)": [[5, "brevitas.core.scaling.standalone.ParameterFromRuntimeStatsScaling.state_dict"]], "state_dict() (brevitas.core.scaling.standalone.parameterfromstatsfromparameterscaling method)": [[5, "brevitas.core.scaling.standalone.ParameterFromStatsFromParameterScaling.state_dict"]], "training_forward() (brevitas.core.scaling.standalone.parameterfromruntimestatsscaling method)": [[5, "brevitas.core.scaling.standalone.ParameterFromRuntimeStatsScaling.training_forward"]], "absave (class in brevitas.core.stats.stats_op)": [[6, "brevitas.core.stats.stats_op.AbsAve"]], "absmax (class in brevitas.core.stats.stats_op)": [[6, "brevitas.core.stats.stats_op.AbsMax"]], "absmaxave (class in brevitas.core.stats.stats_op)": [[6, "brevitas.core.stats.stats_op.AbsMaxAve"]], "absmaxl2 (class in brevitas.core.stats.stats_op)": [[6, "brevitas.core.stats.stats_op.AbsMaxL2"]], "absminmax (class in brevitas.core.stats.stats_op)": [[6, "brevitas.core.stats.stats_op.AbsMinMax"]], "abspercentile (class in brevitas.core.stats.stats_op)": [[6, "brevitas.core.stats.stats_op.AbsPercentile"]], "klminimizerthreshold (class in brevitas.core.stats.stats_op)": [[6, "brevitas.core.stats.stats_op.KLMinimizerThreshold"]], "l1norm (class in brevitas.core.stats.stats_op)": [[6, "brevitas.core.stats.stats_op.L1Norm"]], "l2norm (class in brevitas.core.stats.stats_op)": [[6, "brevitas.core.stats.stats_op.L2Norm"]], "mse (class in brevitas.core.stats.stats_op)": [[6, "brevitas.core.stats.stats_op.MSE"]], "meanlearnedsigmastd (class in brevitas.core.stats.stats_op)": [[6, "brevitas.core.stats.stats_op.MeanLearnedSigmaStd"]], "meansigmastd (class in brevitas.core.stats.stats_op)": [[6, "brevitas.core.stats.stats_op.MeanSigmaStd"]], "negativeminorzero (class in brevitas.core.stats.stats_op)": [[6, "brevitas.core.stats.stats_op.NegativeMinOrZero"]], "negativepercentileorzero (class in brevitas.core.stats.stats_op)": [[6, "brevitas.core.stats.stats_op.NegativePercentileOrZero"]], "percentileinterval (class in brevitas.core.stats.stats_op)": [[6, "brevitas.core.stats.stats_op.PercentileInterval"]], "brevitas.core.stats": [[6, "module-brevitas.core.stats"]], "brevitas.core.stats.stats_op": [[6, "module-brevitas.core.stats.stats_op"]], "brevitas.core.stats.stats_wrapper": [[6, "module-brevitas.core.stats.stats_wrapper"]], "brevitas.core.stats.view_wrapper": [[6, "module-brevitas.core.stats.view_wrapper"]], "evaluate_loss() (brevitas.core.stats.stats_op.mse method)": [[6, "brevitas.core.stats.stats_op.MSE.evaluate_loss"]], "forward() (brevitas.core.stats.stats_op.absave method)": [[6, "brevitas.core.stats.stats_op.AbsAve.forward"]], "forward() (brevitas.core.stats.stats_op.absmax method)": [[6, "brevitas.core.stats.stats_op.AbsMax.forward"]], "forward() (brevitas.core.stats.stats_op.absmaxave method)": [[6, "brevitas.core.stats.stats_op.AbsMaxAve.forward"]], "forward() (brevitas.core.stats.stats_op.absmaxl2 method)": [[6, "brevitas.core.stats.stats_op.AbsMaxL2.forward"]], "forward() (brevitas.core.stats.stats_op.absminmax method)": [[6, "brevitas.core.stats.stats_op.AbsMinMax.forward"]], "forward() (brevitas.core.stats.stats_op.abspercentile method)": [[6, "brevitas.core.stats.stats_op.AbsPercentile.forward"]], "forward() (brevitas.core.stats.stats_op.klminimizerthreshold method)": [[6, "brevitas.core.stats.stats_op.KLMinimizerThreshold.forward"]], "forward() (brevitas.core.stats.stats_op.l1norm method)": [[6, "brevitas.core.stats.stats_op.L1Norm.forward"]], "forward() (brevitas.core.stats.stats_op.l2norm method)": [[6, "brevitas.core.stats.stats_op.L2Norm.forward"]], "forward() (brevitas.core.stats.stats_op.mse method)": [[6, "brevitas.core.stats.stats_op.MSE.forward"]], "forward() (brevitas.core.stats.stats_op.meanlearnedsigmastd method)": [[6, "brevitas.core.stats.stats_op.MeanLearnedSigmaStd.forward"]], "forward() (brevitas.core.stats.stats_op.meansigmastd method)": [[6, "brevitas.core.stats.stats_op.MeanSigmaStd.forward"]], "forward() (brevitas.core.stats.stats_op.negativeminorzero method)": [[6, "brevitas.core.stats.stats_op.NegativeMinOrZero.forward"]], "forward() (brevitas.core.stats.stats_op.negativepercentileorzero method)": [[6, "brevitas.core.stats.stats_op.NegativePercentileOrZero.forward"]], "forward() (brevitas.core.stats.stats_op.percentileinterval method)": [[6, "brevitas.core.stats.stats_op.PercentileInterval.forward"]], "mse_fib_search() (brevitas.core.stats.stats_op.mse method)": [[6, "brevitas.core.stats.stats_op.MSE.mse_fib_search"]], "mse_grid_search() (brevitas.core.stats.stats_op.mse method)": [[6, "brevitas.core.stats.stats_op.MSE.mse_grid_search"]], "mse_loss_fn() (brevitas.core.stats.stats_op.mse method)": [[6, "brevitas.core.stats.stats_op.MSE.mse_loss_fn"]], "mse_search() (brevitas.core.stats.stats_op.mse method)": [[6, "brevitas.core.stats.stats_op.MSE.mse_search"]], "smooth_normalize_distribution() (brevitas.core.stats.stats_op.klminimizerthreshold method)": [[6, "brevitas.core.stats.stats_op.KLMinimizerThreshold.smooth_normalize_distribution"]], "abs_binary_sign_grad() (in module brevitas.function.ops_ste)": [[7, "brevitas.function.ops_ste.abs_binary_sign_grad"]], "binary_sign() (in module brevitas.function.ops)": [[7, "brevitas.function.ops.binary_sign"]], "binary_sign_ste() (in module brevitas.function.ops_ste)": [[7, "brevitas.function.ops_ste.binary_sign_ste"]], "brevitas.function.ops": [[7, "module-brevitas.function.ops"]], "brevitas.function.ops_ste": [[7, "module-brevitas.function.ops_ste"]], "brevitas.function.shape": [[7, "module-brevitas.function.shape"]], "ceil_ste() (in module brevitas.function.ops_ste)": [[7, "brevitas.function.ops_ste.ceil_ste"]], "dpu_round() (in module brevitas.function.ops)": [[7, "brevitas.function.ops.dpu_round"]], "dpu_round_ste() (in module brevitas.function.ops_ste)": [[7, "brevitas.function.ops_ste.dpu_round_ste"]], "floor_ste() (in module brevitas.function.ops_ste)": [[7, "brevitas.function.ops_ste.floor_ste"]], "get_upper_bound_on_l1_norm() (in module brevitas.function.ops)": [[7, "brevitas.function.ops.get_upper_bound_on_l1_norm"]], "identity() (in module brevitas.function.ops)": [[7, "brevitas.function.ops.identity"]], "max_float() (in module brevitas.function.ops)": [[7, "brevitas.function.ops.max_float"]], "max_int() (in module brevitas.function.ops)": [[7, "brevitas.function.ops.max_int"]], "min_int() (in module brevitas.function.ops)": [[7, "brevitas.function.ops.min_int"]], "over_batch_over_output_channels() (in module brevitas.function.shape)": [[7, "brevitas.function.shape.over_batch_over_output_channels"]], "over_batch_over_tensor() (in module brevitas.function.shape)": [[7, "brevitas.function.shape.over_batch_over_tensor"]], "over_output_channels() (in module brevitas.function.shape)": [[7, "brevitas.function.shape.over_output_channels"]], "over_output_features() (in module brevitas.function.shape)": [[7, "brevitas.function.shape.over_output_features"]], "over_tensor() (in module brevitas.function.shape)": [[7, "brevitas.function.shape.over_tensor"]], "round_ste() (in module brevitas.function.ops_ste)": [[7, "brevitas.function.ops_ste.round_ste"]], "round_to_zero() (in module brevitas.function.ops)": [[7, "brevitas.function.ops.round_to_zero"]], "round_to_zero_ste() (in module brevitas.function.ops_ste)": [[7, "brevitas.function.ops_ste.round_to_zero_ste"]], "scalar_clamp_min_ste() (in module brevitas.function.ops_ste)": [[7, "brevitas.function.ops_ste.scalar_clamp_min_ste"]], "scalar_clamp_ste() (in module brevitas.function.ops_ste)": [[7, "brevitas.function.ops_ste.scalar_clamp_ste"]], "tensor_clamp() (in module brevitas.function.ops)": [[7, "brevitas.function.ops.tensor_clamp"]], "tensor_clamp_() (in module brevitas.function.ops)": [[7, "brevitas.function.ops.tensor_clamp_"]], "tensor_clamp_ste() (in module brevitas.function.ops_ste)": [[7, "brevitas.function.ops_ste.tensor_clamp_ste"]], "tensor_clamp_ste_() (in module brevitas.function.ops_ste)": [[7, "brevitas.function.ops_ste.tensor_clamp_ste_"]], "ternary_sign_ste() (in module brevitas.function.ops_ste)": [[7, "brevitas.function.ops_ste.ternary_sign_ste"]], "absbinarysigngradfn (class in brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.AbsBinarySignGradFn"]], "binarysignstefn (class in brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.BinarySignSteFn"]], "ceilstefn (class in brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.CeilSteFn"]], "dpuroundstefn (class in brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.DPURoundSteFn"]], "floorstefn (class in brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.FloorSteFn"]], "inplacetensorclampstefn (class in brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.InplaceTensorClampSteFn"]], "roundstefn (class in brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.RoundSteFn"]], "roundtozerostefn (class in brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.RoundToZeroSteFn"]], "scalarclampminstefn (class in brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.ScalarClampMinSteFn"]], "scalarclampstefn (class in brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.ScalarClampSteFn"]], "tensorclampstefn (class in brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.TensorClampSteFn"]], "ternarysignstefn (class in brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.TernarySignSteFn"]], "abs_binary_sign_grad_impl() (in module brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.abs_binary_sign_grad_impl"]], "binary_sign_ste_impl() (in module brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.binary_sign_ste_impl"]], "brevitas.ops.autograd_ste_ops": [[8, "module-brevitas.ops.autograd_ste_ops"]], "ceil_ste_impl() (in module brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.ceil_ste_impl"]], "dpu_round_ste_impl() (in module brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.dpu_round_ste_impl"]], "floor_ste_impl() (in module brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.floor_ste_impl"]], "round_ste_impl() (in module brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.round_ste_impl"]], "round_to_zero_ste_impl() (in module brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.round_to_zero_ste_impl"]], "scalar_clamp_min_ste_impl() (in module brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.scalar_clamp_min_ste_impl"]], "scalar_clamp_ste_impl() (in module brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.scalar_clamp_ste_impl"]], "tensor_clamp_ste_impl() (in module brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.tensor_clamp_ste_impl"]], "ternary_sign_ste_impl() (in module brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.ternary_sign_ste_impl"]]}}) \ No newline at end of file +Search.setIndex({"docnames": ["about", "api_reference/brevitas.core", "api_reference/brevitas.core.bit_width", "api_reference/brevitas.core.function_wrapper", "api_reference/brevitas.core.quant", "api_reference/brevitas.core.scaling", "api_reference/brevitas.core.stats", "api_reference/brevitas.function", "api_reference/brevitas.ops", "api_reference/index", "architecture", "faq", "getting_started", "index", "settings", "setup", "tutorials/anatomy_quantizer", "tutorials/index", "tutorials/onnx_export", "tutorials/quant_activation_overview", "tutorials/quant_recurrent", "tutorials/quant_tensor_quant_conv2d_overview", "tutorials/tvmcon2021", "user_guide/datatypes", "user_guide/export", "user_guide/graph_transformations", "user_guide/index", "user_guide/precision", "user_guide/proxies", "user_guide/quantized_layers", "user_guide/quantized_tensor", "user_guide/quantizers", "user_guide/scaling", "user_guide/zero_point"], "filenames": ["about.rst", "api_reference/brevitas.core.rst", "api_reference/brevitas.core.bit_width.rst", "api_reference/brevitas.core.function_wrapper.rst", "api_reference/brevitas.core.quant.rst", "api_reference/brevitas.core.scaling.rst", "api_reference/brevitas.core.stats.rst", "api_reference/brevitas.function.rst", "api_reference/brevitas.ops.rst", "api_reference/index.rst", "architecture.rst", "faq.rst", "getting_started.rst", "index.rst", "settings.rst", "setup.rst", "tutorials/anatomy_quantizer.nblink", "tutorials/index.rst", "tutorials/onnx_export.nblink", "tutorials/quant_activation_overview.nblink", "tutorials/quant_recurrent.nblink", "tutorials/quant_tensor_quant_conv2d_overview.nblink", "tutorials/tvmcon2021.nblink", "user_guide/datatypes.rst", "user_guide/export.rst", "user_guide/graph_transformations.rst", "user_guide/index.rst", "user_guide/precision.rst", "user_guide/proxies.rst", "user_guide/quantized_layers.rst", "user_guide/quantized_tensor.rst", "user_guide/quantizers.rst", "user_guide/scaling.rst", "user_guide/zero_point.rst"], "titles": ["About", "brevitas.core package", "brevitas.core.bit_width package", "brevitas.core.function_wrapper package", "brevitas.core.quant package", "brevitas.core.scaling package", "brevitas.core.stats package", "brevitas.function package", "brevitas.ops package", "API reference", "Architecture", "F.A.Q.", "Getting started", "Brevitas", "Settings", "Setup", "Anatomy of a Quantizer", "Tutorials", "ONNX Export", "An Overview of Quantized Activations", "Quantized RNNs and LSTMs", "An overview of QuantTensor and QuantConv2d", "Brevitas TVMCon 2021 tutorial", "<no title>", "Export Compatibility", "<no title>", "User Guide", "<no title>", "<no title>", "<no title>", "<no title>", "<no title>", "<no title>", "<no title>"], "terms": {"alessandro": [0, 22], "pappalardo": [0, 7], "volcaciu": 0, "xilinx": [0, 13, 15, 22, 24], "research": [0, 13], "lab": 0, "For": [0, 10, 12, 13, 16, 18, 19, 20, 21, 22], "privat": 0, "commun": 0, "you": [0, 10, 11, 15, 18, 22], "can": [0, 3, 7, 10, 11, 12, 14, 15, 16, 18, 19, 20, 21, 22, 24], "reach": 0, "me": 0, "alessand": [0, 18, 20], "name_of_my_employ": 0, "dot": 0, "com": [0, 15, 22], "softwar": [0, 10], "brevita": [0, 9, 10, 11, 14, 15, 16, 18, 19, 20, 21, 24], "titl": 0, "publish": 0, "zenodo": 0, "doi": 0, "10": [0, 3, 12, 16, 18, 19, 20, 21, 22], "5281": 0, "3333552": 0, "url": 0, "http": [0, 15, 18, 20, 21, 22], "org": 0, "bit_width": [1, 4, 5, 7, 12, 16, 18, 19, 20, 21, 22], "const": [1, 5, 22], "bitwidthconst": [1, 2, 4, 16, 22], "forward": [1, 2, 3, 4, 5, 6, 12, 16, 18, 20, 21, 22], "bitwidthstatefulconst": [1, 2], "msbclampbitwidth": [1, 2], "paramet": [1, 4, 5, 7, 10, 14, 16, 19, 20, 22], "bitwidthparamet": [1, 2], "removebitwidthparamet": [1, 2], "function_wrapp": [1, 4, 22], "clamp": [1, 4, 7, 8, 16], "clampmin": [1, 3], "scalarclamp": [1, 3], "tensorclamp": [1, 3, 4, 16, 22], "misc": 1, "ident": [1, 3, 4, 5, 7, 16, 19, 22], "inplacelogtwo": [1, 3], "logtwo": [1, 3], "poweroftwo": [1, 3], "ops_st": [1, 10, 16], "ceilst": [1, 3], "dpuroundst": [1, 3], "floorst": [1, 3], "inplacetensorclampst": [1, 3], "roundst": [1, 2, 3, 4, 16, 22], "roundtozerost": [1, 3], "scalarclampminst": [1, 3, 16], "tensorclampst": [1, 3], "shape": [1, 5, 12, 16, 18, 20, 22], "overbatchoveroutputchannelview": [1, 3], "overbatchovertensorview": [1, 3, 5], "overoutputchannelview": [1, 3], "overtensorview": [1, 3, 16, 22], "permutedim": [1, 3], "statsinputviewshapeimpl": [1, 3], "over_batch_over_output_channel": [1, 3, 7], "over_batch_over_tensor": [1, 3, 7], "over_output_channel": [1, 3, 7], "over_tensor": [1, 3, 7], "quant": [1, 12, 16, 18, 19, 20, 21, 22], "binari": [1, 7, 8, 10, 15, 20], "binaryqu": [1, 4, 10, 16], "clampedbinaryqu": [1, 4, 10, 16], "delai": [1, 16, 22], "delaywrapp": [1, 4, 16, 22], "int": [1, 2, 5, 7, 16, 18, 20, 21, 22], "decoupledrescalingintqu": [1, 4], "decoupledrescalingintquantwithinput": [1, 4], "prescaledrestrictintqu": [1, 4], "prescaledrestrictintquantwithinputbitwidth": [1, 4], "rescalingintqu": [1, 4, 16, 22], "truncintqu": [1, 4], "int_bas": 1, "decoupledintqu": [1, 4], "max_int": [1, 4, 7], "min_int": [1, 4, 7], "to_int": [1, 4], "intquant": [1, 4, 16, 22], "ternari": 1, "ternaryqu": [1, 4], "scale": [1, 4, 10, 12, 13, 14, 16, 18, 19, 20, 21], "int_scal": 1, "intscal": [1, 4, 5, 16, 22], "poweroftwointsc": [1, 5], "runtim": [1, 10, 12, 14, 15, 21, 22], "runtimestatssc": [1, 5], "statsfromparametersc": [1, 5], "standalon": [1, 18, 22], "constscal": [1, 4, 5, 16], "parameterfromruntimestatssc": [1, 5, 16, 22], "state_dict": [1, 5, 10, 16, 20, 22], "training_forward": [1, 5], "parameterfromstatsfromparametersc": [1, 5], "parametersc": [1, 5, 16, 22], "stat": [1, 10, 16, 22], "stats_op": 1, "absav": [1, 6], "absmax": [1, 5, 6, 14, 22], "absmaxav": [1, 6], "absmaxl2": [1, 6], "absminmax": [1, 6], "abspercentil": [1, 6, 16, 22], "klminimizerthreshold": [1, 6], "smooth_normalize_distribut": [1, 6], "l1norm": [1, 6], "l2norm": [1, 6], "mse": [1, 5, 6], "evaluate_loss": [1, 6], "mse_fib_search": [1, 6], "mse_grid_search": [1, 6], "mse_loss_fn": [1, 6], "mse_search": [1, 6], "meanlearnedsigmastd": [1, 6], "meansigmastd": [1, 6], "negativeminorzero": [1, 6], "negativepercentileorzero": [1, 6], "percentileinterv": [1, 6], "stats_wrapp": 1, "view_wrapp": 1, "class": [1, 2, 3, 4, 5, 6, 8, 10, 12, 16, 18, 22], "floatrestrictvalu": [1, 16, 22], "sourc": [1, 2, 3, 4, 5, 6, 7, 8, 10, 16, 20, 21, 22], "base": [1, 2, 3, 4, 5, 6, 8, 10, 14, 15, 16, 18, 19, 20], "x": [1, 3, 4, 5, 6, 7, 8, 11, 12, 16, 21, 22], "defin": [1, 2, 3, 4, 5, 6, 10, 16, 18, 19, 20, 21], "comput": [1, 2, 3, 4, 5, 6, 7, 10, 11, 12, 16, 18, 20, 21, 22], "perform": [1, 2, 3, 4, 5, 6, 7, 10, 12, 15, 16, 18, 19, 20, 21, 22], "everi": [1, 2, 3, 4, 5, 6, 10, 12, 15, 22], "call": [1, 2, 3, 4, 5, 6, 10, 12, 15, 16, 19, 21], "should": [1, 2, 3, 4, 5, 6, 10, 11, 16, 21, 22], "overridden": [1, 2, 3, 4, 5, 6], "all": [1, 2, 3, 4, 5, 6, 7, 10, 12, 16, 18, 19, 20, 21, 22], "subclass": [1, 2, 3, 4, 5, 6, 16], "rtype": [1, 2, 3, 4, 5, 6], "tensor": [1, 2, 3, 4, 5, 6, 7, 10, 12, 14, 16, 18, 19, 20, 21], "although": [1, 2, 3, 4, 5, 6, 18, 19, 22], "recip": [1, 2, 3, 4, 5, 6], "pass": [1, 2, 3, 4, 5, 6, 10, 12, 18, 19, 20, 21, 22], "need": [1, 2, 3, 4, 5, 6, 7, 10, 12, 16, 18, 21, 22], "within": [1, 2, 3, 4, 5, 6, 10, 12, 14, 16, 18, 20, 21, 22], "thi": [1, 2, 3, 4, 5, 6, 7, 10, 12, 14, 15, 16, 18, 19, 20, 21, 22], "function": [1, 2, 3, 4, 5, 6, 8, 9, 12, 15, 16, 18, 19, 20, 21, 24], "one": [1, 2, 3, 4, 5, 6, 7, 10, 12, 16, 18, 19, 20, 21, 22], "instanc": [1, 2, 3, 4, 5, 6, 20, 21, 22], "afterward": [1, 2, 3, 4, 5, 6, 22], "instead": [1, 2, 3, 4, 5, 6, 10, 16, 18, 20, 21], "sinc": [1, 2, 3, 4, 5, 6, 7, 15, 16, 18, 19, 21, 22], "former": [1, 2, 3, 4, 5, 6, 16], "take": [1, 2, 3, 4, 5, 6, 10, 12, 16, 18, 19, 20, 21, 22], "care": [1, 2, 3, 4, 5, 6, 16, 22], "run": [1, 2, 3, 4, 5, 6, 10, 11, 13, 16, 18, 20, 22], "regist": [1, 2, 3, 4, 5, 6, 22], "hook": [1, 2, 3, 4, 5, 6], "while": [1, 2, 3, 4, 5, 6, 7, 8, 10, 12, 15, 16, 19, 20, 21, 22], "latter": [1, 2, 3, 4, 5, 6], "silent": [1, 2, 3, 4, 5, 6, 14], "ignor": [1, 2, 3, 4, 5, 6, 14, 22], "them": [1, 2, 3, 4, 5, 6, 10, 12, 16, 21, 22], "restrict_init_float": 1, "return": [1, 2, 4, 5, 7, 10, 12, 16, 18, 19, 20, 21, 22], "type": [1, 2, 4, 5, 7, 10, 12, 13, 16, 18, 21, 22], "float": [1, 2, 4, 5, 7, 10, 11, 14, 15, 16, 18, 19, 20, 21], "restrict_init_inplace_modul": 1, "restrict_init_modul": 1, "restrict_init_tensor": 1, "intrestrictvalu": [1, 2], "restrict_value_float_to_int_impl": 1, "logfloatrestrictvalu": 1, "poweroftworestrictvalu": [1, 5], "parameterwrapp": 1, "valu": [1, 2, 5, 7, 10, 12, 16, 18, 19, 20, 21, 22], "singleargstatelessbuff": 1, "placehold": [1, 5, 21], "slicetensor": 1, "eager_forward": 1, "statelessbuff": [1, 16, 22], "destin": [1, 5], "none": [1, 2, 3, 4, 5, 6, 7, 8, 10, 16, 18, 19, 20, 21, 22], "prefix": [1, 5, 10, 16, 20, 22], "keep_var": [1, 5], "fals": [1, 2, 5, 6, 7, 10, 14, 16, 18, 19, 20, 21, 22], "dictionari": [1, 5], "contain": [1, 5, 7, 10, 14, 16, 22], "refer": [1, 5, 13, 16], "whole": [1, 5, 16], "state": [1, 2, 5, 10, 12, 14, 16, 20, 22], "both": [1, 5, 10, 12, 13, 15, 16, 18, 19, 20, 21, 22], "persist": [1, 5], "buffer": [1, 5, 22], "e": [1, 5, 10, 11, 14, 16, 18, 19, 21, 22, 24], "g": [1, 