Releases: IBM/aihwkit
Releases · IBM/aihwkit
IBM Analog Hardware Accelerator Kit 0.9.2
[0.9.2] - 2024/09/18
Added
- Added new Hermes noise model and related notebooks (#685)
- Added new conductance converters (#685)
- Make Conv layers also compatible with non-batched inputs (#685)
- Added per column drift compensation (#685)
- Added custom drifts (#685)
Changed
- Update requirements-examples.txt (#685)
IBM Analog Hardware Accelerator Kit 0.9.1
[0.9.1] - 2024/05/16
Added
- Added column wise scaling logic to fusion import/export to improve accuracy (#652)
- Added a new example that demonstrates how to import and perform inference using a model which has been trained in a hardware-aware fashion using an external library (#648)
- Added a wew WeightModifierType.DISCRETIZE_PER_CHANNEL type and a test case to validate the correctness against manual quantization in PyTorch (#618)
Fixed
- Fix sub-optimal mapping of conductances to weights for fusion by regressing weights per column (#653)
- Documentation correction: Use
ADDITIVE_CONSTANT instead of
ADD_NORMAL` in WeightNoiseType (#630) - Fix continuing training based on checkpoint using torch tile (#626)
- Fix the support of different dtypes for the torch model (#625)
- Fixes the fall-through to the default error message when using drop connect (#624)
- Update
analog_fusion
notebook (#611)
IBM Analog Hardware Accelerator Kit 0.9.0
[0.9.0] - 2024/01/26
Added
- On-the-fly change of some RPUConfig fields (#539)
- Fusion chip CSV file model weights exporter functionality (#538)
- Experimental support for RPU data types (#563)
- Optional AIHWKIT C++ extension module (#563)
- Variable mantissa / exponent tensor conversion operator (#563)
- To digital feature for analog layers (#563)
- New PCM_NOISE type for hardware-aware training for inference (#563)
- Transfer compounds using torch implementation (TorchTransferTile) (#567)
- Weight programming error plotting utility (#572)
- Add optimizer checkpoint in example 20 (#573)
- Inference tile with time-dependent IR-drop (#587)
- Linear algebra module (#588)
- New Jupyter notebook for Fusion chip access (#601)
Fixed
- Repeated call of cuda() reset the weights for InferenceTile (#540)
- Custom tile bugfixes (#563)
- Bug-fixes for specialized learning algorithms (#563)
- Bug-fix for data-parallel hardware-aware training for inference (#569)
- Fix docker build stubgen (#581)
- Fix readthedoc builds (#586)
- Fix the backward of the input ranges in the torch tile (#606)
Changed
- Parameter structure changed into separate files to reduce file sizes (#563)
- RPUConfig has a new runtime field and inherits from additional base classes (#563)
- AnalogWrapper now directly adds module classes to subclasses (#563)
- RNN linear layers more custonable (#563)
- Parameters for specialized learning algorithms changed somwhat (#563)
- RNN modules inherit from Module or AnalogContainerBase instead of AnalogSequential (#563)
- Adjustment of parameter to bindings for various number formats (#563)
- Documentation updates and fixes (#562, #564, #570, #575, #576, #580, #585, #586)
- Updated installation instructions in Readthedoc (#594)
IBM Analog Hardware Acceleration Kit 0.8.0
[0.8.0] - 2023/07/14
Added
- Added new tutorial notebooks to cover the concepts of training,
hardware-aware training, post-training calibration, and extending aihwkit functionality (#518, #523, #526) - Calibration of input ranges for inference (#512)
- New analog in-memory training algorithms: Chopped Tiki-taka II (#512)
- New analog in-memory training algorithms: AGAD (#512)
- New training presets:
ReRamArrayOMPresetDevice
,
ReRamArrayHfO2PresetDevice
,ChoppedTTv2*
,AGAD*
(#512) - New correlation detection example for comparing specialized analog SGD
algorithms (#512) - Simplified
build_rpu_config
script for generatingRPUConfigs
for
analog in-memory SGD (#512) CustomTile
for customization of in-memory training algorithms (#512)- Pulse counters for pulsed analog training (#512)
TorchInferenceTile
for a fully torch-based analog tile for
inference (not using the C++ RPUCuda engine), supporting a subset of MVM nonidealities (#512)- New inference preset
StandardHWATrainingPreset
(#512) - New inference noise model
ReRamWan2022NoiseModel
(#512) - Improved HWA-training for inference featuring input and output range
learning and more (#512) - Improved CUDA memory management (using torch cached GPU memory for
internal RPUCuda buffer) (#512) - New layer generator:
analog_layers()
loops over layer modules (except
container) (#512) AnalogWrapper
for wrapping a full torch module (Without using
AnalogSequential
) (#512)convert_to_digital
utility (# 512)TileModuleArray
for logical weight matrices larger than a single tile. (#512)- Dumping of all C++ fields for accurate analog training saving and
training continuation after checkpoint load. (#512) apply_write_noise_on_set
for pulsed devices. (#512)- Reset device now also for simple devices. (#512)
SoftBoundsReference
,PowStepReference
for explicit reference
subtraction of symmetry point in Tiki-taka (#512)- Analog MVM with output-to-output std-deviation variability
(output_noise_std
) (#512) - Plotting utility for weight errors (#512)
per_batch_sample
weight noise injections forTorchInferenceRPUConfig
(#512)
Fixed
- BERT example 24 using
AnalogWrapper
(#514) - Cuda supported testing in examples based on AIHWKIT compilation (#513)
- Fixed compilation error for CUDA 12.1. (#500)
- Realistic read weights could have applied the scales wrongly (#512)
Changed
- Major re-organization of
AnalogTiles
for increased modularity
(TileWithPeriphery
,SimulatorTile
,SimulatorTileWrapper
). Analog