5, 10, 11, 14, 16, 18, 21, 22, 24], "averag": [1, 5, 16, 19, 21, 22], "ar": [1, 4, 5, 8, 9, 10, 12, 15, 16, 18, 19, 20, 21, 22, 24], "includ": [1, 5, 10, 11, 16, 19, 22], "kei": [1, 5, 14, 16, 20, 22], "correspond": [1, 4, 5, 7, 14, 15, 22], "name": [1, 5, 10, 16, 18, 20, 22], "set": [1, 2, 4, 5, 10, 11, 12, 13, 15, 16, 18, 19, 20, 21, 22, 24], "The": [1, 2, 5, 7, 10, 11, 12, 13, 14, 15, 16, 19, 20, 21, 22], "object": [1, 3, 5, 10, 16], "i": [1, 2, 3, 4, 5, 7, 8, 10, 11, 12, 13, 14, 15, 16, 18, 19, 20, 21, 24], "shallow": [1, 5], "copi": [1, 5], "It": [1, 5, 10, 13, 16, 18, 21, 22], "": [1, 2, 5, 7, 10, 11, 12, 13, 14, 15, 19, 20, 21, 22, 24], "current": [1, 5, 10, 11, 13, 15, 16, 18, 19, 20, 21, 22, 24], "also": [1, 5, 10, 16, 18, 19, 20, 21, 22], "accept": [1, 5, 12, 16, 19, 20, 22], "posit": [1, 5, 19], "argument": [1, 5, 7, 10, 16, 18, 19, 20, 21], "order": [1, 5, 7, 8, 10, 12, 16, 21, 22, 24], "howev": [1, 5, 10, 16, 18, 19, 20, 21, 22], "being": [1, 5, 10, 11, 14, 16, 18, 20, 21, 22], "deprec": [1, 5, 20], "keyword": [1, 5, 10, 16, 19, 20, 21], "enforc": [1, 5, 10, 21, 22], "futur": [1, 5, 20], "releas": [1, 5, 15, 22, 24], "pleas": [1, 5, 20, 21, 22], "avoid": [1, 2, 5, 7, 15, 16, 22], "us": [1, 3, 4, 5, 7, 10, 11, 12, 14, 16, 18, 19, 20, 21, 22], "design": [1, 5, 10, 11, 15, 16, 22], "end": [1, 5, 7, 11, 12, 15, 16, 18, 19, 20, 22], "user": [1, 5, 10, 12, 15, 16, 18, 19, 20, 21, 22], "dict": [1, 2, 5, 10, 14, 16, 22], "option": [1, 5, 16, 18, 20, 21, 22], "If": [1, 5, 10, 16, 18, 21, 22], "provid": [1, 5, 7, 10, 11, 12, 13, 15, 16, 18, 22], "updat": [1, 5, 14, 16, 21, 22], "same": [1, 3, 5, 10, 11, 16, 18, 19, 20, 21, 22], "otherwis": [1, 5, 10, 18, 20, 22], "an": [1, 3, 4, 5, 7, 10, 11, 12, 14, 15, 18, 20, 22, 24], "ordereddict": [1, 5], "creat": [1, 5, 10, 16, 21, 22], "default": [1, 2, 4, 5, 10, 12, 14, 15, 16, 18, 19, 20, 21], "str": [1, 5, 20, 21], "ad": [1, 5, 18, 20, 21], "compos": [1, 5, 10], "bool": [1, 2, 4, 5, 7, 20, 21, 22], "detach": [1, 5, 21, 22], "from": [1, 2, 4, 5, 7, 10, 14, 16, 18, 19, 20, 21], "autograd": [1, 5, 7, 8, 10, 15], "true": [1, 3, 4, 5, 7, 10, 12, 14, 16, 18, 19, 20, 21, 22], "exampl": [1, 2, 3, 4, 5, 7, 8, 10, 12, 19, 20, 21, 22, 24], "xdoctest": [1, 5], "skip": [1, 5, 10, 12, 16, 19], "undefin": [1, 5], "var": [1, 5], "bia": [1, 5, 10, 12, 18, 20], "weight": [1, 4, 5, 6, 7, 10, 14, 18, 21], "inplace_momentum_upd": 1, "momentum": [1, 5], "counter": 1, "new_count": 1, "inplace_tensor_add": 1, "inplace_tensor_mul": 1, "parameterfromruntimezeropoint": 1, "collect_stats_step": [1, 5, 22], "int_quant": [1, 4, 16, 22], "quantize_zero_point": 1, "zero_point_stats_impl": 1, "zero_point_shap": 1, "zero_point_stats_input_view_shape_impl": 1, "zero_point_stats_momentum": 1, "0": [1, 2, 3, 4, 5, 6, 7, 8, 10, 12, 14, 16, 18, 19, 20, 21, 22], "1": [1, 2, 4, 5, 7, 8, 10, 12, 14, 15, 16, 18, 19, 20, 21, 22], "dtype": [1, 2, 5, 6, 12, 18, 22], "devic": [1, 2, 5, 6], "parameterfromstatsfromparameterzeropoint": 1, "zero_point_stats_input_concat_dim": 1, "tracked_parameter_list": [1, 5], "scriptmodul": [1, 2, 3, 4, 5, 6, 16, 22], "implement": [1, 3, 4, 5, 6, 7, 8, 10, 11, 13, 14, 15, 16, 19, 21, 22, 24], "learn": [1, 2, 5, 10, 14, 16, 19, 21], "factor": [1, 4, 5, 10, 16, 18, 19, 21, 22], "initi": [1, 2, 5, 10, 14, 18, 19, 20, 21, 22], "statist": [1, 5, 10, 19, 22], "parameterzeropoint": 1, "zero_point_init": 1, "statsfromparameterzeropoint": 1, "zerozeropoint": [1, 4, 16, 22], "constant": [2, 5, 18, 22], "bit": [2, 4, 7, 10, 12, 16, 18, 19, 20, 21, 24], "width": [2, 4, 7, 10, 18, 20, 21, 24], "wrap": [2, 4, 5, 8, 10, 16, 21], "torch": [2, 3, 4, 5, 7, 8, 10, 12, 14, 15, 16, 18, 19, 20, 21], "8": [2, 3, 5, 7, 12, 15, 16, 19, 20, 21, 22], "part": [2, 7, 10, 15, 16, 22], "mean": [2, 5, 10, 12, 16, 18, 19, 20, 21, 22], "won": [2, 18, 19], "t": [2, 4, 7, 8, 10, 11, 12, 16, 18, 19, 20, 21, 22], "save": [2, 18, 20], "checkpoint": 2, "map": [2, 4, 5, 16, 22], "bit_width_impl_typ": [2, 22], "bitwidthimpltyp": [2, 22], "higher": [2, 4, 5, 16, 18], "level": [2, 4, 5, 10, 13, 16, 20, 22], "api": [2, 4, 5, 10, 13, 18, 22], "retain": [2, 5], "counterpart": [2, 10, 18], "differ": [2, 3, 5, 10, 12, 13, 16, 18, 19, 20, 21, 22], "stateful_const": 2, "bit_width_to_remove_impl": 2, "min_overall_bit_width": 2, "max_overall_bit_width": 2, "input_bit_width": [2, 4, 7, 21, 22], "min_bit_width": 2, "2": [2, 3, 4, 5, 7, 12, 15, 16, 18, 19, 20, 21, 22], "restrict_bit_width_impl": 2, "float_to_int_impl": [2, 4, 16, 22], "override_pretrained_bit_width": 2, "learnabl": 2, "output": [2, 4, 5, 7, 10, 12, 16, 18, 19, 20], "lower": [2, 5, 7, 20], "bound": [2, 5, 7], "restrict": [2, 4, 5, 7, 10, 12, 18, 21, 22, 24], "subset": [2, 16, 24], "pretrain": [2, 10, 12, 14, 16, 20, 22], "load": [2, 10, 14, 15, 16, 22], "backend": [2, 7, 10, 11, 14], "nn": [2, 5, 10, 14, 16, 18, 19, 20, 21, 22], "rais": [2, 5, 16, 18, 19, 20, 21, 22], "runtimeerror": [2, 5, 16, 19, 20, 21, 22], "bit_width_paramet": 2, "grad_fn": [2, 4, 5, 7, 16, 19, 20, 21, 22], "roundstefnbackward": [2, 7, 22], "env": [2, 4, 5, 7, 8, 15, 16, 20, 21, 22], "variabl": [2, 4, 5, 14, 15, 16, 20, 22], "brevitas_ignore_missing_kei": [2, 5, 14, 16, 22], "error": [2, 5, 14, 16, 18, 19, 20, 21, 22], "when": [2, 4, 5, 7, 8, 10, 14, 15, 16, 18, 19, 20, 21, 22], "retrain": [2, 5], "point": [1, 2, 4, 5, 10, 11, 12, 14, 15, 16, 18, 19, 20, 21], "bit_width_to_remov": 2, "non_zero_epsilon": 2, "1e": [2, 6, 20], "06": 2, "remove_zero_bit_width": 2, "wrapper": [3, 7, 10, 22], "variou": [3, 7, 8, 10, 12, 13, 16, 21, 22, 24], "variant": [3, 7, 10, 12, 16, 19, 20, 22], "min_val": [3, 7, 8, 18, 19], "clamp_min": [3, 7, 8], "3": [3, 4, 5, 7, 10, 12, 15, 16, 18, 19, 20, 21, 22], "max_val": [3, 7, 8, 18, 19], "scalar_clamp": 3, "tensor_clamp": [3, 7, 8], "A": [3, 7, 10, 12, 18, 21, 22], "collect": [3, 5, 6, 10, 19, 22], "miscellan": 3, "quantiz": [3, 4, 6, 7, 11, 13, 14, 15, 18, 24], "randn": [3, 7, 12, 16, 18, 19, 20, 21, 22], "size": [3, 10, 11, 20, 21, 22], "y": [3, 4, 7, 8, 11, 16], "log2_": 3, "inplace_log_two": 3, "note": [3, 7, 12, 16, 18, 20, 21, 22, 24], "inplac": 3, "oper": [3, 7, 11, 12, 13, 15, 18, 19, 20, 21, 24], "torchscript": [3, 4, 5, 7, 10, 14, 16, 20], "problemat": 3, "compil": [3, 4, 7, 10, 13, 14, 15, 16, 22], "disabl": [3, 10, 14, 19, 20, 21, 22], "log2": [3, 22], "log_two": 3, "power_of_two": 3, "5": [3, 4, 7, 12, 16, 18, 19, 20, 21, 22], "32": [3, 12, 16, 22], "ceil_st": [3, 7, 8], "dpu_round_st": [3, 7, 8], "floor_st": [3, 7, 8], "tensor_clamp_ste_": [3, 7], "round_st": [3, 7, 8], "round_to_zero_st": [3, 7, 8], "scalar_clamp_min_st": [3, 7, 8], "tensor_clamp_st": [3, 7], "view": [3, 5, 7, 16, 19], "accord": [3, 4, 5, 7, 10, 12, 16, 21, 22], "criteria": [3, 5, 12], "permute_dim": 3, "input": [3, 4, 5, 7, 10, 12, 16, 18, 19, 20], "view_modul": 3, "empti": [3, 5, 10], "25": [3, 16, 20, 22], "scriptmoodul": 3, "250": 3, "16": [3, 12, 16, 19, 20, 22], "200": [3, 18], "6": [3, 4, 5, 12, 16, 18, 19, 20, 21, 22], "2400": 3, "enum": 3, "like": [3, 4, 10, 12, 16, 19, 20, 21, 22], "pointer": [3, 16, 19, 22], "adher": [3, 7, 10, 11], "interfac": [3, 8, 10, 12, 18, 19, 20, 22], "alia": [3, 8, 22], "scaling_impl": [4, 5, 16, 22], "quant_delay_step": [4, 16], "uniform": [4, 10, 11, 13, 16, 20], "binary_sign_st": [4, 7, 8, 16], "number": [4, 5, 7, 10, 16, 19, 20, 22], "train": [4, 5, 10, 11, 12, 13, 16, 19, 20, 21], "step": [4, 16, 18, 19, 22], "de": [4, 10, 16], "format": [4, 10, 11, 12, 16, 18, 19, 21, 22], "zero": [1, 4, 7, 10, 16, 18, 19, 20, 21, 22], "tupl": [4, 5, 7, 10, 16, 18, 20, 21, 22], "import": [4, 12, 16, 18, 19, 20, 21, 22], "binary_qu": [4, 16], "inp": [4, 16, 18, 19, 20, 21, 22], "04": [4, 16, 21], "out": [4, 12, 13, 16, 18, 19, 21, 22], "zero_point": [4, 16, 19, 20, 21, 22], "1000": [4, 7, 16, 20, 22], "quant_typ": [4, 10, 16, 22], "quanttyp": [4, 10, 16, 22], "appli": [4, 5, 7, 8, 10, 13, 16, 19, 20, 21, 22], "brevitas_jit": [4, 7, 8, 10, 11, 14, 15, 16, 20, 22], "enabl": [4, 7, 10, 11, 12, 14, 15, 16, 19, 21, 22], "tensor_clamp_impl": [4, 16, 22], "befor": [4, 5, 16, 18, 20, 21, 22], "go": [4, 5, 10, 16, 18, 19, 20, 21, 22], "through": [4, 7, 8, 10, 12, 14, 15, 16, 18, 19, 20, 22, 24], "between": [4, 7, 10, 12, 16, 18, 19, 21, 22], "which": [4, 7, 10, 11, 14, 16, 18, 19, 20, 21, 22], "backward": [4, 7, 8], "gradient": [4, 7, 8, 14, 21], "outsid": [4, 11, 16, 22], "rang": [4, 18, 19, 22], "wise": [4, 10, 11], "requires_grad_": 4, "mulbackward0": [4, 16, 19, 21, 22], "grad": [4, 7], "0000": [4, 7, 19, 20, 22], "activ": [4, 5, 10, 14, 18, 20, 21], "decoupled_int_qu": 4, "pre_scaling_impl": 4, "int_scaling_impl": [4, 16, 22], "pre_zero_point_impl": 4, "zero_point_impl": [4, 16, 22], "bit_width_impl": [4, 6, 22], "input_is_sign": [4, 7], "around": [4, 10, 12, 16, 22], "integ": [4, 7, 11, 18, 19, 21, 22], "taken": [4, 22], "narrow_rang": [4, 5, 7, 18, 22], "sign": [4, 5, 6, 7, 8, 10, 16, 18, 19, 21, 22], "int_quant_wrapp": 4, "01": [4, 16], "4": [4, 5, 7, 12, 16, 18, 19, 20, 21, 22], "042": 4, "053": 4, "31": [4, 16, 22], "44": [4, 22], "0400": 4, "0500": 4, "0700": [4, 21], "0100": [4, 16], "respect": [4, 10, 12, 16, 21, 22], "here": [4, 16, 22], "interpret": [4, 18], "threshold": 4, "0429": 4, "0571": 4, "0143": 4, "shift": [4, 19], "pre": [1, 4, 10, 16, 22], "flag": [4, 14, 16, 20, 21, 22], "determin": 4, "whether": [4, 7, 10, 14, 16, 18, 21], "narrow": [4, 7, 18], "convers": [4, 19, 20], "represent": [4, 7, 10, 12, 18, 19, 21, 22], "pre_scal": 4, "pre_zero_point": 4, "02": [4, 21], "0200": 4, "0300": [4, 20], "ternary_sign_st": [4, 7, 8], "ternar": 4, "w": [4, 7, 8, 10, 11], "r": [4, 7, 8, 10], "ternary_qu": 4, "scaling_stats_impl": [5, 22], "scaling_stats_input_view_shape_impl": [5, 22], "restrict_scaling_impl": [5, 22], "scaling_shap": [5, 16, 22], "affine_resc": 5, "affine_shift_scal": 5, "scaling_stats_momentum": 5, "scaling_min_v": 5, "scaling_stats_input_concat_dim": 5, "scaling_init": [5, 16, 22], "union": [5, 21, 22], "some": [5, 10, 11, 12, 15, 18, 19, 20, 21, 22], "forc": [5, 10, 14, 20, 21, 22], "method": [5, 16, 20, 21], "singl": [5, 10, 13, 15, 16, 18, 20, 22], "requir": [5, 10, 14, 16, 19, 20, 21, 22, 24], "earli": 5, "version": [5, 10, 15, 16, 18, 20, 22], "consist": [5, 16], "across": [5, 10, 12, 13, 20, 22], "scaling_impl_typ": [5, 22], "scalingimpltyp": [5, 10, 22], "permute_impl": 5, "work": [5, 10, 16, 20, 22], "two": [5, 10, 12, 16, 18, 19, 20, 21, 22], "phase": [5, 19, 22], "dure": [5, 11, 16, 18, 19, 22], "first": [5, 8, 10, 12, 16, 18, 19, 20, 21, 22], "fashion": [5, 22], "batchnorm": 5, "mode": [5, 10, 19, 21, 22], "per": [5, 6, 16, 18, 20, 21], "batch": [5, 7, 11, 18, 19, 20, 22], "background": [5, 16], "infer": [5, 10, 11, 12, 13, 22, 24], "second": [5, 16, 19, 20, 21, 22], "accumul": [5, 7, 10, 19, 21, 22], "behaviour": [5, 10, 14, 16, 19, 20, 21, 22], "scalar_shap": 5, "move": [5, 19, 22], "default_momentum": 5, "scaling_stats_permute_dim": [5, 22], "arang": 5, "randn_lik": 5, "absbinarysigngradfnbackward": [5, 7, 16], "parameter_from_stat": [5, 10, 22], "stats_input": 5, "extend": [5, 10, 12, 16, 18, 22], "scalar": [5, 7, 18], "non": [5, 10, 13, 16, 19, 21], "powbackward1": 5, "stats_reduce_dim": [1, 6, 22], "keepdim": 6, "high_percentile_q": [6, 22], "percentile_q": 6, "num_bin": 6, "1001": 6, "smoothing_ep": 6, "0001": [6, 21], "apach": 6, "incub": 6, "mxnet": 6, "p": [6, 20], "ep": 6, "channel": [6, 7, 10, 11, 16, 18, 20], "l1": [6, 7], "normal": [6, 16, 18, 19, 22], "l2": 6, "proxy_modul": 6, "mse_init_op": 6, "inner_stats_input_view_shape_impl": 6, "mse_search_method": 6, "fibonacci": 6, "mse_it": 6, "20": [6, 16, 19, 20, 22], "candid": 6, "xl": 6, "xr": 6, "quant_valu": [6, 21], "sigma": 6, "stats_output_shap": 6, "std_dev_epsilon": 6, "08": 6, "low_percentile_q": 6, "core": [7, 9, 16, 22], "often": [7, 16], "impos": [7, 10, 16, 24], "pytorch": [7, 10, 11, 12, 13, 15, 16, 18, 20, 21, 22, 24], "binary_sign": [7, 8], "dpu_round": [7, 8], "dpu": [7, 24], "round": [7, 8, 21, 22], "get_upper_bound_on_l1_norm": 7, "accumulator_bit_width": 7, "calcul": 7, "upper": 7, "norm": [7, 18], "deriv": [7, 10, 12, 22], "neural": [12, 22], "network": [12, 20, 22], "low": [11, 12, 18, 21, 22, 24], "precis": [10, 11, 12, 13, 18, 22], "guarante": 7, "overflow": 7, "colbert": 7, "j": 7, "petri": 7, "koenig": 7, "THe": 7, "7": [7, 16, 18, 19, 20, 21, 22], "max_float": 7, "exponent_bit_width": 7, "mantissa_bit_width": 7, "exponent_bia": 7, "maximum": [7, 16, 22], "given": [7, 18, 20, 21, 22], "indic": 7, "repres": [7, 10, 12, 18, 20, 21, 22], "unsign": [7, 12, 18, 19, 22], "avail": [7, 16, 22], "127": [7, 18, 21, 22], "254": 7, "255": 7, "minimum": [7, 18], "128": [7, 18, 22], "round_to_zero": [7, 8], "toward": [7, 10, 11], "gener": [7, 10, 11, 13, 14, 15, 16, 18, 19, 20, 21, 22, 24], "support": [7, 10, 11, 12, 13, 14, 15, 16, 18, 20, 21, 22, 24], "broadcast": 7, "differenti": [7, 22], "tensor_clamp_": [7, 8], "In": [7, 10, 12, 14, 16, 18, 19, 20, 21, 22, 24], "place": [7, 12, 16, 22], "Not": [7, 21], "wrt": 7, "ani": [7, 10, 11, 12, 14, 16, 18, 19, 20, 21, 22, 24], "straight": [7, 8, 10, 16], "estim": [7, 8, 10, 14], "dispatch": 7, "either": [7, 10, 12, 16, 22], "nativ": [7, 14], "just": [7, 10, 12, 14, 16, 19, 21, 22], "time": [7, 10, 11, 12, 13, 14, 15, 16, 18, 21, 22], "autograd_ste_op": [7, 10], "allow": [7, 10, 16, 18, 19, 20, 21, 22], "built": [7, 10, 14, 18, 22], "abs_binary_sign_grad": [7, 8], "ab": [7, 8, 16, 22], "have": [7, 8, 10, 11, 12, 16, 18, 19, 20, 21, 22], "subgradi": [7, 8], "compar": [7, 8, 12, 18, 19, 20], "abs_binary_sign_grad_impl": [7, 8], "its": [7, 10, 11, 12, 16, 18, 19, 20, 21, 22], "requires_grad": [7, 22], "item": [7, 16, 19, 21], "binary_sign_ste_impl": [7, 8], "binarysignstefnbackward": 7, "ceil": [7, 8], "ceil_ste_impl": [7, 8], "ceilstefnbackward": [7, 22], "dpu_round_ste_impl": [7, 8], "dpuroundstefnbackward": 7, "floor": [7, 8], "floor_ste_impl": [7, 8], "floorstefnbackward": 7, "round_ste_impl": [7, 8], "round_to_zero_ste_impl": [7, 8], "roundtozerostefnbackward": 7, "alwai": [7, 8, 10, 12, 18, 19, 20, 21, 22], "scalar_clamp_min_ste_impl": [7, 8], "c": [7, 10, 11, 14, 15, 16, 19, 20, 22], "5000": 7, "4000": 7, "scalarclampminstefnbackward": 7, "scalar_clamp_st": [7, 8], "scalar_clamp_ste_impl": [7, 8], "scalarclampstefnbackward": 7, "tensor_clamp_ste_impl": [7, 8], "tensorclampstefnbackward": 7, "tensor_clamp_ste_impl_": [7, 8], "inplacetensorclampstefnbackward": 7, "ternary_sign_ste_impl": [7, 8], "ternarysignstefnbackward": 7, "induc": 7, "flatten": 7, "along": [7, 10], "certain": [7, 10, 13, 16, 20, 22, 24], "dimens": 7, "dim": [7, 16, 18, 20, 22], "other": [7, 10, 12, 16, 18, 19, 20, 21, 22, 24], "featur": [7, 10, 16, 18, 20, 22, 24], "arg": [7, 8, 16, 18, 22], "flat": [7, 10, 16], "absbinarysigngradfn": 8, "kwarg": [8, 12, 16, 19, 20, 21, 22], "alias": 8, "see": [8, 11, 16, 18, 19, 21, 22], "detail": [8, 10, 12], "binarysignstefn": 8, "ceilstefn": 8, "dpuroundstefn": 8, "floorstefn": 8, "inplacetensorclampstefn": 8, "roundstefn": 8, "roundtozerostefn": 8, "scalarclampminstefn": 8, "invok": [8, 10, 22], "scalarclampstefn": 8, "tensorclampstefn": 8, "ternarysignstefn": 8, "wip": 9, "most": [9, 10, 16, 19, 21, 22], "packag": [9, 10, 14, 15, 16, 21, 22], "still": [9, 11, 16, 18, 20, 22], "miss": [9, 14, 16, 19, 22], "op": [9, 10, 20, 21], "organ": [10, 20], "few": [10, 14, 16, 18, 20, 21, 22], "concept": [10, 20], "list": [10, 16, 19], "below": [10, 22], "found": [10, 14, 16, 20, 21, 22], "under": [10, 13, 16, 21, 22], "python": [10, 14, 15, 16, 18, 20, 21, 22], "csrc": [10, 20], "becaus": [10, 15, 16, 18, 19, 20, 21, 22, 24], "date": 10, "cannot": [10, 14, 16, 22], "jit": [10, 14, 15, 16, 20, 21, 22], "extens": [10, 16, 18, 22], "simplifi": [10, 15], "distribut": [10, 15], "onli": [10, 16, 18, 19, 21, 22, 24], "cpp": [10, 15, 20], "file": [10, 15], "appropri": [10, 14, 15, 16, 22], "so": [10, 12, 14, 15, 16, 18, 19, 20, 21, 22], "fallback": 10, "long": 10, "switch": [10, 16, 22], "happen": [10, 12, 14, 16, 18, 19, 21, 22], "piec": [10, 16, 22], "commonli": [10, 12], "thei": [10, 16, 18, 20, 21, 22], "algorithm": [10, 11, 12, 16, 19, 22], "build": [10, 13, 19, 22], "block": [10, 13], "assembl": [10, 16], "affin": [10, 11, 13, 16, 21, 22], "old": 10, "style": [10, 13, 16, 18, 19, 22], "scripe": 10, "inherit": 10, "mani": [10, 16, 21, 22], "describ": 10, "section": [10, 11, 12, 22], "abov": [10, 18, 21, 22], "modul": [10, 12, 16, 18, 19, 20, 21, 22], "compat": [10, 15], "everyth": [10, 16, 20, 22], "memori": 10, "optim": [10, 16, 18, 20], "more": [10, 11, 12, 15, 16, 18, 20, 22, 24], "complic": [10, 16], "pipelin": 10, "quit": [10, 16, 19, 20, 21], "signific": 10, "thu": [10, 18], "reduc": [10, 11, 13], "intrins": 10, "cost": [10, 11, 22], "awar": [7, 10, 11, 12, 13, 16, 20, 21, 22], "pose": [10, 16], "challang": 10, "term": [10, 11, 12, 15, 16, 21, 22], "how": [10, 11, 12, 16, 18, 20, 21], "achiev": [10, 11, 16], "flexibl": [10, 11, 16], "minim": [10, 19, 20, 22], "redund": 10, "flavour": [10, 22], "adopt": [10, 13, 15, 16, 19, 20, 21, 22], "doe": [10, 11, 12, 16, 18, 22], "highli": 10, "bias": [10, 11, 16, 21, 22], "leverag": [10, 12, 16, 18, 20, 22], "composit": [10, 22], "particular": [10, 11, 16, 18, 19, 20, 22], "favour": [10, 22], "invers": 10, "control": [10, 14, 16, 20, 22], "depend": [10, 19, 20, 21], "inject": 10, "di": [10, 16], "manual": [10, 19, 22], "librari": [10, 12, 16, 18, 20, 22], "explain": [10, 19, 21], "auto": [10, 21, 22], "wire": [10, 22], "machineri": 10, "heart": 10, "ever": 10, "fixtur": 10, "pytest": 10, "alreadi": [10, 16, 19, 20, 21, 22], "know": [10, 16], "high": [10, 21], "case": [10, 12, 14, 16, 18, 19, 20, 21, 22], "idea": [10, 16, 22], "instanti": [10, 12, 16, 20, 22], "togeth": [10, 15, 16, 19, 20, 21, 22], "declar": [10, 16, 22], "attribut": [10, 16, 18, 19, 22], "drive": [10, 22], "mechan": [10, 15, 22], "behind": [10, 16], "process": [10, 16, 22], "match": [10, 12, 16, 18, 20, 22], "throw": 10, "bunch": [10, 16, 20], "compon": [10, 16, 22], "chosen": 10, "automat": [10, 12, 16, 18, 21, 22], "tensor_qu": [10, 16, 19, 21, 22], "expect": [10, 16, 19, 20, 21, 22], "four": [10, 20, 21], "dequant": [10, 12, 18, 19, 21, 22], "power": [10, 16, 22], "wai": [10, 12, 16, 18, 19, 20, 21, 22, 24], "express": [10, 16, 22], "standard": [10, 11, 12, 15, 16, 18, 19, 20, 21, 22, 24], "lend": 10, "themselv": 10, "multipl": [10, 12, 16, 20, 22], "That": [10, 11, 16, 19, 21, 22], "new": [10, 13, 16, 22], "simpli": [10, 16, 21, 22], "exist": [10, 16, 19, 22], "overrid": [10, 16, 18, 21, 22], "smaller": [10, 22], "specif": [10, 11, 16, 22], "shifted_scaled_int": [10, 19], "hold": 10, "scaled_int": [10, 16, 18, 19, 21, 22], "fixed_point": 10, "fix": [10, 16], "older": [10, 16], "v2": 10, "plai": 10, "well": [10, 12, 13, 21, 22], "extendedinjector": [10, 22], "addit": [10, 18, 20, 21, 22, 24], "dynam": 10, "syntax": [10, 16, 19], "intermedi": [10, 12, 18, 22], "abstract": [10, 13, 16, 22], "try": [10, 11, 16, 19, 20, 21, 22], "navig": 10, "confus": 10, "lack": 10, "clear": [10, 22], "hierarchi": [10, 16], "obviou": [10, 21, 24], "addition": [10, 12, 19, 22], "might": [10, 16, 18, 19, 20, 21, 22], "fit": [10, 11], "pure": [10, 22], "perspect": 10, "machin": 10, "awai": [10, 22], "top": [10, 14, 15, 16, 20, 21, 22], "specifi": [10, 12, 18, 19], "kind": [10, 14, 16, 22], "sai": [10, 12, 16, 19, 21, 22], "v": [10, 16], "gonna": [10, 20, 22], "translat": [10, 22], "target": [10, 11, 13, 22, 24], "relationship": 10, "hyperparamet": [10, 16, 22], "realli": [10, 16, 19], "advantag": [10, 12, 16, 22], "scope": [10, 22], "solv": [10, 21, 22], "against": [10, 19], "enough": [10, 16], "weightquantsolv": [10, 22], "itself": [10, 11, 16, 22, 24], "individu": 10, "task": 10, "look": [10, 12, 16, 18, 19, 20, 21, 22], "seen": [10, 16, 18, 19, 21, 22], "actual": [10, 18, 21, 22], "rather": [10, 16, 18, 22], "than": [10, 11, 12, 15, 16, 18, 20, 22], "directli": [10, 15, 18, 20, 21, 22], "Then": [10, 19], "meant": 10, "blueprint": 10, "understand": [10, 16, 22], "similar": [10, 12, 16, 18, 19, 20, 21], "direct": [10, 20, 22], "scaling_per_output_channel": [10, 16, 22], "ha": [10, 12, 13, 16, 18, 20, 21, 22], "alloc": [10, 16, 20], "much": [10, 16, 21], "known": [10, 22], "capabl": [10, 16, 18], "whose": [10, 14, 22], "again": [10, 16, 19, 21], "thank": [10, 16, 22], "whenev": [10, 14, 16, 20, 22], "last": [7, 10, 16, 18, 19, 21, 22], "possibl": [10, 15, 16, 18, 20, 21, 22], "even": [10, 16, 18, 19, 21, 22], "advanc": [10, 22], "mix": [10, 12, 19, 21], "custom": [10, 13, 15, 18, 20], "final": [10, 16, 18, 19, 20, 21], "possibli": [10, 12, 16], "data": [10, 11, 12, 13, 16, 18, 20, 21, 22], "structur": [10, 12, 16, 20, 21, 22], "quant_tensor": [10, 21, 22], "valid": [10, 16, 21], "had": [10, 21], "been": [10, 11, 12, 13, 16, 20, 21, 22], "previous": [10, 21, 22], "now": [10, 12, 16, 19, 20, 21, 22], "mark": [10, 19, 21, 22], "doesn": [10, 12, 16, 18, 19, 21, 22], "carri": [10, 11, 16], "inform": [10, 12, 16, 21, 22], "back": [10, 14, 20], "wa": [10, 11, 16, 19, 20, 21, 22], "arithmet": [10, 11], "implment": 10, "main": [10, 18], "assumpt": [10, 11, 21, 22], "sum": [10, 20, 21], "constrain": 10, "involv": [10, 11, 20, 22], "deal": 10, "residu": 10, "topologi": 10, "connect": 10, "special": [10, 16, 19], "respons": 10, "make": [10, 11, 12, 15, 16, 18, 19, 20, 21, 22], "sure": [10, 15, 18, 20, 21], "wouldn": [10, 16, 21, 22], "re": [10, 14, 16, 18, 20, 21, 22], "necessari": [10, 15, 16, 20, 21], "priori": 10, "later": [10, 16, 19, 22], "model": [10, 11, 13, 14, 15, 16, 20, 22, 24], "definit": [10, 12, 16, 18, 20, 22], "logic": [10, 16], "recomput": [10, 14, 22], "chang": [10, 12, 16, 18, 21, 22], "do": [10, 11, 12, 16, 18, 19, 21, 22], "reconcil": 10, "inher": 10, "rigid": 10, "typic": [10, 11, 14, 16, 19, 21, 22], "execut": [10, 16, 18, 19, 20, 22], "complex": [10, 11, 16, 21], "scenario": [10, 12, 15, 16, 18, 19, 20, 21, 22], "share": [10, 18, 20], "situat": [10, 11, 16, 21, 22], "where": [10, 16, 18, 19, 20, 21, 22], "branch": 10, "without": [10, 12, 14, 16, 18, 21, 22], "requant": 10, "feed": 10, "accomplish": 10, "weightquantproxi": 10, "among": [10, 16, 18, 20], "what": [10, 11, 12, 18, 21, 22], "affine_stat": 10, "concaten": [10, 20], "start": [10, 13, 18, 20, 22], "track": 10, "underli": [10, 22], "quantlay": 10, "quantconv2d": [10, 12, 16, 18, 19, 22], "conv2d": [10, 16, 18, 21], "plu": [10, 20], "seri": [10, 20], "each": [10, 12, 16, 18, 20, 24], "responsibil": 10, "quantinputmixin": 10, "quantoutputmixin": 10, "_quantwbiol_": 10, "quantweightmixin": 10, "quantbiasmixin": 10, "quantwbiol": [10, 21, 22], "quant_weight": [10, 16, 20, 21, 22], "quant_bia": [10, 21, 22], "quant_input": [10, 20, 22], "quant_output": [10, 22], "suppos": 10, "act": [10, 16, 18, 19], "exactli": [10, 22], "weight_quant": [10, 16, 20, 21, 22], "bias_quant": [10, 12, 18, 20, 21, 22], "input_qu": [10, 16, 18, 19, 21, 22], "output_qu": [10, 16, 18, 19, 21, 22], "behav": [10, 19, 21, 22], "expos": [10, 16, 20, 21, 22], "To": [10, 11, 12, 15, 16, 19, 20, 21, 22], "ux": 10, "weight_": [10, 16, 22], "bias_": [10, 16, 22], "input_": [10, 16, 22], "output_": [10, 16, 22], "quantrelu": [10, 12, 16, 18, 19], "prioriti": [10, 22], "over": [10, 18, 19, 22], "real": 10, "life": 10, "want": [10, 12, 16, 21, 22], "exploratori": 10, "analysi": 10, "accuraci": 10, "correct": [10, 20, 21, 22], "hardwar": [10, 11, 13, 22, 24], "concern": [10, 11, 21], "friction": [10, 22], "remain": [10, 18], "partial": [10, 20], "With": [10, 12, 16, 18, 19, 20, 21, 22], "except": [7, 10, 15, 16, 19, 20, 21, 22], "truncavgpool2d": 10, "receiv": 10, "altough": [10, 21], "nor": [10, 20], "return_quant_tensor": [10, 12, 16, 18, 19, 20, 21, 22], "easili": [10, 16, 19, 21, 22], "todo": 10, "why": [11, 12, 16, 21, 22], "fbgemm": 11, "qnnpack": 11, "platform": [11, 13, 15, 24], "novel": [11, 22], "varieti": [11, 22], "loos": 11, "z": 11, "find": [11, 20], "document": [11, 18, 20, 24], "spars": 11, "until": [11, 22], "improv": 11, "feel": 11, "free": 11, "open": 11, "issu": [11, 16, 20, 21, 22], "ask": 11, "our": [11, 12, 16, 18, 22], "gitter": 11, "slow": [11, 15, 20], "lot": [11, 22], "element": [11, 18], "intens": 11, "contribut": [11, 16], "graph": [11, 16, 18, 22], "backproprag": 11, "As": [11, 16, 18, 19, 20, 21, 22, 24], "up": [11, 16, 19], "slower": [11, 15], "resourc": 11, "greater": 11, "effieci": 11, "principl": [11, 16, 21, 22], "trade": 11, "off": [11, 18], "effici": [11, 15], "mitig": [11, 16, 20], "somewhat": 11, "down": 11, "report": 11, "thought": 11, "my": 11, "faster": 11, "am": 11, "wrong": 11, "path": [11, 13, 18], "acceler": [11, 12, 13, 15, 18, 20, 22, 24], "own": [11, 12, 13, 14, 16, 20, 21, 22], "export": [11, 13, 15, 19], "your": [11, 14, 18, 20], "downstream": [11, 24], "toolchain": [11, 12, 15, 22, 24], "pu": 11, "float16": 11, "bfloat16": 11, "bfloat19": 11, "datatyp": [11, 22, 24], "float32": [11, 20], "test": 11, "tpu": 11, "xla": 11, "math": [11, 22], "reason": [11, 16, 20], "risk": [11, 14], "serv": [12, 16, 18, 20, 22], "goal": [12, 16, 18], "flow": [12, 22, 24], "By": [12, 15, 19, 21, 22], "write": 12, "modifi": [12, 16, 20], "origin": [12, 16, 21, 22], "onc": [12, 18, 21], "post": 12, "qat": 12, "scratch": [12, 16, 21], "finetun": [12, 16], "follow": [12, 16, 18, 19, 20, 21, 22], "combin": [12, 13, 22], "best": [12, 16], "approach": [12, 15, 18, 20, 22], "checkout": 12, "done": [12, 16, 18, 21], "imagenet": 12, "classif": 12, "torchvis": 12, "script": 12, "we": [12, 16, 18, 19, 20, 21, 22], "consid": [12, 16, 18, 19, 20, 21], "classic": 12, "lenet": 12, "let": [12, 16, 19, 21, 22], "interest": [12, 13, 16, 22], "assess": 12, "cifar10": 12, "purpos": [12, 20, 22], "tutori": [12, 13, 16, 18, 19, 20, 21], "tradit": [12, 16, 22], "quantlinear": [12, 18, 20, 21], "weight_bit_width": [12, 18, 20, 22], "relu": [12, 16, 18, 19, 22], "max": [12, 16, 18, 21, 22], "pool": [12, 21], "usual": [12, 21], "max_pool2d": [12, 21], "result": [12, 16, 18, 19, 20, 21, 22, 24], "f": [12, 18, 20, 21, 22], "qnn": [12, 18], "quantweightlenet": 12, "def": [12, 16, 18, 19, 20, 21, 22], "__init__": [12, 16, 18, 19, 20, 21, 22], "self": [12, 16, 18, 19, 20, 21, 22], "super": [12, 13, 16, 18, 19, 20, 24], "conv1": 12, "relu1": 12, "conv2": 12, "relu2": 12, "fc1": 12, "120": [12, 16, 19, 21], "relu3": 12, "fc2": 12, "84": 12, "relu4": 12, "fc3": 12, "reshap": 12, "quant_weight_lenet": 12, "storag": 12, "convert": [12, 18], "practic": [12, 19, 21, 22], "too": [12, 16, 19, 20], "keep": [12, 16, 19, 20, 22], "replac": [12, 18, 20, 22], "veri": [12, 15, 16, 19, 21, 22], "introduc": [12, 16, 18, 19, 20, 22], "quantident": [12, 18, 19, 20, 21], "begin": 12, "int8bia": [12, 21], "biasquant": 12, "quantweightactlenet": 12, "quant_inp": 12, "quant_weight_act_lenet": 12, "coupl": [12, 16, 19, 20, 21], "thing": [12, 16, 18, 19, 20, 21, 22], "fact": [12, 19, 22], "int4": 12, "15": [12, 16, 18, 19, 20, 22], "int32bia": [12, 18, 20], "quantweightactbiaslenet": 12, "lowprecisionlenet": 12, "quant_weight_act_bias_lenet": 12, "previou": [12, 16, 19, 20, 21, 22], "propag": [12, 21], "quanttensor": [12, 16, 18, 19, 20], "next": [12, 18, 19], "metadata": [12, 19, 20, 21], "about": [12, 13, 16, 19, 21, 22], "qint": 12, "friendli": [12, 22], "affect": [12, 18, 20], "bias_scal": 12, "input_scal": [12, 21, 22], "weight_scal": [12, 22], "read": 12, "invari": [12, 16, 21, 22], "extra": [12, 16, 18, 21, 22], "One": 12, "popular": [12, 16], "qdq": [12, 18, 22], "qcdq": [12, 20], "insert": [12, 18], "clip": [12, 20], "node": [12, 18, 20, 22], "export_onnx_qcdq": [12, 18, 20], "export_path": [12, 18, 20, 22], "4b_weight_lenet": 12, "4b_weight_act_lenet": 12, "4b_weight_act_bias_lenet": 12, "check": [12, 20], "recurr": [12, 20], "overview": 12, "tvmcon": 12, "setup": 13, "get": [13, 15, 16, 19, 20, 21, 22], "architectur": 13, "faq": 13, "techinqu": 13, "practition": 13, "techniqu": [13, 16], "scheme": 13, "framework": 13, "unifi": 13, "layer": [13, 14, 18, 19, 20, 21, 24], "finn": [13, 15, 21, 22, 24], "onnxruntim": [13, 15, 18, 20, 24], "successfulli": [13, 16, 20, 22], "project": 13, "larg": 13, "commerci": 13, "deploy": 13, "cpu": 13, "gpu": [13, 15], "fpga": [13, 15, 22, 24], "focu": [13, 21], "box": 13, "boolean": [14, 16, 21], "global": 14, "enviroment": 14, "config": [14, 16, 20, 22], "jit_en": 14, "written": 14, "warn": [14, 18, 20, 22], "fall": 14, "pytorch_jit": 14, "ignore_missing_kei": [14, 16, 20, 22], "intern": [14, 15, 16, 20, 21, 22], "better": 14, "altern": [14, 18, 19, 20, 22], "load_state_dict": [14, 16, 20, 22], "strict": [14, 16, 22], "would": [14, 15, 16, 18, 19, 20, 21, 22], "mismatch": 14, "brevitas_verbos": 14, "verbos": [14, 16, 22], "brevitas_native_ste_backend": 14, "native_ste_backend_en": 14, "brevitas_reinit_on_state_dict_load": 14, "reinit_on_state_dict_load": 14, "trigger": [14, 16, 20, 22], "9": [15, 16, 18, 19, 20, 21, 22], "window": 15, "linux": 15, "maco": 15, "recommend": 15, "latest": 15, "pip": [15, 18, 20, 22], "git": [15, 20, 21], "small": [15, 22], "lt": [15, 16, 18, 19, 20, 21, 22], "benefit": 15, "almost": [15, 22], "never": 15, "util": [15, 20], "cpp_extens": 15, "mantain": 15, "precompil": 15, "put": [15, 22], "burden": 15, "present": [15, 16, 18, 22], "primit": 15, "fake": [15, 18, 22], "dataflow": [15, 22, 24], "integr": [15, 22], "onnx": [15, 19, 20, 21, 24], "broad": 16, "sens": [16, 22], "anyth": [16, 19, 22], "terminologi": 16, "prefer": 16, "said": [16, 19, 20], "int8actpertensorfloat": [16, 18, 19, 20, 21, 22], "issubclass": 16, "delay_wrapp": [16, 22], "delay_impl": [16, 22], "_nodelai": [16, 22], "stats_input_view_shape_impl": [16, 22], "_stat": [16, 22], "stats_impl": [16, 22], "restrict_sc": 16, "_restrictvalu": 16, "restrict_value_impl": [16, 22], "clamp_scal": 16, "_clampvalu": 16, "clamp_min_st": [16, 22], "restrict_inplace_preprocess": [16, 22], "restrict_preprocess": [16, 22], "msb_clamp_bit_width_impl": [16, 22], "explod": 16, "straightforward": 16, "numpi": [16, 18, 20, 22], "come": [16, 18, 19, 22], "program": 16, "problem": [16, 18, 21], "decis": 16, "spot": 16, "line": [16, 22], "code": 16, "yet": [16, 20], "after": [16, 19, 21, 22], "loop": 16, "react": 16, "world": [16, 22], "hard": 16, "impact": [16, 20], "recompil": 16, "lossi": 16, "arbitrarli": 16, "injector": 16, "excel": 16, "interwin": 16, "turn": [16, 22], "assembli": [16, 22], "goe": [16, 18, 22], "context": [16, 21, 22], "gather": 16, "demand": [16, 22], "simpl": 16, "mention": [16, 18, 22], "heavi": 16, "binar": 16, "inspect": [16, 20, 21, 22], "ipython": [16, 18, 19, 20, 21, 22], "displai": [16, 18, 20, 21, 22], "markdown": [16, 20, 21, 22], "pretty_print_sourc": [16, 20, 21, 22], "n": [16, 20, 21, 22], "getsourc": [16, 20, 21, 22], "func": [16, 22], "script_method": 16, "apart": 16, "pick": [16, 19, 21, 22], "equal": [16, 18, 21, 22], "random": [16, 19, 20, 21], "manual_tensor_qu": 16, "gt": [16, 18, 19, 20, 21, 22], "noth": 16, "surpris": 16, "limit": [16, 22], "mybinaryquant": 16, "inj_tensor_qu": 16, "retriev": [16, 22], "mychildbinaryquant": 16, "child_inj_tensor_qu": 16, "Or": [16, 21], "mybinaryimpl": 16, "myscalingimpl": 16, "mycomposedbinaryquant": 16, "comp_inj_tensor_qu": 16, "live": 16, "mostli": 16, "land": 16, "afford": 16, "conveni": [16, 22], "proxy_class": [16, 19, 22], "weightquantproxyfrominjector": [16, 22], "mybinaryweightquant": 16, "11": [16, 18, 19, 20, 21, 22], "binary_weight_quant_conv": 16, "signed_t": [16, 19, 20, 21, 22], "training_t": [16, 19, 20, 21, 22], "properli": 16, "form": [16, 19, 22], "12": [16, 19, 20, 21, 22], "assert": [16, 18, 19, 20, 21], "is_valid": [16, 19, 21], "those": [16, 19, 21, 22], "explicitli": [16, 19, 21, 22], "13": [16, 18, 19, 20, 22], "mysignedbinaryweightquant": 16, "signed_quant_weight": 16, "And": [16, 22], "add": [16, 18, 19, 20, 21], "There": [16, 18, 21], "simpler": 16, "did": [16, 18, 22], "could": [16, 21, 22], "small_scale_quant_conv": 16, "weight_sign": 16, "arbitrari": [16, 21, 22], "17": [16, 19, 20, 22], "actquantproxyfrominjector": [16, 19, 22], "mysignedbinaryactquant": 16, "binary_relu": 16, "act_quant": [16, 19, 22], "isn": 16, "keywork": 16, "18": [16, 19, 20, 22], "small_scale_binary_ident": 16, "001": [16, 18], "0010": [16, 21], "far": [16, 19, 20, 22], "show": [16, 18, 22], "sort": [16, 19, 21, 22], "shine": 16, "19": [16, 19, 20, 22], "paramfrommaxweightquant": 16, "decor": 16, "spirit": 16, "properti": [16, 19, 22], "param_from_max_quant_conv": 16, "1876": [16, 21], "inde": [16, 21], "verifi": [16, 21], "quant_weight_scal": [16, 21], "21": [16, 20, 21, 22], "simuat": 16, "separ": [16, 21, 22], "22": [16, 20, 21, 22], "float_conv": [16, 21], "1897": 16, "maxbackward1": 16, "23": [16, 18, 22], "traceback": [16, 19, 21, 22], "recent": [15, 16, 19, 21, 22], "5b3646241211": 16, "programdata": 16, "miniconda3": [16, 22], "lib": [16, 21, 22], "site": [16, 21, 22], "py": [16, 18, 19, 20, 21, 22], "1405": [16, 22], "len": [16, 22], "error_msg": [16, 22], "1406": [16, 22], "39": [16, 18, 19, 20, 22], "1407": [16, 22], "__class__": [16, 22], "__name__": [16, 22], "34": [16, 18, 19, 21, 22], "join": [16, 22], "1408": [16, 22], "_incompatiblekei": [16, 22], "missing_kei": [16, 22], "unexpected_kei": [16, 22], "1409": [16, 22], "ouch": 16, "forgiv": 16, "strategi": [16, 22], "26": [16, 20, 22], "quant_conv1": 16, "quant_conv2": 16, "27": [16, 20, 22], "assertionerror": 16, "tmp": [16, 21], "ipykernel_58415": 16, "1066539094": 16, "access": 16, "parent": [16, 19], "28": [16, 22], "sharedparamfrommeanweightquant": 16, "isinst": [16, 19], "cat": [16, 20, 22], "els": [16, 19, 20, 21, 22], "old_quant_conv1_scal": 16, "new_quant_conv1_scal": 16, "eager": [16, 22], "don": [16, 19, 20, 22], "semant": [16, 20], "correctli": [16, 19, 21], "easier": 16, "someth": [16, 19, 22], "mind": [16, 19], "linear": [16, 18, 19, 22], "identityqu": 16, "count": 16, "though": [16, 18, 19, 22, 24], "belong": 16, "quant_conv_w_init": 16, "init": [16, 21], "uniform_": 16, "anymor": 16, "33": [16, 22], "init_tensor_qu": [16, 19, 22], "plan": 16, "distinguish": 16, "illustr": [16, 20, 24], "abl": [16, 22], "leav": 16, "commonquant": 16, "is_clamp": 16, "per_channel_broadcastable_shap": 16, "advancedweightquant": 16, "num_ch": 16, "advancedactquant": 16, "unpack": [16, 22], "chain": 16, "35": [16, 22], "per_channel_quant_conv": 16, "weight_is_clamp": 16, "weight_scaling_per_output_channel": [16, 18, 20, 22], "1842": 16, "1838": 16, "vector": 16, "observ": [16, 18, 19, 20, 21], "36": [16, 22], "1875": [16, 22], "similarli": [16, 18, 20, 22], "37": [16, 22], "quant_ident": [16, 19, 20, 21, 22], "dependencyerror": [16, 19], "b3479e90d1a9": 16, "brevitas_fx": [16, 19], "src": [16, 18, 19, 20, 21, 22], "quant_activ": [16, 19], "134": [16, 19], "135": 16, "136": 16, "137": 16, "quant_lay": [16, 19, 21, 22], "act_impl": [16, 19], "passthrough_act": [16, 19], "77": [16, 19, 20], "78": [16, 19], "79": [16, 19], "80": [3, 16, 19], "81": [16, 19], "mixin": [16, 19, 20], "157": [16, 19], "proxy_prefix": [16, 19], "act_": [16, 19], "158": [16, 19], "kwargs_prefix": [16, 19], "159": [16, 19], "160": [16, 19, 21], "161": [16, 19, 21], "proxy_protocol": [16, 19], "none_quant_injector": [16, 19], "98": [16, 19, 22], "quant_injector": [16, 19], "99": [16, 19, 22], "filter_kwarg": [16, 19], "100": [16, 18, 19, 20, 22], "101": [16, 19], "102": [16, 19], "runtime_qu": [16, 19], "108": [16, 19, 21], "109": [16, 19, 21], "110": [16, 19, 21], "111": [16, 19, 21], "is_passthrough_act": [16, 19], "_is_passthrough_act": [16, 19], "112": [16, 19, 20, 21], "quant_proxi": [16, 19], "export_mod": [16, 19, 21, 22], "export_handl": [16, 18, 19, 21, 22], "74": [16, 19, 22], "modulelist": [16, 19, 20], "75": [16, 19], "tracked_module_list": [16, 19], "76": [16, 19], "add_tracked_modul": [16, 19], "130": [16, 19], "append": [16, 19], "131": [16, 19], "update_tracked_modul": [16, 19], "132": [16, 19], "133": [16, 19], "121": [16, 19, 21], "122": [16, 19, 21, 22], "123": [16, 19, 21], "124": [16, 19, 21], "is_act_en": [16, 19], "_is_act_en": [16, 19], "hidden": [16, 19, 20], "frame": [16, 19], "_depend": 16, "__call__": 16, "__self__": 16, "49": 16, "50": 16, "51": 16, "getattr": 16, "symbol": [16, 22], "52": 16, "53": 16, "messag": 16, "resolv": [16, 19], "complet": [16, 22], "instal": [18, 20, 22], "pypi": [18, 22], "notebook": [18, 22], "onnxoptim": [18, 22], "netron": [18, 20, 22], "visual": [18, 20, 22], "qlinearconv": [18, 22], "qlinearmatmul": [18, 22], "cover": [18, 19, 22], "qonnx": [18, 20, 22], "q": 18, "dq": 18, "three": 18, "fp": [18, 22], "u": 18, "int8": [18, 21], "ntenger": 18, "min": 18, "equival": [18, 19, 22], "sever": 18, "implic": [18, 20], "associ": [18, 20, 22], "int32": [18, 22], "deduantizelinear": 18, "desir": 18, "opset": [18, 20, 21, 22], "consider": 18, "quantconv1d": [18, 21, 22], "quantconvtranspose1d": [18, 21, 22], "quantconvtranspose2d": [18, 21, 22], "offer": 18, "ifram": [18, 20, 22], "show_netron": [18, 20, 22], "model_path": [18, 20, 22], "port": [18, 20, 22], "sleep": [18, 20, 22], "address": [18, 20, 22], "localhost": [18, 20, 22], "brows": [18, 20, 22], "height": [18, 20, 22], "400": [18, 20, 22], "in_ch": 18, "out_ch": 18, "batch_siz": [18, 20], "quant_linear_qcdq": 18, "exported_model": [18, 20], "opset_vers": [18, 20], "8082": [18, 20, 22], "moreov": 18, "interv": [18, 21], "perfectli": [18, 19], "symmetr": [18, 22], "absorb": 18, "fusion": 18, "entir": [18, 22], "quantmodel": 18, "quant_model_qcdq": 18, "8083": [18, 20, 22], "furthermor": 18, "uint8": 18, "expand": 18, "imposs": 18, "less": 18, "etc": 18, "eval": [18, 19, 22], "quant_model_3b_4b_qcdq": 18, "8084": [18, 22], "anoth": [18, 21, 22], "preceed": 18, "opposit": 18, "contraint": 18, "export_onnx_qop": [18, 22], "img_siz": 18, "kernel_s": [18, 19, 21, 22], "quant_model_qop": 18, "manag": 18, "userwarn": [18, 20, 22], "default_opset": 18, "ka": 18, "ir_vers": 18, "producer_nam": 18, "producer_vers": 18, "constant_output_0": 18, "op_typ": 18, "data_typ": 18, "raw_data": 18, "000": 18, "constant_1_output_0": 18, "constant_1": 18, "quantizelinear_output_0": 18, "374": 18, "372": 18, "376": 18, "005": 18, "375": 18, "004": 18, "373": 18, "007": 18, "377": 18, "371": 18, "003": 18, "006": 18, "002": 18, "263": 18, "341": 18, "constant_2_output_0": 18, "constant_2": 18, "271": 18, "032": 18, "0009": [18, 21], "302": 18, "031": 18, "024": 18, "000d": 18, "327": 18, "363": 18, "377u": 18, "000t": 18, "321": 18, "236": 18, "241": 18, "237": 18, "010": 18, "350": 18, "267": 18, "355": [18, 22], "000n": 18, "346": 18, "317": 18, "207": 18, "000y": 18, "326": 18, "334": [18, 21], "362": 18, "304": 18, "340": 18, "275": 18, "324": 18, "332": [18, 21], "026": 18, "333": [18, 21], "335": 18, "226": 18, "0002": [18, 21], "000f": 18, "310": 18, "344": 18, "177": 18, "033": 18, "000i": 18, "315": 18, "035": 18, "354": [18, 20], "377z": 18, "377i": 18, "030": 18, "000w": 18, "303": 18, "022": 18, "336": 18, "377p": 18, "351": 18, "000x": 18, "247": 18, "000h": 18, "225": 18, "3776": 18, "301": 18, "210": 18, "307": 18, "320": 18, "377f": 18, "352": 18, "3770": 18, "313": 18, "261": 18, "037": 18, "220": 18, "202": 18, "013": 18, "266": 18, "347": 18, "377o": 18, "244": 18, "214": 18, "014": 18, "qlinearconv_output_0": 18, "dilat": [18, 21, 22], "group": [18, 21, 22], "kernel_shap": 18, "pad": [18, 21, 22], "stride": [18, 21, 22], "torch_jit": 18, "tensor_typ": 18, "elem_typ": 18, "dim_valu": 18, "126": [18, 21], "opset_import": 18, "domain": [18, 22], "8085": [18, 20, 22], "shown": [18, 22], "prevent": 18, "captur": [18, 20, 22], "examin": 18, "ort": [18, 20], "sess_opt": 18, "sessionopt": 18, "sess": [18, 20], "inferencesess": [18, 20], "input_nam": [18, 20], "get_input": [18, 20], "pred_onx": 18, "out_brevita": 18, "out_ort": 18, "allclos": [18, 20], "kernel": [18, 22], "seem": 18, "must": 18, "behavior": 18, "quantconvnd": 18, "qgemm_ort": 18, "unfortun": 18, "log": [18, 22], "unoptim": 18, "conv": 18, "output_bit_width": [18, 20, 21, 22], "quant_model_qops_4b_4b": 18, "atol": [18, 20], "vecaus": 18, "At": 18, "due": 18, "slightli": 18, "closer": 18, "toler": 18, "deeper": 19, "were": [19, 22], "obtain": [19, 21], "manual_se": [19, 20, 21, 22], "output_quant_conv": [19, 21], "in_channel": [19, 21], "out_channel": [19, 21], "default_quant_conv": [19, 21], "output_identity_qu": 19, "out_tensor1": 19, "out_tensor2": 19, "isclos": [19, 21], "input_output_quant_conv": 19, "input_identity_qu": 19, "becom": [19, 21], "clearer": 19, "meantim": 19, "disabled_quant_ident": 19, "return_quant_ident": 19, "out_tensor": [19, 21], "4566": [19, 20], "5707": 19, "5517": [19, 20], "5897": 19, "5409": 19, "5136": 19, "1902": 19, "0761": 19, "4946": 19, "5029": 19, "4376": 19, "3317": 19, "6361": 19, "0736": [19, 20], "7122": 19, "3780": 19, "1224": 19, "3234": 19, "0844": 19, "0951": 19, "7610": 19, "5980": 19, "0190": 19, "7419": 19, "6278": 19, "6468": 19, "2473": 19, "5327": 19, "1605": 19, "7990": 19, "2936": 19, "3127": 19, "2283": 19, "4351": [19, 20], "3615": 19, "2175": 19, "9214": 19, "divbackward0": [19, 20, 21, 22], "whather": 19, "strip": 19, "implicitli": [19, 21, 22], "out_torch_tensor": 19, "return_disabled_quant_ident": 19, "identity_out_tensor": 19, "uint8actpertensorfloat": [19, 20, 22], "return_quant_relu": 19, "5974": 19, "5402": 19, "5041": 19, "1867": 19, "4481": 19, "3255": 19, "0817": 19, "7083": 19, "3804": [19, 20], "0187": 19, "6254": 19, "6348": 19, "1668": 19, "4387": 19, "2334": 19, "7935": 19, "9230": 19, "0093": [19, 21], "preserv": [19, 21], "return_disabled_quant_relu": 19, "relu_out_tensor": 19, "quantsigmoid": 19, "return_disabled_quant_sigmoid": 19, "sigmoid_out_tensor": 19, "3878": 19, "3611": 19, "3655": [19, 20, 21], "6433": [19, 20], "8236": 19, "6257": 19, "3567": [19, 20], "5474": 19, "4810": 19, "3788": 19, "1820": 19, "4526": 19, "6077": 19, "7911": 19, "1630": 19, "8883": 19, "8471": 19, "9151": 19, "2456": 19, "4198": 19, "2527": [19, 20], "4762": 19, "3184": 19, "1683": 19, "5048": 19, "3226": 19, "6520": 19, "6563": 19, "4385": [19, 21], "3699": 19, "7614": 19, "3102": 19, "2152": [19, 20, 21], "2120": 19, "4432": 19, "0805": 19, "5568": 19, "6898": 19, "4106": 19, "2284": 19, "3480": 19, "8723": 19, "sigmoidbackward0": 19, "shifteduint8actpertensorfloat": [19, 22], "neg": 19, "effect": [19, 20, 22], "shifted_quant_ident": 19, "5854": 19, "5485": 19, "5099": 19, "1888": 19, "4532": 19, "3219": 19, "0772": 19, "6996": 19, "3794": 19, "0189": [19, 20], "6232": 19, "6421": 19, "1708": 19, "4343": [19, 22], "2266": 19, "7931": 19, "9262": 19, "relubackward0": 19, "129": 19, "swherebackward0": 19, "common": [19, 21], "3134": 19, "2557": 19, "0392": 19, "4186": 19, "7361": 19, "5340": 19, "8516": 19, "2887": 19, "3175": 19, "8949": 19, "6743": 19, "0722": 19, "0289": 19, "2021": 19, "4907": 19, "14": [19, 20, 22], "default_quant_relu": 19, "3078": [19, 20], "2555": 19, "0397": 19, "4185": 19, "7454": 19, "5427": 19, "8566": 19, "2943": 19, "3269": 19, "8893": 19, "6674": 19, "0785": 19, "0065": [19, 20], "0262": 19, "1962": 19, "4839": 19, "close": 19, "half": 19, "numer": 19, "lost": [19, 22], "treat": 19, "wast": 19, "regard": 19, "premad": 19, "word": 19, "caution": 19, "anticip": 19, "theme": 19, "interact": [19, 22], "30": [19, 22], "basic": [19, 21, 22], "calibr": 19, "These": [19, 22], "exponenti": [19, 22], "inp1": 19, "inp2": 19, "ones": 19, "out1_train": 19, "out2_train": 19, "out1_ev": 19, "out2_ev": 19, "quanthardtanh": 19, "hardtanh": 19, "8145d2f87fcb": 19, "117": [19, 21], "118": [19, 21], "119": [19, 21], "proxi": [19, 21, 22], "int8actpertensorfloatminmaxinit": 19, "scaling_init_impl": 19, "concer": 19, "quant_hard_tanh": 19, "remind": [19, 20], "legal": [19, 21], "encourag": 19, "drop": 20, "further": 20, "upstream": 20, "input_s": 20, "hidden_s": 20, "num_lay": 20, "nonlinear": 20, "tanh": [20, 21], "batch_first": 20, "bidirect": 20, "int8weightpertensorfloat": [20, 21, 22], "io_quant": 20, "gate_acc_qu": 20, "shared_input_hidden_weight": 20, "layer_impl": 20, "_quantrnnlay": 20, "gate": 20, "fuse": [20, 22], "fed": 20, "stack": 20, "potenti": 20, "nest": [20, 22], "rnn_sublay": 20, "sublayer_numb": 20, "right_to_left_direct": 20, "quant_rnn": 20, "quant_rnn_0_left_to_right": 20, "quant_rnn_0_right_to_left": 20, "quant_rnn_1_left_to_right": 20, "quant_rnn_1_right_to_left": 20, "gate_param": 20, "input_weight": 20, "cell": 20, "bitwidth": 20, "deepspeech": 20, "dai": 20, "qualiti": 20, "count_weight": 20, "numel": 20, "named_paramet": 20, "quant_rnn_single_direct": 20, "quant_rnn_bidirect": 20, "quant_rnn_bidirectional_shared_input_hidden": 20, "print": [20, 21, 22], "600": 20, "1200": 20, "4b": [20, 22], "6b": 20, "io": [20, 21], "quant_rnn_4b": 20, "io_bit_width": 20, "quant_rnn_4b_0_left_to_right": 20, "input_hidden_weight": 20, "hidden_hidden_weight": 20, "hidden_weight": 20, "o": 20, "0316": 20, "0317": 20, "0319": 20, "0318": 20, "0314": 20, "0298": 20, "0285": 20, "0306": 20, "0312": [20, 21], "0315": 20, "0293": 20, "0310": 20, "0309": 20, "pack": 20, "length": 20, "unbatch": 20, "sequenc": 20, "hidden_st": 20, "num_direct": 20, "343": 20, "__torch_function__": [20, 21, 22], "plain": 20, "classmethod": 20, "cb": 20, "pytorch_1000000000000": 20, "python_arg_pars": 20, "seq_dim": 20, "4458": [20, 21], "1651": 20, "7045": 20, "5889": 20, "2532": 20, "0330": 20, "1706": 20, "1376": [20, 21], "4348": 20, "5834": 20, "3577": 20, "2807": 20, "1046": [20, 21], "4293": 20, "1486": 20, "1569": 20, "3530": 20, "6995": 20, "0458": 20, "5295": 20, "3007": [20, 21], "7257": 20, "2877": 20, "1308": 20, "6603": 20, "0196": 20, "8237": 20, "4380": 20, "2615": 20, "3138": 20, "0850": 20, "1961": 20, "1929": 20, "5981": 20, "2508": 20, "2251": 20, "5917": 20, "0257": 20, "3023": 20, "2830": 20, "3344": 20, "4309": 20, "0836": 20, "2701": 20, "3666": 20, "1351": 20, "1736": 20, "1286": 20, "6174": 20, "4682": [20, 21], "1804": 20, "2780": 20, "4974": 20, "4389": 20, "0585": 20, "6242": 20, "0098": 20, "2341": 20, "3511": 20, "2926": 20, "4925": 20, "1414": 20, "4633": 20, "0683": 20, "2633": 20, "3024": 20, "1951": 20, "1707": 20, "0852": 20, "0965": 20, "4656": 20, "3180": 20, "3464": 20, "2782": 20, "1931": 20, "6360": 20, "3293": 20, "7211": 20, "4316": 20, "4145": 20, "3066": 20, "5224": [20, 22], "5849": 20, "1420": 20, "5669": 20, "2367": 20, "3027": 20, "3137": 20, "3632": 20, "5999": 20, "2036": 20, "2201": 20, "2862": 20, "3908": 20, "2091": 20, "2697": 20, "0055": [20, 21], "1761": 20, "1242": 20, "4184": 20, "6472": [20, 22], "4707": 20, "5034": [20, 21], "8368": 20, "1504": 20, "0654": 20, "7714": 20, "4903": 20, "6015": 20, "3596": 20, "2484": 20, "2942": 20, "3409": [20, 21], "8168": 20, "7396": 20, "2958": 20, "7782": 20, "1994": 20, "7846": 20, "3087": 20, "1029": 20, "1479": 20, "3216": 20, "2315": 20, "5209": 20, "0878": 20, "0390": 20, "1365": [20, 21], "2243": 20, "2390": 20, "3706": 20, "1609": 20, "5511": 20, "4096": 20, "5121": 20, "5901": 20, "3609": 20, "3755": 20, "0780": 20, "2829": 20, "1987": 20, "0057": 20, "1306": 20, "3861": 20, "2839": 20, "5962": 20, "1647": 20, "0227": 20, "4372": 20, "3748": 20, "1703": 20, "0738": [20, 21], "catbackward0": 20, "0058": 20, "unsqueezebackward0": 20, "1760": 20, "2670": 20, "1214": 20, "3702": 20, "3884": 20, "4127": [20, 21], "0243": 20, "0425": 20, "2246": 20, "0910": 20, "4734": 20, "0971": 20, "3824": 20, "1396": 20, "6858": 20, "0061": [20, 22], "1275": 20, "5037": 20, "2831": 20, "0566": 20, "2661": [20, 21], "0793": 20, "4926": 20, "0510": 20, "6455": 20, "7191": 20, "1812": 20, "6172": 20, "1529": 20, "4077": 20, "7078": 20, "0453": 20, "0963": 20, "4983": 20, "3977": 20, "0947": 20, "1894": 20, "3725": 20, "2589": 20, "3914": 20, "0063": [20, 21], "2652": 20, "5177": 20, "4230": 20, "0821": 20, "0631": 20, "0505": 20, "0253": [20, 22], "1578": 20, "4988": 20, "5556": 20, "4809": 20, "8144": 20, "6925": 20, "4360": 20, "0256": 20, "5130": 20, "2501": 20, "1347": 20, "7631": 20, "5386": 20, "2437": 20, "4296": 20, "1988": 20, "7246": 20, "1154": 20, "0641": 20, "3142": 20, "0706": 20, "0192": 20, "7185": 20, "8211": 20, "5709": 20, "1155": 20, "4683": 20, "3400": 20, "3015": 20, "3528": 20, "3143": 20, "1411": 20, "2309": 20, "5132": 20, "3721": 20, "5196": 20, "5453": 20, "4066": 20, "7768": 20, "6008": 20, "0546": 20, "0182": [20, 22], "1821": 20, "3763": 20, "3520": 20, "0486": 20, "2124": 20, "3641": 20, "4248": 20, "0789": 20, "2321": 20, "1982": 20, "1302": 20, "0283": 20, "4869": 20, "5379": 20, "6964": 20, "0340": 20, "2944": 20, "4643": 20, "3454": 20, "3284": 20, "3341": 20, "5945": 20, "2020": 20, "0379": 20, "8081": 20, "7260": 20, "0694": 20, "5430": 20, "8018": 20, "2273": 20, "3472": 20, "4924": 20, "4735": 20, "5745": 20, "5619": 20, "6313": 20, "1768": 20, "6541": [20, 22], "0385": [20, 21], "5835": 20, "0449": [20, 21], "3270": 20, "7951": 20, "3591": 20, "2757": 20, "7567": 20, "5194": 20, "7438": 20, "7695": 20, "5451": 20, "2630": 20, "4747": 20, "2245": 20, "3336": 20, "4490": 20, "4619": 20, "1796": 20, "3913": 20, "2053": 20, "2823": [20, 22], "6992": 20, "6607": 20, "1989": 20, "6928": 20, "5581": 20, "5966": 20, "0062": 20, "0064": 20, "1984": 20, "2499": [20, 21], "1102": [20, 21], "0955": 20, "4630": 20, "8672": 20, "1911": 20, "4851": 20, "6982": 20, "5806": 20, "4189": 20, "7423": 20, "9260": 20, "0147": 20, "0514": 20, "2167": 20, "5092": 20, "3846": 20, "0650": [20, 21], "6717": 20, "2492": 20, "0867": 20, "3900": 20, "3521": 20, "4767": 20, "1137": 20, "6879": [20, 21], "1733": 20, "0596": 20, "4279": 20, "5471": 20, "2762": 20, "5904": 20, "3737": 20, "1335": 20, "0140": [20, 22], "2810": 20, "5339": 20, "0562": 20, "7236": 20, "1264": 20, "0211": 20, "3021": 20, "1124": 20, "4777": 20, "3793": [20, 22], "2388": [20, 21], "0702": 20, "4847": 20, "3340": 20, "5225": 20, "1499": 20, "3083": 20, "1756": 20, "1713": 20, "3512": 20, "3041": 20, "3126": 20, "5482": 20, "4882": 20, "1028": 20, "4796": 20, "3640": 20, "0471": 20, "4438": 20, "2686": 20, "3095": 20, "2978": 20, "0993": 20, "0584": 20, "4846": 20, "0526": 20, "4496": 20, "1109": 20, "7416": 20, "3445": 20, "4963": 20, "2803": 20, "1927": 20, "6131": 20, "0661": 20, "3013": 20, "2646": 20, "6321": [20, 21], "4557": 20, "0294": 20, "9407": 20, "7350": 20, "6027": 20, "4116": 20, "6835": 20, "1787": 20, "0271": 20, "1354": 20, "3033": 20, "6229": 20, "3250": 20, "0812": [20, 21], "5633": 20, "0325": 20, "2383": 20, "5850": 20, "6771": 20, "3196": 20, "3934": 20, "3231": 20, "0492": 20, "5128": 20, "8149": 20, "7517": 20, "8711": 20, "4004": 20, "8992": 20, "2178": 20, "8851": 20, "5760": 20, "1054": 20, "5198": 20, "2612": 20, "2570": 20, "1542": 20, "1071": 20, "3854": 20, "0685": 20, "0728": 20, "4240": 20, "1627": 20, "3383": 20, "0428": 20, "1199": 20, "3683": 20, "3298": 20, "4204": 20, "2452": 20, "0934": [20, 21], "2336": 20, "1285": [20, 22], "2044": 20, "0701": 20, "3971": 20, "0175": 20, "1810": 20, "5547": 20, "0467": [20, 21], "0060": [20, 21, 22], "simul": [20, 22], "untrain": 20, "float_rnn": 20, "recov": 20, "123456": 20, "prebuilt": 20, "signedbinaryweightpertensorconst": [20, 22], "binary_rnn": 20, "3684": 20, "0946": 20, "4480": 20, "0050": [20, 21], "1543": 20, "6322": 20, "1643": 20, "1693": 20, "2937": 20, "5227": 20, "2290": 20, "3534": 20, "3883": 20, "4331": 20, "3634": 20, "1941": 20, "2240": 20, "0199": 20, "3485": 20, "1145": 20, "4082": 20, "2987": 20, "0647": 20, "0498": 20, "1493": 20, "0299": 20, "1195": 20, "0776": 20, "5670": 20, "4178": 20, "0239": 20, "4476": 20, "2029": 20, "7042": 20, "6326": 20, "4058": 20, "4118": 20, "0477": 20, "2387": 20, "0179": 20, "4416": 20, "4237": 20, "3282": 20, "1074": 20, "2626": 20, "3581": [20, 21], "2328": 20, "2268": 20, "3103": 20, "4536": 20, "3461": 20, "3163": 20, "7639": 20, "5252": 20, "1790": 20, "2984": 20, "5411": 20, "3147": 20, "6184": 20, "3037": 20, "1877": 20, "1767": 20, "1491": 20, "1049": [20, 22], "2871": 20, "0552": 20, "0883": 20, "0331": 20, "4749": 20, "7013": 20, "2264": 20, "0773": 20, "4583": 20, "0166": 20, "5743": 20, "1160": 20, "0442": 20, "1325": 20, "1657": 20, "0718": 20, "1215": 20, "6240": 20, "3092": 20, "0627": 20, "1882": 20, "4642": 20, "1443": 20, "4705": 20, "2447": 20, "1129": 20, "3011": 20, "2572": 20, "2384": 20, "0376": 20, "1380": 20, "0251": 20, "6399": 20, "5771": 20, "2133": 20, "7967": 20, "1631": 20, "4078": 20, "3199": 20, "0753": 20, "6524": [20, 22], "0690": 20, "1819": 20, "2258": 20, "3889": 20, "3764": 20, "5458": 20, "5704": 20, "6139": 20, "1209": 20, "5173": 20, "4447": 20, "0048": [20, 21], "3481": 20, "5946": 20, "5221": 20, "1644": 20, "2949": 20, "1789": 20, "2707": 20, "2900": 20, "5124": 20, "4399": 20, "0725": 20, "6091": 20, "0435": 20, "2030": 20, "2659": 20, "1547": 20, "0580": 20, "4254": 20, "5559": 20, "1740": 20, "4592": 20, "2369": 20, "3046": 20, "3626": 20, "2079": 20, "4641": 20, "sigmoid_qu": 20, "tanh_quant": 20, "cell_state_qu": 20, "coupled_input_forget_g": 20, "cat_output_cell_st": 20, "shared_intra_layer_weight_qu": 20, "shared_intra_layer_gate_acc_qu": 20, "shared_cell_state_qu": 20, "_quantlstmlay": 20, "cifg": 20, "ourselv": [20, 21, 22], "concen": 20, "illeg": 20, "sigmoid": 20, "forget": 20, "forget_g": 20, "input_g": 20, "orthogon": 20, "wors": 20, "calibration_mod": [20, 22], "bias_correction_mod": [20, 22], "progress": 20, "proof": 20, "quantizelinear": [20, 22], "dequantizelinear": [20, 22], "qeight": 20, "quant_lstm_weight_onli": 20, "quant_lstm_weight_only_4b": 20, "8080": 20, "np": 20, "np_input": 20, "astyp": 20, "seq_len": 20, "pred_onnx": 20, "quant_lstm_weight_only_cifg": 20, "quant_lstm_weight_only_cifg_4b": 20, "24": [7, 20, 22], "quant_lstm_weight_only_bidirectional_2_lay": 20, "home": [20, 21], "giusepp": 20, "quant_lstm_weight_only_bidirectional_2_layers_shar": 20, "quant_lstm_weight_only_bidirectional_2_layers_shared_ih": 20, "zp": 20, "quant_lstm_weight_only_bidirectional_2_layers_shared_q": 20, "8086": 20, "side": 20, "quant_lstm_weight_only_bidirectional_2_layers_shared_q_ih": 20, "8087": 20, "quantlstmcel": 20, "export_qonnx": [20, 22], "quant_lstm": 20, "8088": 20, "quant_hidden_st": 20, "quant_cell_st": 20, "quant_weight_ii": 20, "quant_weight_if": 20, "quant_weight_": 20, "quant_weight_io": 20, "quant_weight_hi": 20, "quant_weight_hf": 20, "quant_weight_hc": 20, "quant_weight_ho": 20, "quant_bias_input": 20, "quant_bias_forget": 20, "quant_bias_cel": 20, "quant_bias_output": 20, "output_scal": [20, 21, 22], "output_zero_point": 20, "cell_state_scal": 20, "cell_state_zero_point": 20, "cell_state_bit_width": 20, "input_acc_scal": 20, "input_acc_zero_point": 20, "input_acc_bit_width": 20, "forget_acc_scal": 20, "forget_acc_zero_point": 20, "forget_acc_bit_width": 20, "cell_acc_scal": 20, "cell_acc_zero_point": 20, "cell_acc_bit_width": 20, "output_acc_scal": 20, "output_acc_zero_point": 20, "output_acc_bit_width": 20, "input_sigmoid_scal": 20, "input_sigmoid_zero_point": 20, "input_sigmoid_bit_width": 20, "forget_sigmoid_scal": 20, "forget_sigmoid_zero_point": 20, "forget_sigmoid_bit_width": 20, "cell_tanh_scal": 20, "cell_tanh_zero_point": 20, "cell_tanh_bit_width": 20, "output_sigmoid_scal": 20, "output_sigmoid_zero_point": 20, "output_sigmoid_bit_width": 20, "hidden_state_tanh_scal": 20, "hidden_state_tanh_zero_point": 20, "hidden_state_tanh_bit_width": 20, "quantweightbiasinputoutputlay": [21, 22], "notic": [21, 22], "local": [21, 22], "python3": 21, "tqdm": 21, "tqdmwarn": 21, "iprogress": 21, "jupyt": 21, "ipywidget": 21, "readthedoc": 21, "en": 21, "stabl": [21, 22], "user_instal": 21, "html": 21, "autonotebook": 21, "notebook_tqdm": 21, "padding_typ": 21, "weightquanttyp": [21, 22], "biasquanttyp": [21, 22], "actquanttyp": [21, 22], "padding_mod": 21, "Its": 21, "intercept": [21, 22], "dedic": 21, "is_weight_quant_en": 21, "is_bias_quant_en": 21, "is_input_quant_en": 21, "is_output_quant_en": 21, "convolut": [21, 22], "2594": 21, "5392": 21, "5916": 21, "3493": 21, "6813": 21, "3732": 21, "1229": 21, "0084": [21, 22], "0031": 21, "1702": 21, "1069": 21, "8181": 21, "8056": 21, "4738": 21, "0589": 21, "1278": 21, "1718": 21, "1162": 21, "1526": 21, "9903": 21, "3541": 21, "1645": 21, "0557": 21, "2080": 21, "thnnconv2dbackward0": 21, "unquant": [21, 22], "everywher": 21, "seed": 21, "reproduc": 21, "disabled_quant_conv": 21, "freedom": 21, "experi": [21, 22], "littl": 21, "formula": 21, "integer_valu": 21, "quant_conv": 21, "short": 21, "0790": 21, "0503": 21, "1149": 21, "1903": 21, "1329": 21, "1813": 21, "0108": 21, "0593": 21, "0970": 21, "0215": 21, "0144": 21, "2280": 21, "1239": 21, "0090": 21, "1957": 21, "2011": 21, "0018": 21, "1993": 21, "0359": 21, "1778": 21, "1400": 21, "0916": 21, "1059": 21, "2173": 21, "1670": 21, "1939": 21, "2191": 21, "1688": 21, "1383": 21, "1185": 21, "1742": 21, "0808": 21, "1652": 21, "0233": 21, "0485": 21, "1418": 21, "1077": 21, "0036": 21, "1508": 21, "1616": 21, "0287": 21, "int_weight": 21, "quant_weight_zero_point": 21, "quant_weight_manu": 21, "popul": 21, "field": 21, "account": 21, "rel": 21, "expens": 21, "sparingli": 21, "quant_act": 21, "out_tensor_0": 21, "out_tensor_1": 21, "0173": 21, "0307": 21, "evalu": [21, 22], "9489": 21, "9111": 21, "0536": 21, "5788": 21, "3645": 21, "3401": 21, "4325": 21, "6498": 21, "6411": 21, "4390": 21, "9029": 21, "7012": [21, 22], "1591": 21, "9235": 21, "5883": 21, "7258": 21, "5330": 21, "9165": 21, "0820": 21, "4148": 21, "3651": 21, "0164": 21, "9567": 21, "2758": 21, "2414": 21, "2111": 21, "9124": 21, "3814": 21, "8805": 21, "3191": 21, "8965": 21, "2048": 21, "8113": 21, "1142": 21, "3381": 21, "2238": 21, "0068": 21, "2567": 21, "0731": 21, "4280": 21, "0909": 21, "0875": 21, "6851": 21, "7744": 21, "8143": 21, "3557": 21, "2802": 21, "addbackward0": 21, "0240": 21, "consequ": 21, "longer": 21, "handl": 21, "5800": 21, "0157": 21, "4445": 21, "8577": 21, "5643": 21, "0383": 21, "9028": 21, "5191": 21, "6546": 21, "1442": 21, "5868": 21, "maxpool2dwithindicesbackward0": 21, "0226": 21, "decai": [21, 22], "4943": 21, "9938": 21, "9073": 21, "7681": 21, "3262": 21, "9186": 21, "1786": 21, "3659": 21, "7489": 21, "8946": 21, "0451": 21, "5594": 21, "1346": 21, "4770": 21, "6951": 21, "0676": 21, "5111": 21, "8459": 21, "8990": 21, "9426": 21, "7945": 21, "9220": 21, "7772": 21, "7177": 21, "4414": 21, "2220": 21, "5747": 21, "6710": 21, "4594": 21, "3462": 21, "9729": 21, "5896": 21, "5276": 21, "0900": 21, "8852": 21, "tanhbackward0": 21, "input_quant_conv": 21, "9693": 21, "9431": 21, "2459": 21, "5416": 21, "9037": 21, "5278": 21, "6207": 21, "3578": 21, "4815": 21, "4551": 21, "4065": 21, "8889": 21, "3393": 21, "0803": 21, "1748": 21, "0977": 21, "6284": 21, "7193": [21, 22], "7626": 21, "2634": 21, "3453": 21, "3349": 21, "1923": 21, "5993": 21, "9579": 21, "2208e": 21, "05": [21, 22], "mimpli": 21, "reflect": [21, 22], "worst": 21, "largest": 21, "int_valu": 21, "randint": 21, "quant_tensor_input": 21, "7000e": 21, "03": 21, "5000e": 21, "2400e": 21, "2000e": 21, "3000e": 21, "0000e": 21, "2700e": 21, "9000e": 21, "6000e": 21, "4000e": 21, "1100e": 21, "1500e": 21, "8000e": 21, "0600e": 21, "1000e": 21, "1300e": 21, "0100e": 21, "1900e": 21, "113": 21, "return_quant_conv": 21, "114": [21, 22], "0085": 21, "0066": 21, "0038": 21, "0115": 21, "0037": 21, "0015": [21, 22], "0027": 21, "0079": 21, "0034": 21, "0043": 21, "0008": 21, "0052": 21, "0033": 21, "0082": 21, "0021": [21, 22], "0004": 21, "0054": 21, "0013": [21, 22], "8448e": 21, "07": 21, "115": 21, "116": 21, "0035": 21, "0051": 21, "0047": 21, "0017": [21, 22], "0028": 21, "0044": [21, 22], "0024": 21, "0011": 21, "0039": 21, "7410e": 21, "4060": 21, "3654": 21, "7876": 21, "8119": 21, "9825": 21, "5115": 21, "3979": 21, "3248": 21, "3816": 21, "0568": 21, "8038": 21, "3491": 21, "4141": 21, "5846": 21, "4222": 21, "7389": 21, "2517": 21, "1624": 21, "7308": 21, "0081": 21, "got": 21, "saw": [21, 22], "assum": [7, 21, 22], "somehow": [21, 22], "predefin": [21, 22], "tri": [21, 22], "bias_quant_conv": 21, "ipykernel_48365": 21, "2280634207": 21, "opt": 21, "conda": 21, "torch_1": 21, "_call_impl": [21, 22], "1100": 21, "_backward_hook": [21, 22], "_forward_hook": [21, 22], "_forward_pre_hook": [21, 22], "_global_backward_hook": [21, 22], "1101": 21, "_global_forward_hook": [21, 22], "_global_forward_pre_hook": [21, 22], "forward_cal": [21, 22], "1103": 21, "1104": 21, "full_backward_hook": [21, 22], "non_full_backward_hook": [21, 22], "workspac": 21, "fork_brevita": 21, "190": 21, "191": 21, "192": 21, "forward_impl": [21, 22], "193": 21, "194": [21, 22], "inner_forward_impl": [21, 22], "330": 21, "331": 21, "cache_inference_quant_bia": [21, 22], "_cached_bia": [21, 22], "_cachedio": [21, 22], "metadata_onli": [21, 22], "parameter_qu": [21, 22], "impl": [21, 22], "requires_input_scal": [21, 22], "162": 21, "163": 21, "requires_input_bit_width": [21, 22], "164": 21, "0005": 21, "0106": 21, "0012": 21, "0007": 21, "0067": 21, "0059": 21, "0071": 21, "8108e": 21, "input_bias_quant_conv": 21, "3825": 21, "1371": 21, "9135": 21, "2016": 21, "7495": 21, "4071": 21, "0755": 21, "5283": 21, "0788": 21, "3802": 21, "2234": 21, "8678": 21, "5546": 21, "4408": 21, "6788": 21, "4422": 21, "4412": 21, "3205": 21, "0083": 21, "3295": 21, "2076": 21, "4417": 21, "8610e": 21, "0080": 21, "0014": 21, "0029": 21, "0003": 21, "0019": 21, "0056": 21, "0091": 21, "0095": 21, "8384e": 21, "give": 21, "output_bias_quant_conv": 21, "2990591641": 21, "int8biaspertensorfloatinternalsc": 21, "bias_internal_scale_quant_conv": 21, "8346": 21, "0746": 21, "5212": 21, "1019": 21, "6004": 21, "1500": 21, "1453": 21, "1551": 21, "3458": 21, "1312": 21, "2502": 21, "5267": 21, "2412": 21, "3556": 21, "3289": 21, "2276": 21, "4599": 21, "6094": 21, "5064": 21, "6768": 21, "6638": 21, "2359": 21, "lead": 21, "125": 21, "unquant_bias_input_quant_conv": 21, "6632": 21, "2411": 21, "2064": 21, "7371": 21, "3910": 21, "9533": 21, "2994": 21, "4684": 21, "4495": 21, "5021": 21, "5738": 21, "4199": 21, "3380": 21, "6218": 21, "0408": 21, "8483": 21, "5625": [21, 22], "1837": 21, "0575": 21, "2816": [21, 22], "4993": 21, "4556": 21, "4269": 21, "5369": 21, "0975e": 21, "1276": 21, "0774": 21, "3152": 21, "4585": 21, "7320": 21, "2324": 21, "subject": [21, 22], "retur": 21, "bias_input_quant_conv": 21, "8357": 21, "0733": 21, "9527": 21, "1803": 21, "2154": 21, "7598": 21, "1121": 21, "8728": 21, "7917": 21, "6516": 21, "1852": 21, "7263": 21, "0956": 21, "2747": 21, "1617": 21, "8299": 21, "9934": 21, "3821": 21, "4865": 21, "9309": 21, "7924": 21, "4201": 21, "2343": 21, "1532": 21, "github": 22, "tree": 22, "master": 22, "in_featur": 22, "out_featur": 22, "quant_linear": 22, "0053": 22, "5820": 22, "5204": 22, "2723": 22, "1896": 22, "5607": 22, "0046": 22, "3803": 22, "2704": 22, "1879": 22, "0137": 22, "5591": 22, "absolut": 22, "full": 22, "corrispond": 22, "83": 22, "59": 22, "41": 22, "float_input": 22, "float_output": 22, "5410": 22, "2934": 22, "1788": 22, "5684": 22, "0845": 22, "3986": 22, "9036": 22, "4586": 22, "3096": 22, "2058": 22, "6525": 22, "3723": 22, "8677": 22, "3873": 22, "2801": 22, "9009": 22, "9507": 22, "mmbackward": 22, "made": 22, "int8weightpertensorfixedpoint": 22, "0078": 22, "3828": 22, "5781": 22, "5234": 22, "2734": 22, "0156": 22, "handi": 22, "overal": 22, "quant_linear1": 22, "quant_linear1_scale_before_shar": 22, "quant_linear2": 22, "4f": 22, "9109": 22, "4609": 22, "3135": 22, "6523": 22, "2089": 22, "3752": 22, "8697": 22, "3893": 22, "9011": 22, "9521": 22, "0542e": 22, "5490": 22, "2894": 22, "5617": 22, "0894": 22, "3958": 22, "0170": 22, "quant_relu": 22, "5681": 22, "signed_quant_output": 22, "unsigned_quant_output": 22, "5588": 22, "999": 22, "percentil": 22, "300": 22, "proper": 22, "int16bia": 22, "alessa": 22, "appdata": 22, "temp": 22, "ipykernel_18920": 22, "2660651517": 22, "pt190": 22, "1050": 22, "1051": 22, "1052": 22, "1053": 22, "documenti": 22, "brevitas_tvmcon": 22, "96": 22, "97": 22, "356": 22, "357": 22, "358": 22, "359": 22, "195": 22, "196": 22, "197": 22, "198": 22, "1263": 22, "1680": 22, "1231": 22, "4658": 22, "2395": 22, "5207": 22, "3989": 22, "6461": 22, "8687": 22, "0466": 22, "4813": 22, "addmmbackward": 22, "constraint": 22, "operand": 22, "allign": 22, "easi": 22, "float_inp1": 22, "float_inp2": 22, "train_quant_inp1": 22, "train_quant_inp2": 22, "train_mode_add": 22, "ema": 22, "eval_quant_inp1": 22, "eval_quant_inp2": 22, "eval_mode_add": 22, "5335": 22, "2875": 22, "0447": 22, "5751": 22, "0863": 22, "4057": 22, "0160": 22, "3994": 22, "8307": 22, "7188": 22, "5910": 22, "1757": 22, "9329": 22, "5431": 22, "7636": 22, "6773": 22, "2300": 22, "input_dequant_valu": 22, "input_integer_valu": 22, "output_integer_valu": 22, "max_pool1d": 22, "float_inp": 22, "1218": 22, "1580": 22, "2533": 22, "8504": 22, "6876": 22, "3076": 22, "1170": 22, "4704": 22, "1628": 22, "4475": 22, "2714": 22, "8685": 22, "1448": 22, "1086": 22, "9228": 22, "2666": 22, "0543": 22, "6152": 22, "4162": 22, "8323": 22, "3160": 22, "0181": 22, "squeezebackward1": 22, "652": 22, "experiment": [1, 22, 24], "c10": 22, "tensorimpl": 22, "h": 22, "1156": 22, "ceil_mod": 22, "8204": 22, "2480": 22, "4089": 22, "6913": 22, "5964": 22, "2983": 22, "9714": 22, "4386": 22, "1614": 22, "8952": 22, "2649": 22, "7006": 22, "1438": 22, "1081": 22, "7272": 22, "8529": 22, "9646": 22, "0542": 22, "5478": 22, "3937": 22, "6817": 22, "9807": 22, "tanhbackward": 22, "simiarli": 22, "train_mode_cat": 22, "eval_mode_cat": 22, "concat": 22, "easiest": 22, "3880": 22, "5044": 22, "2716": 22, "1940": 22, "5432": 22, "0388": 22, "1816": 22, "0374": 22, "6341": 22, "5447": 22, "uint8actpertensorfloatmaxinit": 22, "5294": 22, "5647": 22, "0235": 22, "static": 22, "depthwis": 22, "per_channel_depthwise_quant_conv": 22, "input_scaling_per_output_channel": 22, "input_scaling_stats_permute_dim": 22, "permut": 22, "input_per_channel_broadcastable_shap": 22, "4033": 22, "8380": 22, "8616": 22, "4503": 22, "4937": 22, "1901": 22, "compact": 22, "usabl": 22, "perchannel3bactqu": 22, "driven": 22, "solver": 22, "actquantsolv": 22, "float_to_int_impl_typ": 22, "floattointimpltyp": 22, "nearest": 22, "scaling_stats_op": 22, "statsop": 22, "restrict_scaling_typ": 22, "restrictvaluetyp": 22, "overriden": 22, "tweak": 22, "8b": 22, "learnedintweightperchannelfloat": 22, "log_fp": 22, "1887": 22, "0132": 22, "0030": 22, "backpropag": 22, "loss": 22, "regular": 22, "push": 22, "larger": 22, "learnedintactpertensorfloat": 22, "4588": 22, "3119": 22, "6530": 22, "6493": 22, "3731": 22, "8706": 22, "8979": 22, "9543": 22, "9068e": 22, "6866e": 22, "4251e": 22, "didn": 22, "29": 22, "float_linear": 22, "1653109852": 22, "fused_activation_quant_proxi": 22, "bit_width_offset": 22, "supress": 22, "accomod": 22, "basi": 22, "0023": 22, "question": 22, "underneath": 22, "immedi": 22, "_zero_hw_sentinel": 22, "fusedactivationquantproxi": 22, "activation_impl": 22, "restrict_clamp_sc": 22, "_restrictclampvalu": 22, "deped": 22, "procedur": 22, "restrict_v": 22, "modular": 22, "degre": 22, "overhead": 22, "scatter": 22, "int8actpertensorfloatfromscratch": 22, "dequant_valu": 22, "convini": 22, "engin": 22, "\u00e8xtendedinjector": 22, "repeat": 22, "recurs": 22, "manner": 22, "indipend": 22, "fine": 22, "grain": 22, "int8actpertensorfloatparameterfromscratch": 22, "quant_identity_bit_width": 22, "quant_linear_bit_width": 22, "perfect": 22, "indipedent": 22, "satisfi": 22, "protobuf": 22, "six": 22, "38": 22, "32b": 22, "16b": 22, "quant_conv_4b8b": 22, "output_path": 22, "qop_onnx_conv_4b8b": 22, "input_t": 22, "relax": 22, "dialect": 22, "brevitas_onnx_conv4b8b": 22, "40": 22, "quant_conv_4b_weight": 22, "brevitas_onnx_conv_4b_weight": 22, "tvm": [22, 24], "7b": 22, "export_torch_qop": 22, "quant_conv_8b7b": 22, "pytorch_qf_conv_8b7b": 22, "pt": 22, "42": 22, "tracerwarn": 22, "trace": 22, "safe": 22, "caus": 22, "incorrect": 22, "incorpor": 22, "programmat": 22, "pattern": 22, "subsystem": 22, "transform": 22, "beyond": 22, "worth": 22, "embrac": 22, "backport": 22, "tracer": 22, "value_trac": 22, "condit": 22, "caveat": 22, "temporar": 22, "quant_model": 22, "calibrate_model": 22, "calibration_load": 22, "no_grad": 22, "imag": 22, "_": 22, "enumer": 22, "iter": 22, "pyxir": 24, "compliant": 24, "necessar": 24, "overoutputfeaturesview": [1, 3], "prezerocenterzeropoint": 1, "pre_zero_point_stats_input_view_shape_impl": 1, "pre_zero_point_shap": 1, "center": 1, "incom": 1, "intend": 1, "get_zero_cent": 1, "over_output_featur": [1, 3, 7], "a2q": 7, "untest": 15}, "objects": {"brevitas": [[1, 0, 0, "-", "core"]], "brevitas.core": [[2, 0, 0, "-", "bit_width"], [3, 0, 0, "-", "function_wrapper"], [4, 0, 0, "-", "quant"], [1, 0, 0, "-", "restrict_val"], [5, 0, 0, "-", "scaling"], [6, 0, 0, "-", "stats"], [1, 0, 0, "-", "utils"], [1, 0, 0, "-", "zero_point"]], "brevitas.core.bit_width": [[2, 0, 0, "-", "const"], [2, 0, 0, "-", "parameter"]], "brevitas.core.bit_width.const": [[2, 1, 1, "", "BitWidthConst"], [2, 1, 1, "", "BitWidthStatefulConst"], [2, 1, 1, "", "MsbClampBitWidth"]], "brevitas.core.bit_width.const.BitWidthConst": [[2, 2, 1, "", "forward"]], "brevitas.core.bit_width.const.BitWidthStatefulConst": [[2, 2, 1, "", "forward"]], "brevitas.core.bit_width.const.MsbClampBitWidth": [[2, 2, 1, "", "forward"]], "brevitas.core.bit_width.parameter": [[2, 1, 1, "", "BitWidthParameter"], [2, 1, 1, "", "RemoveBitwidthParameter"]], "brevitas.core.bit_width.parameter.BitWidthParameter": [[2, 2, 1, "", "forward"]], "brevitas.core.bit_width.parameter.RemoveBitwidthParameter": [[2, 2, 1, "", "forward"]], "brevitas.core.function_wrapper": [[3, 0, 0, "-", "clamp"], [3, 0, 0, "-", "misc"], [3, 0, 0, "-", "ops_ste"], [3, 0, 0, "-", "shape"]], "brevitas.core.function_wrapper.clamp": [[3, 1, 1, "", "ClampMin"], [3, 1, 1, "", "ScalarClamp"], [3, 1, 1, "", "TensorClamp"]], "brevitas.core.function_wrapper.clamp.ClampMin": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.clamp.ScalarClamp": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.clamp.TensorClamp": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.misc": [[3, 1, 1, "", "Identity"], [3, 1, 1, "", "InplaceLogTwo"], [3, 1, 1, "", "LogTwo"], [3, 1, 1, "", "PowerOfTwo"]], "brevitas.core.function_wrapper.misc.Identity": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.misc.InplaceLogTwo": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.misc.LogTwo": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.misc.PowerOfTwo": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.ops_ste": [[3, 1, 1, "", "CeilSte"], [3, 1, 1, "", "DPURoundSte"], [3, 1, 1, "", "FloorSte"], [3, 1, 1, "", "InplaceTensorClampSte"], [3, 1, 1, "", "RoundSte"], [3, 1, 1, "", "RoundToZeroSte"], [3, 1, 1, "", "ScalarClampMinSte"], [3, 1, 1, "", "TensorClampSte"]], "brevitas.core.function_wrapper.ops_ste.CeilSte": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.ops_ste.DPURoundSte": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.ops_ste.FloorSte": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.ops_ste.InplaceTensorClampSte": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.ops_ste.RoundSte": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.ops_ste.RoundToZeroSte": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.ops_ste.ScalarClampMinSte": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.ops_ste.TensorClampSte": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.shape": [[3, 1, 1, "", "OverBatchOverOutputChannelView"], [3, 1, 1, "", "OverBatchOverTensorView"], [3, 1, 1, "", "OverOutputChannelView"], [3, 1, 1, "", "OverOutputFeaturesView"], [3, 1, 1, "", "OverTensorView"], [3, 1, 1, "", "PermuteDims"], [3, 1, 1, "", "StatsInputViewShapeImpl"]], "brevitas.core.function_wrapper.shape.OverBatchOverOutputChannelView": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.shape.OverBatchOverTensorView": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.shape.OverOutputChannelView": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.shape.OverOutputFeaturesView": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.shape.OverTensorView": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.shape.PermuteDims": [[3, 2, 1, "", "forward"]], "brevitas.core.function_wrapper.shape.StatsInputViewShapeImpl": [[3, 3, 1, "", "OVER_BATCH_OVER_OUTPUT_CHANNELS"], [3, 3, 1, "", "OVER_BATCH_OVER_TENSOR"], [3, 3, 1, "", "OVER_OUTPUT_CHANNELS"], [3, 3, 1, "", "OVER_OUTPUT_FEATURES"], [3, 3, 1, "", "OVER_TENSOR"]], "brevitas.core.quant": [[4, 0, 0, "-", "binary"], [4, 0, 0, "-", "delay"], [4, 0, 0, "-", "int"], [4, 0, 0, "-", "int_base"], [4, 0, 0, "-", "ternary"]], "brevitas.core.quant.binary": [[4, 1, 1, "", "BinaryQuant"], [4, 1, 1, "", "ClampedBinaryQuant"]], "brevitas.core.quant.binary.BinaryQuant": [[4, 2, 1, "", "forward"]], "brevitas.core.quant.binary.ClampedBinaryQuant": [[4, 2, 1, "", "forward"]], "brevitas.core.quant.delay": [[4, 1, 1, "", "DelayWrapper"]], "brevitas.core.quant.delay.DelayWrapper": [[4, 2, 1, "", "forward"]], "brevitas.core.quant.int": [[4, 1, 1, "", "DecoupledRescalingIntQuant"], [4, 1, 1, "", "DecoupledRescalingIntQuantWithInput"], [4, 1, 1, "", "PrescaledRestrictIntQuant"], [4, 1, 1, "", "PrescaledRestrictIntQuantWithInputBitWidth"], [4, 1, 1, "", "RescalingIntQuant"], [4, 1, 1, "", "TruncIntQuant"]], "brevitas.core.quant.int.DecoupledRescalingIntQuant": [[4, 2, 1, "", "forward"]], "brevitas.core.quant.int.DecoupledRescalingIntQuantWithInput": [[4, 2, 1, "", "forward"]], "brevitas.core.quant.int.PrescaledRestrictIntQuant": [[4, 2, 1, "", "forward"]], "brevitas.core.quant.int.PrescaledRestrictIntQuantWithInputBitWidth": [[4, 2, 1, "", "forward"]], "brevitas.core.quant.int.RescalingIntQuant": [[4, 2, 1, "", "forward"]], "brevitas.core.quant.int.TruncIntQuant": [[4, 2, 1, "", "forward"]], "brevitas.core.quant.int_base": [[4, 1, 1, "", "DecoupledIntQuant"], [4, 1, 1, "", "IntQuant"]], "brevitas.core.quant.int_base.DecoupledIntQuant": [[4, 2, 1, "", "forward"], [4, 2, 1, "", "max_int"], [4, 2, 1, "", "min_int"], [4, 2, 1, "", "to_int"]], "brevitas.core.quant.int_base.IntQuant": [[4, 2, 1, "", "forward"], [4, 2, 1, "", "max_int"], [4, 2, 1, "", "min_int"], [4, 2, 1, "", "to_int"]], "brevitas.core.quant.ternary": [[4, 1, 1, "", "TernaryQuant"]], "brevitas.core.quant.ternary.TernaryQuant": [[4, 2, 1, "", "forward"]], "brevitas.core.restrict_val": [[1, 1, 1, "", "FloatRestrictValue"], [1, 1, 1, "", "IntRestrictValue"], [1, 1, 1, "", "LogFloatRestrictValue"], [1, 1, 1, "", "PowerOfTwoRestrictValue"]], "brevitas.core.restrict_val.FloatRestrictValue": [[1, 2, 1, "", "forward"], [1, 2, 1, "", "restrict_init_float"], [1, 2, 1, "", "restrict_init_inplace_module"], [1, 2, 1, "", "restrict_init_module"], [1, 2, 1, "", "restrict_init_tensor"]], "brevitas.core.restrict_val.IntRestrictValue": [[1, 2, 1, "", "forward"], [1, 2, 1, "", "restrict_init_float"], [1, 2, 1, "", "restrict_init_inplace_module"], [1, 2, 1, "", "restrict_init_module"], [1, 2, 1, "", "restrict_init_tensor"]], "brevitas.core.restrict_val.LogFloatRestrictValue": [[1, 2, 1, "", "forward"], [1, 2, 1, "", "restrict_init_float"], [1, 2, 1, "", "restrict_init_inplace_module"], [1, 2, 1, "", "restrict_init_module"], [1, 2, 1, "", "restrict_init_tensor"]], "brevitas.core.restrict_val.PowerOfTwoRestrictValue": [[1, 2, 1, "", "forward"], [1, 2, 1, "", "restrict_init_float"], [1, 2, 1, "", "restrict_init_inplace_module"], [1, 2, 1, "", "restrict_init_module"], [1, 2, 1, "", "restrict_init_tensor"]], "brevitas.core.scaling": [[5, 0, 0, "-", "int_scaling"], [5, 0, 0, "-", "runtime"], [5, 0, 0, "-", "standalone"]], "brevitas.core.scaling.int_scaling": [[5, 1, 1, "", "IntScaling"], [5, 1, 1, "", "PowerOfTwoIntScaling"]], "brevitas.core.scaling.int_scaling.IntScaling": [[5, 2, 1, "", "forward"]], "brevitas.core.scaling.int_scaling.PowerOfTwoIntScaling": [[5, 2, 1, "", "forward"]], "brevitas.core.scaling.runtime": [[5, 1, 1, "", "RuntimeStatsScaling"], [5, 1, 1, "", "StatsFromParameterScaling"]], "brevitas.core.scaling.runtime.RuntimeStatsScaling": [[5, 2, 1, "", "forward"]], "brevitas.core.scaling.runtime.StatsFromParameterScaling": [[5, 2, 1, "", "forward"]], "brevitas.core.scaling.standalone": [[5, 1, 1, "", "ConstScaling"], [5, 1, 1, "", "ParameterFromRuntimeStatsScaling"], [5, 1, 1, "", "ParameterFromStatsFromParameterScaling"], [5, 1, 1, "", "ParameterScaling"]], "brevitas.core.scaling.standalone.ConstScaling": [[5, 2, 1, "", "forward"]], "brevitas.core.scaling.standalone.ParameterFromRuntimeStatsScaling": [[5, 2, 1, "", "forward"], [5, 2, 1, "", "state_dict"], [5, 2, 1, "", "training_forward"]], "brevitas.core.scaling.standalone.ParameterFromStatsFromParameterScaling": [[5, 2, 1, "", "forward"], [5, 2, 1, "", "state_dict"]], "brevitas.core.scaling.standalone.ParameterScaling": [[5, 2, 1, "", "forward"]], "brevitas.core.stats": [[6, 0, 0, "-", "stats_op"], [6, 0, 0, "-", "stats_wrapper"], [6, 0, 0, "-", "view_wrapper"]], "brevitas.core.stats.stats_op": [[6, 1, 1, "", "AbsAve"], [6, 1, 1, "", "AbsMax"], [6, 1, 1, "", "AbsMaxAve"], [6, 1, 1, "", "AbsMaxL2"], [6, 1, 1, "", "AbsMinMax"], [6, 1, 1, "", "AbsPercentile"], [6, 1, 1, "", "KLMinimizerThreshold"], [6, 1, 1, "", "L1Norm"], [6, 1, 1, "", "L2Norm"], [6, 1, 1, "", "MSE"], [6, 1, 1, "", "MeanLearnedSigmaStd"], [6, 1, 1, "", "MeanSigmaStd"], [6, 1, 1, "", "NegativeMinOrZero"], [6, 1, 1, "", "NegativePercentileOrZero"], [6, 1, 1, "", "PercentileInterval"]], "brevitas.core.stats.stats_op.AbsAve": [[6, 2, 1, "", "forward"]], "brevitas.core.stats.stats_op.AbsMax": [[6, 2, 1, "", "forward"]], "brevitas.core.stats.stats_op.AbsMaxAve": [[6, 2, 1, "", "forward"]], "brevitas.core.stats.stats_op.AbsMaxL2": [[6, 2, 1, "", "forward"]], "brevitas.core.stats.stats_op.AbsMinMax": [[6, 2, 1, "", "forward"]], "brevitas.core.stats.stats_op.AbsPercentile": [[6, 2, 1, "", "forward"]], "brevitas.core.stats.stats_op.KLMinimizerThreshold": [[6, 2, 1, "", "forward"], [6, 2, 1, "", "smooth_normalize_distribution"]], "brevitas.core.stats.stats_op.L1Norm": [[6, 2, 1, "", "forward"]], "brevitas.core.stats.stats_op.L2Norm": [[6, 2, 1, "", "forward"]], "brevitas.core.stats.stats_op.MSE": [[6, 2, 1, "", "evaluate_loss"], [6, 2, 1, "", "forward"], [6, 2, 1, "", "mse_fib_search"], [6, 2, 1, "", "mse_grid_search"], [6, 2, 1, "", "mse_loss_fn"], [6, 2, 1, "", "mse_search"]], "brevitas.core.stats.stats_op.MeanLearnedSigmaStd": [[6, 2, 1, "", "forward"]], "brevitas.core.stats.stats_op.MeanSigmaStd": [[6, 2, 1, "", "forward"]], "brevitas.core.stats.stats_op.NegativeMinOrZero": [[6, 2, 1, "", "forward"]], "brevitas.core.stats.stats_op.NegativePercentileOrZero": [[6, 2, 1, "", "forward"]], "brevitas.core.stats.stats_op.PercentileInterval": [[6, 2, 1, "", "forward"]], "brevitas.core.utils": [[1, 1, 1, "", "ParameterWrapper"], [1, 1, 1, "", "SingleArgStatelessBuffer"], [1, 1, 1, "", "SliceTensor"], [1, 1, 1, "", "StatelessBuffer"], [1, 4, 1, "", "inplace_momentum_update"], [1, 4, 1, "", "inplace_tensor_add"], [1, 4, 1, "", "inplace_tensor_mul"]], "brevitas.core.utils.ParameterWrapper": [[1, 2, 1, "", "forward"]], "brevitas.core.utils.SingleArgStatelessBuffer": [[1, 2, 1, "", "forward"]], "brevitas.core.utils.SliceTensor": [[1, 2, 1, "", "eager_forward"], [1, 2, 1, "", "forward"]], "brevitas.core.utils.StatelessBuffer": [[1, 2, 1, "", "forward"], [1, 2, 1, "", "state_dict"]], "brevitas.core.zero_point": [[1, 1, 1, "", "ParameterFromRuntimeZeroPoint"], [1, 1, 1, "", "ParameterFromStatsFromParameterZeroPoint"], [1, 1, 1, "", "ParameterZeroPoint"], [1, 1, 1, "", "PreZeroCenterZeroPoint"], [1, 1, 1, "", "StatsFromParameterZeroPoint"], [1, 1, 1, "", "ZeroZeroPoint"]], "brevitas.core.zero_point.ParameterFromRuntimeZeroPoint": [[1, 2, 1, "", "forward"], [1, 2, 1, "", "state_dict"], [1, 2, 1, "", "training_forward"]], "brevitas.core.zero_point.ParameterFromStatsFromParameterZeroPoint": [[1, 2, 1, "", "forward"], [1, 2, 1, "", "state_dict"]], "brevitas.core.zero_point.ParameterZeroPoint": [[1, 2, 1, "", "forward"]], "brevitas.core.zero_point.PreZeroCenterZeroPoint": [[1, 2, 1, "", "forward"], [1, 2, 1, "", "get_zero_center"]], "brevitas.core.zero_point.StatsFromParameterZeroPoint": [[1, 2, 1, "", "forward"]], "brevitas.core.zero_point.ZeroZeroPoint": [[1, 2, 1, "", "forward"]], "brevitas.function": [[7, 0, 0, "-", "ops"], [7, 0, 0, "-", "ops_ste"], [7, 0, 0, "-", "shape"]], "brevitas.function.ops": [[7, 4, 1, "", "binary_sign"], [7, 4, 1, "", "dpu_round"], [7, 4, 1, "", "get_upper_bound_on_l1_norm"], [7, 4, 1, "", "identity"], [7, 4, 1, "", "max_float"], [7, 4, 1, "", "max_int"], [7, 4, 1, "", "min_int"], [7, 4, 1, "", "round_to_zero"], [7, 4, 1, "", "tensor_clamp"], [7, 4, 1, "", "tensor_clamp_"]], "brevitas.function.ops_ste": [[7, 4, 1, "", "abs_binary_sign_grad"], [7, 4, 1, "", "binary_sign_ste"], [7, 4, 1, "", "ceil_ste"], [7, 4, 1, "", "dpu_round_ste"], [7, 4, 1, "", "floor_ste"], [7, 4, 1, "", "round_ste"], [7, 4, 1, "", "round_to_zero_ste"], [7, 4, 1, "", "scalar_clamp_min_ste"], [7, 4, 1, "", "scalar_clamp_ste"], [7, 4, 1, "", "tensor_clamp_ste"], [7, 4, 1, "", "tensor_clamp_ste_"], [7, 4, 1, "", "ternary_sign_ste"]], "brevitas.function.shape": [[7, 4, 1, "", "over_batch_over_output_channels"], [7, 4, 1, "", "over_batch_over_tensor"], [7, 4, 1, "", "over_output_channels"], [7, 4, 1, "", "over_output_features"], [7, 4, 1, "", "over_tensor"]], "brevitas.ops": [[8, 0, 0, "-", "autograd_ste_ops"]], "brevitas.ops.autograd_ste_ops": [[8, 1, 1, "", "AbsBinarySignGradFn"], [8, 1, 1, "", "BinarySignSteFn"], [8, 1, 1, "", "CeilSteFn"], [8, 1, 1, "", "DPURoundSteFn"], [8, 1, 1, "", "FloorSteFn"], [8, 1, 1, "", "InplaceTensorClampSteFn"], [8, 1, 1, "", "RoundSteFn"], [8, 1, 1, "", "RoundToZeroSteFn"], [8, 1, 1, "", "ScalarClampMinSteFn"], [8, 1, 1, "", "ScalarClampSteFn"], [8, 1, 1, "", "TensorClampSteFn"], [8, 1, 1, "", "TernarySignSteFn"], [8, 4, 1, "", "abs_binary_sign_grad_impl"], [8, 4, 1, "", "binary_sign_ste_impl"], [8, 4, 1, "", "ceil_ste_impl"], [8, 4, 1, "", "dpu_round_ste_impl"], [8, 4, 1, "", "floor_ste_impl"], [8, 4, 1, "", "round_ste_impl"], [8, 4, 1, "", "round_to_zero_ste_impl"], [8, 4, 1, "", "scalar_clamp_min_ste_impl"], [8, 4, 1, "", "scalar_clamp_ste_impl"], [8, 4, 1, "", "tensor_clamp_ste_impl"], [8, 4, 1, "", "ternary_sign_ste_impl"]]}, "objtypes": {"0": "py:module", "1": "py:class", "2": "py:method", "3": "py:attribute", "4": "py:function"}, "objnames": {"0": ["py", "module", "Python module"], "1": ["py", "class", "Python class"], "2": ["py", "method", "Python method"], "3": ["py", "attribute", "Python attribute"], "4": ["py", "function", "Python function"]}, "titleterms": {"about": 0, "author": 0, "cite": 0, "brevita": [1, 2, 3, 4, 5, 6, 7, 8, 12, 13, 22], "core": [1, 2, 3, 4, 5, 6, 10], "packag": [1, 2, 3, 4, 5, 6, 7, 8], "subpackag": [1, 9], "submodul": [1, 2, 3, 4, 5, 6, 7, 8], "restrict_v": 1, "modul": [1, 2, 3, 4, 5, 6, 7, 8], "util": 1, "zero_point": 1, "content": [1, 2, 3, 4, 5, 6], "bit_width": 2, "const": 2, "paramet": 2, "function_wrapp": 3, "clamp": 3, "misc": 3, "ops_st": [3, 7], "shape": [3, 7, 10], "quant": [4, 10], "binari": [4, 16, 22], "delai": 4, "int": 4, "int_bas": 4, "ternari": 4, "scale": [5, 22], "int_scal": 5, "runtim": [5, 18], "standalon": 5, "stat": 6, "stats_op": 6, "stats_wrapp": 6, "view_wrapp": 6, "function": [7, 10, 22], "op": [7, 8], "autograd_ste_op": 8, "api": [9, 16], "refer": 9, "architectur": 10, "scriptmodul": 10, "injector": 10, "quantiz": [10, 12, 16, 19, 20, 21, 22], "enum": [10, 22], "solver": 10, "quanttensor": [10, 21, 22], "proxi": [10, 16], "layer": [10, 12, 16, 22], "mixin": 10, "export": [10, 12, 18, 20, 22, 24], "fx": [10, 22], "graph": 10, "trace": 10, "transform": 10, "loss": 10, "f": 11, "A": [11, 16], "q": 11, "get": 12, "start": 12, "ptq": 12, "over": 12, "hand": 12, "programmat": 12, "defin": [12, 22], "model": [12, 18], "nn": 12, "weight": [12, 16, 20, 22], "onli": [12, 20], "float": [12, 22], "activ": [12, 16, 19, 22], "bias": 12, "onnx": [12, 18, 22], "where": 12, "go": 12, "from": [12, 15, 22], "here": 12, "set": 14, "setup": 15, "requir": [15, 18], "instal": 15, "pypi": 15, "github": 15, "option": 15, "train": [15, 22], "infer": 15, "anatomi": 16, "what": 16, "": 16, "auto": 16, "wire": 16, "depend": [16, 22], "inject": [16, 22], "practic": 16, "exampl": [16, 18], "manual": 16, "an": [16, 19, 21], "extendedinjector": 16, "inherit": [16, 22], "composit": 16, "interfac": 16, "pass": 16, "custom": [16, 22], "quantident": [16, 22], "initi": 16, "statist": 16, "share": [16, 22], "instanc": 16, "deal": 16, "build": 16, "tutori": [17, 22], "introduct": 18, "quantizelinear": 18, "clip": 18, "dequantizelinear": 18, "qcdq": 18, "basic": 18, "complet": 18, "The": 18, "c": 18, "bitwidth": 18, "8": 18, "qop": [18, 22], "qgemm": 18, "v": 18, "gemm": 18, "overview": [19, 21], "rnn": 20, "lstm": 20, "quantrnn": 20, "quantlstm": 20, "just": 20, "time": 20, "compil": 20, "calibr": [20, 22], "full": 20, "quantconv2d": 21, "input": [21, 22], "output": [21, 22], "bia": [21, 22], "tvmcon": 22, "2021": 22, "fundament": 22, "quantlinear": 22, "default": 22, "mix": 22, "point": 22, "fix": 22, "quantrelu": 22, "requant": 22, "tensor": 22, "how": 22, "i": 22, "determin": 22, "oper": 22, "element": 22, "wise": 22, "arithmet": 22, "add": 22, "call": 22, "torch": 22, "max_pool": 22, "tanh": 22, "concaten": 22, "common": 22, "keyword": 22, "argument": 22, "bit": 22, "width": 22, "per": 22, "channel": 22, "max_val": 22, "init": 22, "scratch": 22, "learn": 22, "retrain": 22, "among": 22, "torchscript": 22, "backend": 22, "base": 22, "post": 22, "compat": 24, "user": 26, "guid": 26}, "envversion": {"sphinx.domains.c": 2, "sphinx.domains.changeset": 1, "sphinx.domains.citation": 1, "sphinx.domains.cpp": 8, "sphinx.domains.index": 1, "sphinx.domains.javascript": 2, "sphinx.domains.math": 2, "sphinx.domains.python": 3, "sphinx.domains.rst": 2, "sphinx.domains.std": 2, "sphinx.ext.intersphinx": 1, "sphinx.ext.viewcode": 1, "nbsphinx": 4, "sphinx": 57}, "alltitles": {"About": [[0, "about"]], "Author": [[0, "author"]], "Cite as": [[0, "cite-as"]], "brevitas.core package": [[1, "brevitas-core-package"]], "Subpackages": [[1, "subpackages"], [9, "subpackages"]], "Submodules": [[1, "submodules"], [2, "submodules"], [3, "submodules"], [4, "submodules"], [5, "submodules"], [6, "submodules"], [7, "submodules"], [8, "submodules"]], "brevitas.core.restrict_val module": [[1, "module-brevitas.core.restrict_val"]], "brevitas.core.utils module": [[1, "module-brevitas.core.utils"]], "brevitas.core.zero_point module": [[1, "module-brevitas.core.zero_point"]], "Module contents": [[1, "module-brevitas.core"], [2, "module-brevitas.core.bit_width"], [3, "module-brevitas.core.function_wrapper"], [4, "module-brevitas.core.quant"], [5, "module-brevitas.core.scaling"], [6, "module-brevitas.core.stats"]], "brevitas.core.bit_width package": [[2, "brevitas-core-bit-width-package"]], "brevitas.core.bit_width.const module": [[2, "module-brevitas.core.bit_width.const"]], "brevitas.core.bit_width.parameter module": [[2, "module-brevitas.core.bit_width.parameter"]], "brevitas.core.function_wrapper package": [[3, "brevitas-core-function-wrapper-package"]], "brevitas.core.function_wrapper.clamp module": [[3, "module-brevitas.core.function_wrapper.clamp"]], "brevitas.core.function_wrapper.misc module": [[3, "module-brevitas.core.function_wrapper.misc"]], "brevitas.core.function_wrapper.ops_ste module": [[3, "module-brevitas.core.function_wrapper.ops_ste"]], "brevitas.core.function_wrapper.shape module": [[3, "module-brevitas.core.function_wrapper.shape"]], "brevitas.core.quant package": [[4, "brevitas-core-quant-package"]], "brevitas.core.quant.binary module": [[4, "module-brevitas.core.quant.binary"]], "brevitas.core.quant.delay module": [[4, "module-brevitas.core.quant.delay"]], "brevitas.core.quant.int module": [[4, "module-brevitas.core.quant.int"]], "brevitas.core.quant.int_base module": [[4, "module-brevitas.core.quant.int_base"]], "brevitas.core.quant.ternary module": [[4, "module-brevitas.core.quant.ternary"]], "brevitas.core.scaling package": [[5, "brevitas-core-scaling-package"]], "brevitas.core.scaling.int_scaling module": [[5, "module-brevitas.core.scaling.int_scaling"]], "brevitas.core.scaling.runtime module": [[5, "module-brevitas.core.scaling.runtime"]], "brevitas.core.scaling.standalone module": [[5, "module-brevitas.core.scaling.standalone"]], "brevitas.core.stats package": [[6, "brevitas-core-stats-package"]], "brevitas.core.stats.stats_op module": [[6, "module-brevitas.core.stats.stats_op"]], "brevitas.core.stats.stats_wrapper module": [[6, "module-brevitas.core.stats.stats_wrapper"]], "brevitas.core.stats.view_wrapper module": [[6, "module-brevitas.core.stats.view_wrapper"]], "brevitas.function package": [[7, "brevitas-function-package"]], "brevitas.function.ops module": [[7, "module-brevitas.function.ops"]], "brevitas.function.ops_ste module": [[7, "module-brevitas.function.ops_ste"]], "brevitas.function.shape module": [[7, "module-brevitas.function.shape"]], "brevitas.ops package": [[8, "brevitas-ops-package"]], "brevitas.ops.autograd_ste_ops module": [[8, "module-brevitas.ops.autograd_ste_ops"]], "API reference": [[9, "api-reference"]], "Architecture": [[10, "architecture"]], "Functions": [[10, "functions"]], "Core ScriptModules": [[10, "core-scriptmodules"]], "Injectors and Quantizers": [[10, "injectors-and-quantizers"]], "Enums, Shapes and Solvers": [[10, "enums-shapes-and-solvers"]], "QuantTensor": [[10, "quanttensor"], [21, "QuantTensor"]], "Proxies": [[10, "proxies"]], "Quant Layers and Mixins": [[10, "quant-layers-and-mixins"]], "Export": [[10, "export"], [20, "Export"], [22, "Export"]], "FX graph tracing and transformations": [[10, "fx-graph-tracing-and-transformations"]], "Losses": [[10, "losses"]], "F.A.Q.": [[11, "f-a-q"]], "Getting started": [[12, "getting-started"]], "PTQ over hand or programmatically defined quantized models": [[12, "ptq-over-hand-or-programmatically-defined-quantized-models"]], "Defining a quantized model with brevitas.nn layers": [[12, "defining-a-quantized-model-with-brevitas-nn-layers"]], "Weights-only quantization, float activations and biases": [[12, "weights-only-quantization-float-activations-and-biases"]], "Weights and activations quantization, float biases": [[12, "weights-and-activations-quantization-float-biases"]], "Weights, activations, biases quantization": [[12, "weights-activations-biases-quantization"]], "Export to ONNX": [[12, "export-to-onnx"]], "Where to go from here": [[12, "where-to-go-from-here"]], "Brevitas": [[13, "brevitas"]], "Settings": [[14, "settings"]], "Setup": [[15, "setup"]], "Requirements": [[15, "requirements"], [18, "Requirements"]], "Installation Requirements": [[15, "installation-requirements"]], "Installation": [[15, "installation"]], "Installing from PyPI": [[15, "installing-from-pypi"]], "Installing from Github": [[15, "installing-from-github"]], "Optional Training Requirements": [[15, "optional-training-requirements"]], "Optional Inference Requirements": [[15, "optional-inference-requirements"]], "Anatomy of a Quantizer": [[16, "Anatomy-of-a-Quantizer"]], "What\u2019s in a Quantizer?": [[16, "What's-in-a-Quantizer?"]], "Quantization with auto-wiring Dependency Injection": [[16, "Quantization-with-auto-wiring-Dependency-Injection"]], "A Practical Example: Binary Quantization": [[16, "A-Practical-Example:-Binary-Quantization"]], "Manual Binary Quantization": [[16, "Manual-Binary-Quantization"]], "Binary Quantization with an ExtendedInjector": [[16, "Binary-Quantization-with-an-ExtendedInjector"]], "Inheritance and Composition of Quantizers": [[16, "Inheritance-and-Composition-of-Quantizers"]], "Interfacing a Quantizer with a Quantized Layer": [[16, "Interfacing-a-Quantizer-with-a-Quantized-Layer"]], "Passing a custom quantizer to QuantIdentity": [[16, "Passing-a-custom-quantizer-to-QuantIdentity"]], "A Custom Quantizer initialized with Weight Statistics": [[16, "A-Custom-Quantizer-initialized-with-Weight-Statistics"]], "Sharing a Quantizer": [[16, "Sharing-a-Quantizer"]], "Sharing a proxy": [[16, "Sharing-a-proxy"]], "Sharing an instance of Activation Quantization": [[16, "Sharing-an-instance-of-Activation-Quantization"]], "Dealing with Weight Initialization": [[16, "Dealing-with-Weight-Initialization"]], "Building a Custom Quantization API": [[16, "Building-a-Custom-Quantization-API"]], "Tutorials": [[17, "tutorials"]], "Tutorials:": [[17, null]], "ONNX Export": [[18, "ONNX-Export"]], "Introduction": [[18, "Introduction"]], "QuantizeLinear-Clip-DeQuantizeLinear (QCDQ)": [[18, "QuantizeLinear-Clip-DeQuantizeLinear-(QCDQ)"]], "Basic Example": [[18, "Basic-Example"]], "Complete Model": [[18, "Complete-Model"]], "The C in QCDQ (Bitwidth <= 8)": [[18, "The-C-in-QCDQ-(Bitwidth-<=-8)"]], "QOps Export": [[18, "QOps-Export"]], "Clipping in QOps": [[18, "Clipping-in-QOps"]], "ONNX Runtime": [[18, "ONNX-Runtime"]], "QCDQ": [[18, "QCDQ"]], "QGEMM vs GEMM": [[18, "QGEMM-vs-GEMM"]], "QOps": [[18, "QOps"]], "An Overview of Quantized Activations": [[19, "An-Overview-of-Quantized-Activations"]], "Quantized RNNs and LSTMs": [[20, "Quantized-RNNs-and-LSTMs"]], "QuantRNN": [[20, "QuantRNN"]], "QuantLSTM": [[20, "QuantLSTM"]], "Just-in-time compilation": [[20, "Just-in-time-compilation"]], "Calibration": [[20, "Calibration"]], "QuantLSTM weight-only quantization export": [[20, "QuantLSTM-weight-only-quantization-export"]], "QuantLSTM full quantization export": [[20, "QuantLSTM-full-quantization-export"]], "An overview of QuantTensor and QuantConv2d": [[21, "An-overview-of-QuantTensor-and-QuantConv2d"]], "Input Quantization": [[21, "Input-Quantization"]], "Output Quantization": [[21, "Output-Quantization"]], "Bias Quantization": [[21, "Bias-Quantization"], [22, "Bias-Quantization"]], "Brevitas TVMCon 2021 tutorial": [[22, "Brevitas-TVMCon-2021-tutorial"]], "Fundamentals": [[22, "Fundamentals"]], "QuantLinear layer": [[22, "QuantLinear-layer"]], "Weight quantization": [[22, "Weight-quantization"]], "Default weight quantization": [[22, "Default-weight-quantization"]], "Mixing quantized weights and floating-point inputs": [[22, "Mixing-quantized-weights-and-floating-point-inputs"]], "Fixed-point weight quantization": [[22, "Fixed-point-weight-quantization"]], "Binary weight quantization": [[22, "Binary-weight-quantization"]], "Sharing a weight quantizer": [[22, "Sharing-a-weight-quantizer"]], "Inputs/Outputs/Activations quantization:": [[22, "Inputs/Outputs/Activations-quantization:"]], "QuantIdentity layer": [[22, "QuantIdentity-layer"]], "QuantReLU layer": [[22, "QuantReLU-layer"]], "Requantizing a tensor": [[22, "Requantizing-a-tensor"]], "How is the activation scale determined by default?": [[22, "How-is-the-activation-scale-determined-by-default?"]], "Operations on QuantTensor": [[22, "Operations-on-QuantTensor"]], "Element-wise Arithmetic": [[22, "Element-wise-Arithmetic"]], "Element-wise adds": [[22, "Element-wise-adds"]], "Calling torch functions": [[22, "Calling-torch-functions"]], "max_pool on QuantTensor": [[22, "max_pool-on-QuantTensor"]], "tanh on QuantTensor": [[22, "tanh-on-QuantTensor"]], "QuantTensor concatenation": [[22, "QuantTensor-concatenation"]], "Customizing Quantizers": [[22, "Customizing-Quantizers"]], "Common keyword arguments": [[22, "Common-keyword-arguments"]], "Weight bit-width": [[22, "Weight-bit-width"]], "Per-channel weight quantization": [[22, "Per-channel-weight-quantization"]], "Activation bit-width": [[22, "Activation-bit-width"]], "Activation quantization with max_val init": [[22, "Activation-quantization-with-max_val-init"]], "Per-channel activation quantization": [[22, "Per-channel-activation-quantization"]], "Inheriting from a quantizer": [[22, "Inheriting-from-a-quantizer"]], "Defining a quantizer from scratch with enums": [[22, "Defining-a-quantizer-from-scratch-with-enums"]], "Weight quantizer": [[22, "Weight-quantizer"]], "Activation quantizer": [[22, "Activation-quantizer"]], "Learned scale and bit-width quantizer": [[22, "Learned-scale-and-bit-width-quantizer"]], "Retraining from floating-point": [[22, "Retraining-from-floating-point"]], "Defining a quantizer from scratch with dependency-injection": [[22, "Defining-a-quantizer-from-scratch-with-dependency-injection"]], "Activation quantization from scratch": [[22, "Activation-quantization-from-scratch"]], "Weight quantization with learned scale from scratch": [[22, "Weight-quantization-with-learned-scale-from-scratch"]], "Sharing learned bit-width among layers": [[22, "Sharing-learned-bit-width-among-layers"]], "Export to ONNX QOps": [[22, "Export-to-ONNX-QOps"]], "Export to custom Quantized ONNX": [[22, "Export-to-custom-Quantized-ONNX"]], "Export to TorchScript quantization backend": [[22, "Export-to-TorchScript-quantization-backend"]], "Brevitas and FX": [[22, "Brevitas-and-FX"]], "Calibration-based post-training quantization": [[22, "Calibration-based-post-training-quantization"]], "Export Compatibility": [[24, "export-compatibility"]], "User Guide": [[26, "user-guide"]]}, "indexentries": {"floatrestrictvalue (class in brevitas.core.restrict_val)": [[1, "brevitas.core.restrict_val.FloatRestrictValue"]], "intrestrictvalue (class in brevitas.core.restrict_val)": [[1, "brevitas.core.restrict_val.IntRestrictValue"]], "logfloatrestrictvalue (class in brevitas.core.restrict_val)": [[1, "brevitas.core.restrict_val.LogFloatRestrictValue"]], "parameterfromruntimezeropoint (class in brevitas.core.zero_point)": [[1, "brevitas.core.zero_point.ParameterFromRuntimeZeroPoint"]], "parameterfromstatsfromparameterzeropoint (class in brevitas.core.zero_point)": [[1, "brevitas.core.zero_point.ParameterFromStatsFromParameterZeroPoint"]], "parameterwrapper (class in brevitas.core.utils)": [[1, "brevitas.core.utils.ParameterWrapper"]], "parameterzeropoint (class in brevitas.core.zero_point)": [[1, "brevitas.core.zero_point.ParameterZeroPoint"]], "poweroftworestrictvalue (class in brevitas.core.restrict_val)": [[1, "brevitas.core.restrict_val.PowerOfTwoRestrictValue"]], "prezerocenterzeropoint (class in brevitas.core.zero_point)": [[1, "brevitas.core.zero_point.PreZeroCenterZeroPoint"]], "singleargstatelessbuffer (class in brevitas.core.utils)": [[1, "brevitas.core.utils.SingleArgStatelessBuffer"]], "slicetensor (class in brevitas.core.utils)": [[1, "brevitas.core.utils.SliceTensor"]], "statelessbuffer (class in brevitas.core.utils)": [[1, "brevitas.core.utils.StatelessBuffer"]], "statsfromparameterzeropoint (class in brevitas.core.zero_point)": [[1, "brevitas.core.zero_point.StatsFromParameterZeroPoint"]], "zerozeropoint (class in brevitas.core.zero_point)": [[1, "brevitas.core.zero_point.ZeroZeroPoint"]], "brevitas.core": [[1, "module-brevitas.core"]], "brevitas.core.restrict_val": [[1, "module-brevitas.core.restrict_val"]], "brevitas.core.utils": [[1, "module-brevitas.core.utils"]], "brevitas.core.zero_point": [[1, "module-brevitas.core.zero_point"]], "eager_forward() (brevitas.core.utils.slicetensor method)": [[1, "brevitas.core.utils.SliceTensor.eager_forward"]], "forward() (brevitas.core.restrict_val.floatrestrictvalue method)": [[1, "brevitas.core.restrict_val.FloatRestrictValue.forward"]], "forward() (brevitas.core.restrict_val.intrestrictvalue method)": [[1, "brevitas.core.restrict_val.IntRestrictValue.forward"]], "forward() (brevitas.core.restrict_val.logfloatrestrictvalue method)": [[1, "brevitas.core.restrict_val.LogFloatRestrictValue.forward"]], "forward() (brevitas.core.restrict_val.poweroftworestrictvalue method)": [[1, "brevitas.core.restrict_val.PowerOfTwoRestrictValue.forward"]], "forward() (brevitas.core.utils.parameterwrapper method)": [[1, "brevitas.core.utils.ParameterWrapper.forward"]], "forward() (brevitas.core.utils.singleargstatelessbuffer method)": [[1, "brevitas.core.utils.SingleArgStatelessBuffer.forward"]], "forward() (brevitas.core.utils.slicetensor method)": [[1, "brevitas.core.utils.SliceTensor.forward"]], "forward() (brevitas.core.utils.statelessbuffer method)": [[1, "brevitas.core.utils.StatelessBuffer.forward"]], "forward() (brevitas.core.zero_point.parameterfromruntimezeropoint method)": [[1, "brevitas.core.zero_point.ParameterFromRuntimeZeroPoint.forward"]], "forward() (brevitas.core.zero_point.parameterfromstatsfromparameterzeropoint method)": [[1, "brevitas.core.zero_point.ParameterFromStatsFromParameterZeroPoint.forward"]], "forward() (brevitas.core.zero_point.parameterzeropoint method)": [[1, "brevitas.core.zero_point.ParameterZeroPoint.forward"]], "forward() (brevitas.core.zero_point.prezerocenterzeropoint method)": [[1, "brevitas.core.zero_point.PreZeroCenterZeroPoint.forward"]], "forward() (brevitas.core.zero_point.statsfromparameterzeropoint method)": [[1, "brevitas.core.zero_point.StatsFromParameterZeroPoint.forward"]], "forward() (brevitas.core.zero_point.zerozeropoint method)": [[1, "brevitas.core.zero_point.ZeroZeroPoint.forward"]], "get_zero_center() (brevitas.core.zero_point.prezerocenterzeropoint method)": [[1, "brevitas.core.zero_point.PreZeroCenterZeroPoint.get_zero_center"]], "inplace_momentum_update() (in module brevitas.core.utils)": [[1, "brevitas.core.utils.inplace_momentum_update"]], "inplace_tensor_add() (in module brevitas.core.utils)": [[1, "brevitas.core.utils.inplace_tensor_add"]], "inplace_tensor_mul() (in module brevitas.core.utils)": [[1, "brevitas.core.utils.inplace_tensor_mul"]], "module": [[1, "module-brevitas.core"], [1, "module-brevitas.core.restrict_val"], [1, "module-brevitas.core.utils"], [1, "module-brevitas.core.zero_point"], [2, "module-brevitas.core.bit_width"], [2, "module-brevitas.core.bit_width.const"], [2, "module-brevitas.core.bit_width.parameter"], [3, "module-brevitas.core.function_wrapper"], [3, "module-brevitas.core.function_wrapper.clamp"], [3, "module-brevitas.core.function_wrapper.misc"], [3, "module-brevitas.core.function_wrapper.ops_ste"], [3, "module-brevitas.core.function_wrapper.shape"], [4, "module-brevitas.core.quant"], [4, "module-brevitas.core.quant.binary"], [4, "module-brevitas.core.quant.delay"], [4, "module-brevitas.core.quant.int"], [4, "module-brevitas.core.quant.int_base"], [4, "module-brevitas.core.quant.ternary"], [5, "module-brevitas.core.scaling"], [5, "module-brevitas.core.scaling.int_scaling"], [5, "module-brevitas.core.scaling.runtime"], [5, "module-brevitas.core.scaling.standalone"], [6, "module-brevitas.core.stats"], [6, "module-brevitas.core.stats.stats_op"], [6, "module-brevitas.core.stats.stats_wrapper"], [6, "module-brevitas.core.stats.view_wrapper"], [7, "module-brevitas.function.ops"], [7, "module-brevitas.function.ops_ste"], [7, "module-brevitas.function.shape"], [8, "module-brevitas.ops.autograd_ste_ops"]], "restrict_init_float() (brevitas.core.restrict_val.floatrestrictvalue method)": [[1, "brevitas.core.restrict_val.FloatRestrictValue.restrict_init_float"]], "restrict_init_float() (brevitas.core.restrict_val.intrestrictvalue method)": [[1, "brevitas.core.restrict_val.IntRestrictValue.restrict_init_float"]], "restrict_init_float() (brevitas.core.restrict_val.logfloatrestrictvalue method)": [[1, "brevitas.core.restrict_val.LogFloatRestrictValue.restrict_init_float"]], "restrict_init_float() (brevitas.core.restrict_val.poweroftworestrictvalue method)": [[1, "brevitas.core.restrict_val.PowerOfTwoRestrictValue.restrict_init_float"]], "restrict_init_inplace_module() (brevitas.core.restrict_val.floatrestrictvalue method)": [[1, "brevitas.core.restrict_val.FloatRestrictValue.restrict_init_inplace_module"]], "restrict_init_inplace_module() (brevitas.core.restrict_val.intrestrictvalue method)": [[1, "brevitas.core.restrict_val.IntRestrictValue.restrict_init_inplace_module"]], "restrict_init_inplace_module() (brevitas.core.restrict_val.logfloatrestrictvalue method)": [[1, "brevitas.core.restrict_val.LogFloatRestrictValue.restrict_init_inplace_module"]], "restrict_init_inplace_module() (brevitas.core.restrict_val.poweroftworestrictvalue method)": [[1, "brevitas.core.restrict_val.PowerOfTwoRestrictValue.restrict_init_inplace_module"]], "restrict_init_module() (brevitas.core.restrict_val.floatrestrictvalue method)": [[1, "brevitas.core.restrict_val.FloatRestrictValue.restrict_init_module"]], "restrict_init_module() (brevitas.core.restrict_val.intrestrictvalue method)": [[1, "brevitas.core.restrict_val.IntRestrictValue.restrict_init_module"]], "restrict_init_module() (brevitas.core.restrict_val.logfloatrestrictvalue method)": [[1, "brevitas.core.restrict_val.LogFloatRestrictValue.restrict_init_module"]], "restrict_init_module() (brevitas.core.restrict_val.poweroftworestrictvalue method)": [[1, "brevitas.core.restrict_val.PowerOfTwoRestrictValue.restrict_init_module"]], "restrict_init_tensor() (brevitas.core.restrict_val.floatrestrictvalue method)": [[1, "brevitas.core.restrict_val.FloatRestrictValue.restrict_init_tensor"]], "restrict_init_tensor() (brevitas.core.restrict_val.intrestrictvalue method)": [[1, "brevitas.core.restrict_val.IntRestrictValue.restrict_init_tensor"]], "restrict_init_tensor() (brevitas.core.restrict_val.logfloatrestrictvalue method)": [[1, "brevitas.core.restrict_val.LogFloatRestrictValue.restrict_init_tensor"]], "restrict_init_tensor() (brevitas.core.restrict_val.poweroftworestrictvalue method)": [[1, "brevitas.core.restrict_val.PowerOfTwoRestrictValue.restrict_init_tensor"]], "state_dict() (brevitas.core.utils.statelessbuffer method)": [[1, "brevitas.core.utils.StatelessBuffer.state_dict"]], "state_dict() (brevitas.core.zero_point.parameterfromruntimezeropoint method)": [[1, "brevitas.core.zero_point.ParameterFromRuntimeZeroPoint.state_dict"]], "state_dict() (brevitas.core.zero_point.parameterfromstatsfromparameterzeropoint method)": [[1, "brevitas.core.zero_point.ParameterFromStatsFromParameterZeroPoint.state_dict"]], "training_forward() (brevitas.core.zero_point.parameterfromruntimezeropoint method)": [[1, "brevitas.core.zero_point.ParameterFromRuntimeZeroPoint.training_forward"]], "bitwidthconst (class in brevitas.core.bit_width.const)": [[2, "brevitas.core.bit_width.const.BitWidthConst"]], "bitwidthparameter (class in brevitas.core.bit_width.parameter)": [[2, "brevitas.core.bit_width.parameter.BitWidthParameter"]], "bitwidthstatefulconst (class in brevitas.core.bit_width.const)": [[2, "brevitas.core.bit_width.const.BitWidthStatefulConst"]], "msbclampbitwidth (class in brevitas.core.bit_width.const)": [[2, "brevitas.core.bit_width.const.MsbClampBitWidth"]], "removebitwidthparameter (class in brevitas.core.bit_width.parameter)": [[2, "brevitas.core.bit_width.parameter.RemoveBitwidthParameter"]], "brevitas.core.bit_width": [[2, "module-brevitas.core.bit_width"]], "brevitas.core.bit_width.const": [[2, "module-brevitas.core.bit_width.const"]], "brevitas.core.bit_width.parameter": [[2, "module-brevitas.core.bit_width.parameter"]], "forward() (brevitas.core.bit_width.const.bitwidthconst method)": [[2, "brevitas.core.bit_width.const.BitWidthConst.forward"]], "forward() (brevitas.core.bit_width.const.bitwidthstatefulconst method)": [[2, "brevitas.core.bit_width.const.BitWidthStatefulConst.forward"]], "forward() (brevitas.core.bit_width.const.msbclampbitwidth method)": [[2, "brevitas.core.bit_width.const.MsbClampBitWidth.forward"]], "forward() (brevitas.core.bit_width.parameter.bitwidthparameter method)": [[2, "brevitas.core.bit_width.parameter.BitWidthParameter.forward"]], "forward() (brevitas.core.bit_width.parameter.removebitwidthparameter method)": [[2, "brevitas.core.bit_width.parameter.RemoveBitwidthParameter.forward"]], "ceilste (class in brevitas.core.function_wrapper.ops_ste)": [[3, "brevitas.core.function_wrapper.ops_ste.CeilSte"]], "clampmin (class in brevitas.core.function_wrapper.clamp)": [[3, "brevitas.core.function_wrapper.clamp.ClampMin"]], "dpuroundste (class in brevitas.core.function_wrapper.ops_ste)": [[3, "brevitas.core.function_wrapper.ops_ste.DPURoundSte"]], "floorste (class in brevitas.core.function_wrapper.ops_ste)": [[3, "brevitas.core.function_wrapper.ops_ste.FloorSte"]], "identity (class in brevitas.core.function_wrapper.misc)": [[3, "brevitas.core.function_wrapper.misc.Identity"]], "inplacelogtwo (class in brevitas.core.function_wrapper.misc)": [[3, "brevitas.core.function_wrapper.misc.InplaceLogTwo"]], "inplacetensorclampste (class in brevitas.core.function_wrapper.ops_ste)": [[3, "brevitas.core.function_wrapper.ops_ste.InplaceTensorClampSte"]], "logtwo (class in brevitas.core.function_wrapper.misc)": [[3, "brevitas.core.function_wrapper.misc.LogTwo"]], "over_batch_over_output_channels (brevitas.core.function_wrapper.shape.statsinputviewshapeimpl attribute)": [[3, "brevitas.core.function_wrapper.shape.StatsInputViewShapeImpl.OVER_BATCH_OVER_OUTPUT_CHANNELS"]], "over_batch_over_tensor (brevitas.core.function_wrapper.shape.statsinputviewshapeimpl attribute)": [[3, "brevitas.core.function_wrapper.shape.StatsInputViewShapeImpl.OVER_BATCH_OVER_TENSOR"]], "over_output_channels (brevitas.core.function_wrapper.shape.statsinputviewshapeimpl attribute)": [[3, "brevitas.core.function_wrapper.shape.StatsInputViewShapeImpl.OVER_OUTPUT_CHANNELS"]], "over_output_features (brevitas.core.function_wrapper.shape.statsinputviewshapeimpl attribute)": [[3, "brevitas.core.function_wrapper.shape.StatsInputViewShapeImpl.OVER_OUTPUT_FEATURES"]], "over_tensor (brevitas.core.function_wrapper.shape.statsinputviewshapeimpl attribute)": [[3, "brevitas.core.function_wrapper.shape.StatsInputViewShapeImpl.OVER_TENSOR"]], "overbatchoveroutputchannelview (class in brevitas.core.function_wrapper.shape)": [[3, "brevitas.core.function_wrapper.shape.OverBatchOverOutputChannelView"]], "overbatchovertensorview (class in brevitas.core.function_wrapper.shape)": [[3, "brevitas.core.function_wrapper.shape.OverBatchOverTensorView"]], "overoutputchannelview (class in brevitas.core.function_wrapper.shape)": [[3, "brevitas.core.function_wrapper.shape.OverOutputChannelView"]], "overoutputfeaturesview (class in brevitas.core.function_wrapper.shape)": [[3, "brevitas.core.function_wrapper.shape.OverOutputFeaturesView"]], "overtensorview (class in brevitas.core.function_wrapper.shape)": [[3, "brevitas.core.function_wrapper.shape.OverTensorView"]], "permutedims (class in brevitas.core.function_wrapper.shape)": [[3, "brevitas.core.function_wrapper.shape.PermuteDims"]], "poweroftwo (class in brevitas.core.function_wrapper.misc)": [[3, "brevitas.core.function_wrapper.misc.PowerOfTwo"]], "roundste (class in brevitas.core.function_wrapper.ops_ste)": [[3, "brevitas.core.function_wrapper.ops_ste.RoundSte"]], "roundtozeroste (class in brevitas.core.function_wrapper.ops_ste)": [[3, "brevitas.core.function_wrapper.ops_ste.RoundToZeroSte"]], "scalarclamp (class in brevitas.core.function_wrapper.clamp)": [[3, "brevitas.core.function_wrapper.clamp.ScalarClamp"]], "scalarclampminste (class in brevitas.core.function_wrapper.ops_ste)": [[3, "brevitas.core.function_wrapper.ops_ste.ScalarClampMinSte"]], "statsinputviewshapeimpl (class in brevitas.core.function_wrapper.shape)": [[3, "brevitas.core.function_wrapper.shape.StatsInputViewShapeImpl"]], "tensorclamp (class in brevitas.core.function_wrapper.clamp)": [[3, "brevitas.core.function_wrapper.clamp.TensorClamp"]], "tensorclampste (class in brevitas.core.function_wrapper.ops_ste)": [[3, "brevitas.core.function_wrapper.ops_ste.TensorClampSte"]], "brevitas.core.function_wrapper": [[3, "module-brevitas.core.function_wrapper"]], "brevitas.core.function_wrapper.clamp": [[3, "module-brevitas.core.function_wrapper.clamp"]], "brevitas.core.function_wrapper.misc": [[3, "module-brevitas.core.function_wrapper.misc"]], "brevitas.core.function_wrapper.ops_ste": [[3, "module-brevitas.core.function_wrapper.ops_ste"]], "brevitas.core.function_wrapper.shape": [[3, "module-brevitas.core.function_wrapper.shape"]], "forward() (brevitas.core.function_wrapper.clamp.clampmin method)": [[3, "brevitas.core.function_wrapper.clamp.ClampMin.forward"]], "forward() (brevitas.core.function_wrapper.clamp.scalarclamp method)": [[3, "brevitas.core.function_wrapper.clamp.ScalarClamp.forward"]], "forward() (brevitas.core.function_wrapper.clamp.tensorclamp method)": [[3, "brevitas.core.function_wrapper.clamp.TensorClamp.forward"]], "forward() (brevitas.core.function_wrapper.misc.identity method)": [[3, "brevitas.core.function_wrapper.misc.Identity.forward"]], "forward() (brevitas.core.function_wrapper.misc.inplacelogtwo method)": [[3, "brevitas.core.function_wrapper.misc.InplaceLogTwo.forward"]], "forward() (brevitas.core.function_wrapper.misc.logtwo method)": [[3, "brevitas.core.function_wrapper.misc.LogTwo.forward"]], "forward() (brevitas.core.function_wrapper.misc.poweroftwo method)": [[3, "brevitas.core.function_wrapper.misc.PowerOfTwo.forward"]], "forward() (brevitas.core.function_wrapper.ops_ste.ceilste method)": [[3, "brevitas.core.function_wrapper.ops_ste.CeilSte.forward"]], "forward() (brevitas.core.function_wrapper.ops_ste.dpuroundste method)": [[3, "brevitas.core.function_wrapper.ops_ste.DPURoundSte.forward"]], "forward() (brevitas.core.function_wrapper.ops_ste.floorste method)": [[3, "brevitas.core.function_wrapper.ops_ste.FloorSte.forward"]], "forward() (brevitas.core.function_wrapper.ops_ste.inplacetensorclampste method)": [[3, "brevitas.core.function_wrapper.ops_ste.InplaceTensorClampSte.forward"]], "forward() (brevitas.core.function_wrapper.ops_ste.roundste method)": [[3, "brevitas.core.function_wrapper.ops_ste.RoundSte.forward"]], "forward() (brevitas.core.function_wrapper.ops_ste.roundtozeroste method)": [[3, "brevitas.core.function_wrapper.ops_ste.RoundToZeroSte.forward"]], "forward() (brevitas.core.function_wrapper.ops_ste.scalarclampminste method)": [[3, "brevitas.core.function_wrapper.ops_ste.ScalarClampMinSte.forward"]], "forward() (brevitas.core.function_wrapper.ops_ste.tensorclampste method)": [[3, "brevitas.core.function_wrapper.ops_ste.TensorClampSte.forward"]], "forward() (brevitas.core.function_wrapper.shape.overbatchoveroutputchannelview method)": [[3, "brevitas.core.function_wrapper.shape.OverBatchOverOutputChannelView.forward"]], "forward() (brevitas.core.function_wrapper.shape.overbatchovertensorview method)": [[3, "brevitas.core.function_wrapper.shape.OverBatchOverTensorView.forward"]], "forward() (brevitas.core.function_wrapper.shape.overoutputchannelview method)": [[3, "brevitas.core.function_wrapper.shape.OverOutputChannelView.forward"]], "forward() (brevitas.core.function_wrapper.shape.overoutputfeaturesview method)": [[3, "brevitas.core.function_wrapper.shape.OverOutputFeaturesView.forward"]], "forward() (brevitas.core.function_wrapper.shape.overtensorview method)": [[3, "brevitas.core.function_wrapper.shape.OverTensorView.forward"]], "forward() (brevitas.core.function_wrapper.shape.permutedims method)": [[3, "brevitas.core.function_wrapper.shape.PermuteDims.forward"]], "binaryquant (class in brevitas.core.quant.binary)": [[4, "brevitas.core.quant.binary.BinaryQuant"]], "clampedbinaryquant (class in brevitas.core.quant.binary)": [[4, "brevitas.core.quant.binary.ClampedBinaryQuant"]], "decoupledintquant (class in brevitas.core.quant.int_base)": [[4, "brevitas.core.quant.int_base.DecoupledIntQuant"]], "decoupledrescalingintquant (class in brevitas.core.quant.int)": [[4, "brevitas.core.quant.int.DecoupledRescalingIntQuant"]], "decoupledrescalingintquantwithinput (class in brevitas.core.quant.int)": [[4, "brevitas.core.quant.int.DecoupledRescalingIntQuantWithInput"]], "delaywrapper (class in brevitas.core.quant.delay)": [[4, "brevitas.core.quant.delay.DelayWrapper"]], "intquant (class in brevitas.core.quant.int_base)": [[4, "brevitas.core.quant.int_base.IntQuant"]], "prescaledrestrictintquant (class in brevitas.core.quant.int)": [[4, "brevitas.core.quant.int.PrescaledRestrictIntQuant"]], "prescaledrestrictintquantwithinputbitwidth (class in brevitas.core.quant.int)": [[4, "brevitas.core.quant.int.PrescaledRestrictIntQuantWithInputBitWidth"]], "rescalingintquant (class in brevitas.core.quant.int)": [[4, "brevitas.core.quant.int.RescalingIntQuant"]], "ternaryquant (class in brevitas.core.quant.ternary)": [[4, "brevitas.core.quant.ternary.TernaryQuant"]], "truncintquant (class in brevitas.core.quant.int)": [[4, "brevitas.core.quant.int.TruncIntQuant"]], "brevitas.core.quant": [[4, "module-brevitas.core.quant"]], "brevitas.core.quant.binary": [[4, "module-brevitas.core.quant.binary"]], "brevitas.core.quant.delay": [[4, "module-brevitas.core.quant.delay"]], "brevitas.core.quant.int": [[4, "module-brevitas.core.quant.int"]], "brevitas.core.quant.int_base": [[4, "module-brevitas.core.quant.int_base"]], "brevitas.core.quant.ternary": [[4, "module-brevitas.core.quant.ternary"]], "forward() (brevitas.core.quant.binary.binaryquant method)": [[4, "brevitas.core.quant.binary.BinaryQuant.forward"]], "forward() (brevitas.core.quant.binary.clampedbinaryquant method)": [[4, "brevitas.core.quant.binary.ClampedBinaryQuant.forward"]], "forward() (brevitas.core.quant.delay.delaywrapper method)": [[4, "brevitas.core.quant.delay.DelayWrapper.forward"]], "forward() (brevitas.core.quant.int.decoupledrescalingintquant method)": [[4, "brevitas.core.quant.int.DecoupledRescalingIntQuant.forward"]], "forward() (brevitas.core.quant.int.decoupledrescalingintquantwithinput method)": [[4, "brevitas.core.quant.int.DecoupledRescalingIntQuantWithInput.forward"]], "forward() (brevitas.core.quant.int.prescaledrestrictintquant method)": [[4, "brevitas.core.quant.int.PrescaledRestrictIntQuant.forward"]], "forward() (brevitas.core.quant.int.prescaledrestrictintquantwithinputbitwidth method)": [[4, "brevitas.core.quant.int.PrescaledRestrictIntQuantWithInputBitWidth.forward"]], "forward() (brevitas.core.quant.int.rescalingintquant method)": [[4, "brevitas.core.quant.int.RescalingIntQuant.forward"]], "forward() (brevitas.core.quant.int.truncintquant method)": [[4, "brevitas.core.quant.int.TruncIntQuant.forward"]], "forward() (brevitas.core.quant.int_base.decoupledintquant method)": [[4, "brevitas.core.quant.int_base.DecoupledIntQuant.forward"]], "forward() (brevitas.core.quant.int_base.intquant method)": [[4, "brevitas.core.quant.int_base.IntQuant.forward"]], "forward() (brevitas.core.quant.ternary.ternaryquant method)": [[4, "brevitas.core.quant.ternary.TernaryQuant.forward"]], "max_int() (brevitas.core.quant.int_base.decoupledintquant method)": [[4, "brevitas.core.quant.int_base.DecoupledIntQuant.max_int"]], "max_int() (brevitas.core.quant.int_base.intquant method)": [[4, "brevitas.core.quant.int_base.IntQuant.max_int"]], "min_int() (brevitas.core.quant.int_base.decoupledintquant method)": [[4, "brevitas.core.quant.int_base.DecoupledIntQuant.min_int"]], "min_int() (brevitas.core.quant.int_base.intquant method)": [[4, "brevitas.core.quant.int_base.IntQuant.min_int"]], "to_int() (brevitas.core.quant.int_base.decoupledintquant method)": [[4, "brevitas.core.quant.int_base.DecoupledIntQuant.to_int"]], "to_int() (brevitas.core.quant.int_base.intquant method)": [[4, "brevitas.core.quant.int_base.IntQuant.to_int"]], "constscaling (class in brevitas.core.scaling.standalone)": [[5, "brevitas.core.scaling.standalone.ConstScaling"]], "intscaling (class in brevitas.core.scaling.int_scaling)": [[5, "brevitas.core.scaling.int_scaling.IntScaling"]], "parameterfromruntimestatsscaling (class in brevitas.core.scaling.standalone)": [[5, "brevitas.core.scaling.standalone.ParameterFromRuntimeStatsScaling"]], "parameterfromstatsfromparameterscaling (class in brevitas.core.scaling.standalone)": [[5, "brevitas.core.scaling.standalone.ParameterFromStatsFromParameterScaling"]], "parameterscaling (class in brevitas.core.scaling.standalone)": [[5, "brevitas.core.scaling.standalone.ParameterScaling"]], "poweroftwointscaling (class in brevitas.core.scaling.int_scaling)": [[5, "brevitas.core.scaling.int_scaling.PowerOfTwoIntScaling"]], "runtimestatsscaling (class in brevitas.core.scaling.runtime)": [[5, "brevitas.core.scaling.runtime.RuntimeStatsScaling"]], "statsfromparameterscaling (class in brevitas.core.scaling.runtime)": [[5, "brevitas.core.scaling.runtime.StatsFromParameterScaling"]], "brevitas.core.scaling": [[5, "module-brevitas.core.scaling"]], "brevitas.core.scaling.int_scaling": [[5, "module-brevitas.core.scaling.int_scaling"]], "brevitas.core.scaling.runtime": [[5, "module-brevitas.core.scaling.runtime"]], "brevitas.core.scaling.standalone": [[5, "module-brevitas.core.scaling.standalone"]], "forward() (brevitas.core.scaling.int_scaling.intscaling method)": [[5, "brevitas.core.scaling.int_scaling.IntScaling.forward"]], "forward() (brevitas.core.scaling.int_scaling.poweroftwointscaling method)": [[5, "brevitas.core.scaling.int_scaling.PowerOfTwoIntScaling.forward"]], "forward() (brevitas.core.scaling.runtime.runtimestatsscaling method)": [[5, "brevitas.core.scaling.runtime.RuntimeStatsScaling.forward"]], "forward() (brevitas.core.scaling.runtime.statsfromparameterscaling method)": [[5, "brevitas.core.scaling.runtime.StatsFromParameterScaling.forward"]], "forward() (brevitas.core.scaling.standalone.constscaling method)": [[5, "brevitas.core.scaling.standalone.ConstScaling.forward"]], "forward() (brevitas.core.scaling.standalone.parameterfromruntimestatsscaling method)": [[5, "brevitas.core.scaling.standalone.ParameterFromRuntimeStatsScaling.forward"]], "forward() (brevitas.core.scaling.standalone.parameterfromstatsfromparameterscaling method)": [[5, "brevitas.core.scaling.standalone.ParameterFromStatsFromParameterScaling.forward"]], "forward() (brevitas.core.scaling.standalone.parameterscaling method)": [[5, "brevitas.core.scaling.standalone.ParameterScaling.forward"]], "state_dict() (brevitas.core.scaling.standalone.parameterfromruntimestatsscaling method)": [[5, "brevitas.core.scaling.standalone.ParameterFromRuntimeStatsScaling.state_dict"]], "state_dict() (brevitas.core.scaling.standalone.parameterfromstatsfromparameterscaling method)": [[5, "brevitas.core.scaling.standalone.ParameterFromStatsFromParameterScaling.state_dict"]], "training_forward() (brevitas.core.scaling.standalone.parameterfromruntimestatsscaling method)": [[5, "brevitas.core.scaling.standalone.ParameterFromRuntimeStatsScaling.training_forward"]], "absave (class in brevitas.core.stats.stats_op)": [[6, "brevitas.core.stats.stats_op.AbsAve"]], "absmax (class in brevitas.core.stats.stats_op)": [[6, "brevitas.core.stats.stats_op.AbsMax"]], "absmaxave (class in brevitas.core.stats.stats_op)": [[6, "brevitas.core.stats.stats_op.AbsMaxAve"]], "absmaxl2 (class in brevitas.core.stats.stats_op)": [[6, "brevitas.core.stats.stats_op.AbsMaxL2"]], "absminmax (class in brevitas.core.stats.stats_op)": [[6, "brevitas.core.stats.stats_op.AbsMinMax"]], "abspercentile (class in brevitas.core.stats.stats_op)": [[6, "brevitas.core.stats.stats_op.AbsPercentile"]], "klminimizerthreshold (class in brevitas.core.stats.stats_op)": [[6, "brevitas.core.stats.stats_op.KLMinimizerThreshold"]], "l1norm (class in brevitas.core.stats.stats_op)": [[6, "brevitas.core.stats.stats_op.L1Norm"]], "l2norm (class in brevitas.core.stats.stats_op)": [[6, "brevitas.core.stats.stats_op.L2Norm"]], "mse (class in brevitas.core.stats.stats_op)": [[6, "brevitas.core.stats.stats_op.MSE"]], "meanlearnedsigmastd (class in brevitas.core.stats.stats_op)": [[6, "brevitas.core.stats.stats_op.MeanLearnedSigmaStd"]], "meansigmastd (class in brevitas.core.stats.stats_op)": [[6, "brevitas.core.stats.stats_op.MeanSigmaStd"]], "negativeminorzero (class in brevitas.core.stats.stats_op)": [[6, "brevitas.core.stats.stats_op.NegativeMinOrZero"]], "negativepercentileorzero (class in brevitas.core.stats.stats_op)": [[6, "brevitas.core.stats.stats_op.NegativePercentileOrZero"]], "percentileinterval (class in brevitas.core.stats.stats_op)": [[6, "brevitas.core.stats.stats_op.PercentileInterval"]], "brevitas.core.stats": [[6, "module-brevitas.core.stats"]], "brevitas.core.stats.stats_op": [[6, "module-brevitas.core.stats.stats_op"]], "brevitas.core.stats.stats_wrapper": [[6, "module-brevitas.core.stats.stats_wrapper"]], "brevitas.core.stats.view_wrapper": [[6, "module-brevitas.core.stats.view_wrapper"]], "evaluate_loss() (brevitas.core.stats.stats_op.mse method)": [[6, "brevitas.core.stats.stats_op.MSE.evaluate_loss"]], "forward() (brevitas.core.stats.stats_op.absave method)": [[6, "brevitas.core.stats.stats_op.AbsAve.forward"]], "forward() (brevitas.core.stats.stats_op.absmax method)": [[6, "brevitas.core.stats.stats_op.AbsMax.forward"]], "forward() (brevitas.core.stats.stats_op.absmaxave method)": [[6, "brevitas.core.stats.stats_op.AbsMaxAve.forward"]], "forward() (brevitas.core.stats.stats_op.absmaxl2 method)": [[6, "brevitas.core.stats.stats_op.AbsMaxL2.forward"]], "forward() (brevitas.core.stats.stats_op.absminmax method)": [[6, "brevitas.core.stats.stats_op.AbsMinMax.forward"]], "forward() (brevitas.core.stats.stats_op.abspercentile method)": [[6, "brevitas.core.stats.stats_op.AbsPercentile.forward"]], "forward() (brevitas.core.stats.stats_op.klminimizerthreshold method)": [[6, "brevitas.core.stats.stats_op.KLMinimizerThreshold.forward"]], "forward() (brevitas.core.stats.stats_op.l1norm method)": [[6, "brevitas.core.stats.stats_op.L1Norm.forward"]], "forward() (brevitas.core.stats.stats_op.l2norm method)": [[6, "brevitas.core.stats.stats_op.L2Norm.forward"]], "forward() (brevitas.core.stats.stats_op.mse method)": [[6, "brevitas.core.stats.stats_op.MSE.forward"]], "forward() (brevitas.core.stats.stats_op.meanlearnedsigmastd method)": [[6, "brevitas.core.stats.stats_op.MeanLearnedSigmaStd.forward"]], "forward() (brevitas.core.stats.stats_op.meansigmastd method)": [[6, "brevitas.core.stats.stats_op.MeanSigmaStd.forward"]], "forward() (brevitas.core.stats.stats_op.negativeminorzero method)": [[6, "brevitas.core.stats.stats_op.NegativeMinOrZero.forward"]], "forward() (brevitas.core.stats.stats_op.negativepercentileorzero method)": [[6, "brevitas.core.stats.stats_op.NegativePercentileOrZero.forward"]], "forward() (brevitas.core.stats.stats_op.percentileinterval method)": [[6, "brevitas.core.stats.stats_op.PercentileInterval.forward"]], "mse_fib_search() (brevitas.core.stats.stats_op.mse method)": [[6, "brevitas.core.stats.stats_op.MSE.mse_fib_search"]], "mse_grid_search() (brevitas.core.stats.stats_op.mse method)": [[6, "brevitas.core.stats.stats_op.MSE.mse_grid_search"]], "mse_loss_fn() (brevitas.core.stats.stats_op.mse method)": [[6, "brevitas.core.stats.stats_op.MSE.mse_loss_fn"]], "mse_search() (brevitas.core.stats.stats_op.mse method)": [[6, "brevitas.core.stats.stats_op.MSE.mse_search"]], "smooth_normalize_distribution() (brevitas.core.stats.stats_op.klminimizerthreshold method)": [[6, "brevitas.core.stats.stats_op.KLMinimizerThreshold.smooth_normalize_distribution"]], "abs_binary_sign_grad() (in module brevitas.function.ops_ste)": [[7, "brevitas.function.ops_ste.abs_binary_sign_grad"]], "binary_sign() (in module brevitas.function.ops)": [[7, "brevitas.function.ops.binary_sign"]], "binary_sign_ste() (in module brevitas.function.ops_ste)": [[7, "brevitas.function.ops_ste.binary_sign_ste"]], "brevitas.function.ops": [[7, "module-brevitas.function.ops"]], "brevitas.function.ops_ste": [[7, "module-brevitas.function.ops_ste"]], "brevitas.function.shape": [[7, "module-brevitas.function.shape"]], "ceil_ste() (in module brevitas.function.ops_ste)": [[7, "brevitas.function.ops_ste.ceil_ste"]], "dpu_round() (in module brevitas.function.ops)": [[7, "brevitas.function.ops.dpu_round"]], "dpu_round_ste() (in module brevitas.function.ops_ste)": [[7, "brevitas.function.ops_ste.dpu_round_ste"]], "floor_ste() (in module brevitas.function.ops_ste)": [[7, "brevitas.function.ops_ste.floor_ste"]], "get_upper_bound_on_l1_norm() (in module brevitas.function.ops)": [[7, "brevitas.function.ops.get_upper_bound_on_l1_norm"]], "identity() (in module brevitas.function.ops)": [[7, "brevitas.function.ops.identity"]], "max_float() (in module brevitas.function.ops)": [[7, "brevitas.function.ops.max_float"]], "max_int() (in module brevitas.function.ops)": [[7, "brevitas.function.ops.max_int"]], "min_int() (in module brevitas.function.ops)": [[7, "brevitas.function.ops.min_int"]], "over_batch_over_output_channels() (in module brevitas.function.shape)": [[7, "brevitas.function.shape.over_batch_over_output_channels"]], "over_batch_over_tensor() (in module brevitas.function.shape)": [[7, "brevitas.function.shape.over_batch_over_tensor"]], "over_output_channels() (in module brevitas.function.shape)": [[7, "brevitas.function.shape.over_output_channels"]], "over_output_features() (in module brevitas.function.shape)": [[7, "brevitas.function.shape.over_output_features"]], "over_tensor() (in module brevitas.function.shape)": [[7, "brevitas.function.shape.over_tensor"]], "round_ste() (in module brevitas.function.ops_ste)": [[7, "brevitas.function.ops_ste.round_ste"]], "round_to_zero() (in module brevitas.function.ops)": [[7, "brevitas.function.ops.round_to_zero"]], "round_to_zero_ste() (in module brevitas.function.ops_ste)": [[7, "brevitas.function.ops_ste.round_to_zero_ste"]], "scalar_clamp_min_ste() (in module brevitas.function.ops_ste)": [[7, "brevitas.function.ops_ste.scalar_clamp_min_ste"]], "scalar_clamp_ste() (in module brevitas.function.ops_ste)": [[7, "brevitas.function.ops_ste.scalar_clamp_ste"]], "tensor_clamp() (in module brevitas.function.ops)": [[7, "brevitas.function.ops.tensor_clamp"]], "tensor_clamp_() (in module brevitas.function.ops)": [[7, "brevitas.function.ops.tensor_clamp_"]], "tensor_clamp_ste() (in module brevitas.function.ops_ste)": [[7, "brevitas.function.ops_ste.tensor_clamp_ste"]], "tensor_clamp_ste_() (in module brevitas.function.ops_ste)": [[7, "brevitas.function.ops_ste.tensor_clamp_ste_"]], "ternary_sign_ste() (in module brevitas.function.ops_ste)": [[7, "brevitas.function.ops_ste.ternary_sign_ste"]], "absbinarysigngradfn (class in brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.AbsBinarySignGradFn"]], "binarysignstefn (class in brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.BinarySignSteFn"]], "ceilstefn (class in brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.CeilSteFn"]], "dpuroundstefn (class in brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.DPURoundSteFn"]], "floorstefn (class in brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.FloorSteFn"]], "inplacetensorclampstefn (class in brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.InplaceTensorClampSteFn"]], "roundstefn (class in brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.RoundSteFn"]], "roundtozerostefn (class in brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.RoundToZeroSteFn"]], "scalarclampminstefn (class in brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.ScalarClampMinSteFn"]], "scalarclampstefn (class in brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.ScalarClampSteFn"]], "tensorclampstefn (class in brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.TensorClampSteFn"]], "ternarysignstefn (class in brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.TernarySignSteFn"]], "abs_binary_sign_grad_impl() (in module brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.abs_binary_sign_grad_impl"]], "binary_sign_ste_impl() (in module brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.binary_sign_ste_impl"]], "brevitas.ops.autograd_ste_ops": [[8, "module-brevitas.ops.autograd_ste_ops"]], "ceil_ste_impl() (in module brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.ceil_ste_impl"]], "dpu_round_ste_impl() (in module brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.dpu_round_ste_impl"]], "floor_ste_impl() (in module brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.floor_ste_impl"]], "round_ste_impl() (in module brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.round_ste_impl"]], "round_to_zero_ste_impl() (in module brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.round_to_zero_ste_impl"]], "scalar_clamp_min_ste_impl() (in module brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.scalar_clamp_min_ste_impl"]], "scalar_clamp_ste_impl() (in module brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.scalar_clamp_ste_impl"]], "tensor_clamp_ste_impl() (in module brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.tensor_clamp_ste_impl"]], "ternary_sign_ste_impl() (in module brevitas.ops.autograd_ste_ops)": [[8, "brevitas.ops.autograd_ste_ops.ternary_sign_ste_impl"]]}}) \ No newline at end of file diff --git a/docs/settings.html b/docs/settings.html index 8c71cc998..6cde012b1 100644 --- a/docs/settings.html +++ b/docs/settings.html @@ -9,7 +9,7 @@ - Settings — Brevitas 0.10.1 documentation + Settings — Brevitas 0.10.2 documentation @@ -126,8 +126,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/setup.html b/docs/setup.html index 2597082d1..8fd8453b4 100644 --- a/docs/setup.html +++ b/docs/setup.html @@ -9,7 +9,7 @@ - Setup — Brevitas 0.10.1 documentation + Setup — Brevitas 0.10.2 documentation @@ -126,8 +126,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + @@ -418,8 +418,8 @@

Requirements

Installation Requirements#

diff --git a/docs/tutorials/anatomy_quantizer.html b/docs/tutorials/anatomy_quantizer.html index 0b17e219f..eb82ca551 100644 --- a/docs/tutorials/anatomy_quantizer.html +++ b/docs/tutorials/anatomy_quantizer.html @@ -9,7 +9,7 @@ - Anatomy of a Quantizer — Brevitas 0.10.1 documentation + Anatomy of a Quantizer — Brevitas 0.10.2 documentation @@ -129,8 +129,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/tutorials/index.html b/docs/tutorials/index.html index 2f5efbf1e..094cc0463 100644 --- a/docs/tutorials/index.html +++ b/docs/tutorials/index.html @@ -9,7 +9,7 @@ - Tutorials — Brevitas 0.10.1 documentation + Tutorials — Brevitas 0.10.2 documentation @@ -127,8 +127,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/tutorials/onnx_export.html b/docs/tutorials/onnx_export.html index 840933467..58010c658 100644 --- a/docs/tutorials/onnx_export.html +++ b/docs/tutorials/onnx_export.html @@ -9,7 +9,7 @@ - ONNX Export — Brevitas 0.10.1 documentation + ONNX Export — Brevitas 0.10.2 documentation @@ -129,8 +129,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/tutorials/quant_activation_overview.html b/docs/tutorials/quant_activation_overview.html index 1f85c472e..00f28f6e9 100644 --- a/docs/tutorials/quant_activation_overview.html +++ b/docs/tutorials/quant_activation_overview.html @@ -9,7 +9,7 @@ - An Overview of Quantized Activations — Brevitas 0.10.1 documentation + An Overview of Quantized Activations — Brevitas 0.10.2 documentation @@ -129,8 +129,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/tutorials/quant_recurrent.html b/docs/tutorials/quant_recurrent.html index 6563c3621..7c6d89e27 100644 --- a/docs/tutorials/quant_recurrent.html +++ b/docs/tutorials/quant_recurrent.html @@ -9,7 +9,7 @@ - Quantized RNNs and LSTMs — Brevitas 0.10.1 documentation + Quantized RNNs and LSTMs — Brevitas 0.10.2 documentation @@ -129,8 +129,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/tutorials/quant_tensor_quant_conv2d_overview.html b/docs/tutorials/quant_tensor_quant_conv2d_overview.html index cd138761d..5d004802a 100644 --- a/docs/tutorials/quant_tensor_quant_conv2d_overview.html +++ b/docs/tutorials/quant_tensor_quant_conv2d_overview.html @@ -9,7 +9,7 @@ - An overview of QuantTensor and QuantConv2d — Brevitas 0.10.1 documentation + An overview of QuantTensor and QuantConv2d — Brevitas 0.10.2 documentation @@ -129,8 +129,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/tutorials/tvmcon2021.html b/docs/tutorials/tvmcon2021.html index 609ff9116..dc1065045 100644 --- a/docs/tutorials/tvmcon2021.html +++ b/docs/tutorials/tvmcon2021.html @@ -9,7 +9,7 @@ - Brevitas TVMCon 2021 tutorial — Brevitas 0.10.1 documentation + Brevitas TVMCon 2021 tutorial — Brevitas 0.10.2 documentation @@ -129,8 +129,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/user_guide/datatypes.html b/docs/user_guide/datatypes.html index acef757ca..d634b371d 100644 --- a/docs/user_guide/datatypes.html +++ b/docs/user_guide/datatypes.html @@ -9,7 +9,7 @@ - <no title> — Brevitas 0.10.1 documentation + <no title> — Brevitas 0.10.2 documentation @@ -124,8 +124,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/user_guide/export.html b/docs/user_guide/export.html index 4f0da1a1d..b9e756003 100644 --- a/docs/user_guide/export.html +++ b/docs/user_guide/export.html @@ -9,7 +9,7 @@ - Export Compatibility — Brevitas 0.10.1 documentation + Export Compatibility — Brevitas 0.10.2 documentation @@ -124,8 +124,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/user_guide/graph_transformations.html b/docs/user_guide/graph_transformations.html index 997784741..388a67efd 100644 --- a/docs/user_guide/graph_transformations.html +++ b/docs/user_guide/graph_transformations.html @@ -9,7 +9,7 @@ - <no title> — Brevitas 0.10.1 documentation + <no title> — Brevitas 0.10.2 documentation @@ -124,8 +124,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/user_guide/index.html b/docs/user_guide/index.html index 8c6233ed6..656d7bbc6 100644 --- a/docs/user_guide/index.html +++ b/docs/user_guide/index.html @@ -9,7 +9,7 @@ - User Guide — Brevitas 0.10.1 documentation + User Guide — Brevitas 0.10.2 documentation @@ -124,8 +124,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/user_guide/precision.html b/docs/user_guide/precision.html index 08a695fcb..5aa07e1ed 100644 --- a/docs/user_guide/precision.html +++ b/docs/user_guide/precision.html @@ -9,7 +9,7 @@ - <no title> — Brevitas 0.10.1 documentation + <no title> — Brevitas 0.10.2 documentation @@ -124,8 +124,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/user_guide/proxies.html b/docs/user_guide/proxies.html index 1388af04f..3f1569d17 100644 --- a/docs/user_guide/proxies.html +++ b/docs/user_guide/proxies.html @@ -9,7 +9,7 @@ - <no title> — Brevitas 0.10.1 documentation + <no title> — Brevitas 0.10.2 documentation @@ -124,8 +124,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/user_guide/quantized_layers.html b/docs/user_guide/quantized_layers.html index 24786eb4c..a29656481 100644 --- a/docs/user_guide/quantized_layers.html +++ b/docs/user_guide/quantized_layers.html @@ -9,7 +9,7 @@ - <no title> — Brevitas 0.10.1 documentation + <no title> — Brevitas 0.10.2 documentation @@ -124,8 +124,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/user_guide/quantized_tensor.html b/docs/user_guide/quantized_tensor.html index 42e152e6c..968c360c0 100644 --- a/docs/user_guide/quantized_tensor.html +++ b/docs/user_guide/quantized_tensor.html @@ -9,7 +9,7 @@ - <no title> — Brevitas 0.10.1 documentation + <no title> — Brevitas 0.10.2 documentation @@ -124,8 +124,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/user_guide/quantizers.html b/docs/user_guide/quantizers.html index a8d7b187f..a6b38634c 100644 --- a/docs/user_guide/quantizers.html +++ b/docs/user_guide/quantizers.html @@ -9,7 +9,7 @@ - <no title> — Brevitas 0.10.1 documentation + <no title> — Brevitas 0.10.2 documentation @@ -124,8 +124,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/user_guide/scaling.html b/docs/user_guide/scaling.html index 0ec93ba8f..d78cfbbbd 100644 --- a/docs/user_guide/scaling.html +++ b/docs/user_guide/scaling.html @@ -9,7 +9,7 @@ - <no title> — Brevitas 0.10.1 documentation + <no title> — Brevitas 0.10.2 documentation @@ -124,8 +124,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home + diff --git a/docs/user_guide/zero_point.html b/docs/user_guide/zero_point.html index bd30d53e1..4d1e76fc8 100644 --- a/docs/user_guide/zero_point.html +++ b/docs/user_guide/zero_point.html @@ -9,7 +9,7 @@ - <no title> — Brevitas 0.10.1 documentation + <no title> — Brevitas 0.10.2 documentation @@ -124,8 +124,8 @@ - Brevitas 0.10.1 documentation - Home - + Brevitas 0.10.2 documentation - Home +