tile modules (possibly array of analog tiles) are now also torchModule
. (#512) - Change in tile generators:
analog_model.analog_tiles()
now loops over
all available tiles (in all modules) (#512) - Import and file position changes. However, user can still import
RPUConfig
related modules fromaihwkit.simulator.config
(#512) convert_to_analog
now also considered mapping. Set
mapping.max_out_size = 0
andmapping.max_out_size = 0
to avoid this. (#512)- Mapped layers now use
TileModuleArray
array by default. (#512) - Checkpoint structure is different than previous
versions.utils.legacy_load
provides a way to load old checkpoints. (#512)
Removed
realistic_read_write
is removed from some high-level function. Use
program_weights
(after setting the weights) orread_weights
for realistic reading (using weight estimation technique). (#512)
IBM Analog Hardware Acceleration Kit 0.7.1
Added
- Updated the CLI Cloud runner code to support inference experiment result. (#491)
- Read weights is done with least-square estimation method. (#489)
Fixed
- Realistic read / write behavior was broken for some tiles. (#489)
Changed
- Torch minimal version has changed to version 1.9. (#489)
- Realistic read / write is now achieved by read_weights and program_weights. (#489)
Removed
- The tile methods get/set_weights_realistic are removed. (#489)
IBM Analog Hardware Acceleration Kit 0.7.0
[0.7.0] - 2023/01/30
Added
- Reset tiles method (#456)
- Added many new analog MAC non-linearties (forward / backward pass) (#456)
- Polynomial weight noise for hardware-aware training (#456)
- Remap functionality for hardware-aware training (#456)
- Input range estimation for InferenceRPUConfig (#456)
- CUDA always syncs and added non-blocking option if not wished (#456)
- Fitting utility for fitting any device model to conductance measurements (#456)
- Added
PowStepReferenceDevice
for easy subtraction of symmetry
point (#456) - Added
SoftBoundsReferenceDevice
for easy subtraction of symmetry
point (#456) - Added stand-alone functions for applying inference drift to any model (#419)
- Added Example 24: analog inference and hardware-aware training on BERT with the SQUAD task (#440)
- Added Example 23: how to use
AnalogTile
directly to implement an
analog matrix-vector product without using pytorch modules. (#393) - Added Example 22: 2 layer LSTM network trained on War and Peace dataset. (#391)
- Added a new notebook for exploring analog sensitivities. (#380)
- Remapping functionality for
InferenceRPUConfig
. (#388) - Inference cloud experiment and runners. (#410)
- Added
analog_modules
generator inAnalogSequential
. (#410) - Added
SKIP_CUDA_TESTS
to manually switch off the CUDA tests. - Enabling comparisons of
RPUConfig
instances. (#410) - Specific user-defined function for layer-wise setting for RPUConfigs
in conversions. (#412) - Added stochastic rounding options for
MixedPrecisionCompound
. (#418) - New
remap
parameter field and functionality in
InferenceRPUConfig
(#423). - Tile-level weight getter and setter have
apply_weight_scaling
argument. (#423) - Pre and post-update / backward / forward methods in
BaseTile
for
easier user-defined modification of pre and/or post-processings of a tile. (#423) - Type-checking for
RPUConfig
fields. (#424)
Fixed
- Decay fix for compound devices (#463)
RPUCuda
backend update with many fixes (#456)- Missing zero-grad call in example 02 (#446)
- Indexing error in
OneSidedDevice
for CPU (#447) - Analog summary error when model is on cuda device. (#392)
- Index error when loading the state dict with a model use previously. (#387)
- Weights that were not contiguous could have been set wrongly. (#388)
- Programming noise would not be applied if drift compensation was not
used. (#389) - Loading a new model state dict for inference does not overwrite the noise
model setting. (#410) - Avoid
AnalogContext
copying of self pointers. (#410) - Fix issue that drift compensation is not applied to conv-layers. (#412)
- Fix issue that noise modifiers are not applied to conv-layers. (#412)
- The CPU
AnalogConv2d
layer now uses unfolded convolutions instead of
indexed covolutions (that are efficient only for GPUs). (#415) - Fix issue that write noise hidden weights are not transferred to
pytorch when usingget_hidden_parameters
in case of CUDA. (#417) - Learning rate scaling due to output scales. (#423)
WeightModifiers
of theInferenceRPUConfig
are no longer called
in the forward pass, but instead in thepost_update_step
method to avoid issues with repeated forward calls. (#423)- Fix training
learn_out_scales
issue after checkpoint load. (#434)
Changed
- Pylint / mypy / pycodestyle / protobuf version bump (#456)
- All configs related classes can now be imported from
aihwkit.simulator.config
(#456) - Weight noise visualization now shows the programming noise and drift
noise differences. (#389) - Concatenate the gradients before applying to the tile update
function (some speedup for CUDA expected). (#390) - Drift compensation uses eye instead of ones for readout. (#412)
weight_scaling_omega_columnwise
parameter inMappingParameter
is now called
weight_scaling_columnwise
. (#423)- Tile-level weight getter and setter now use Tensors instead of numpy
arrays. (#423) - Output scaling and mapping scales are now distiniguished, only the
former is learnable. (#423) - Renamed
learn_out_scaling_alpha
parameter inMappingParameter
to
learn_out_scaling
and columnwise learning has a separate switch
out_scaling_columnwise
. (#423)
Deprecated
- Input
weight_scaling_omega
argument in analog layers is deprecated. (#423)
Removed
- The
_scaled
versions of the weight getter and setter methods are
removed. (#423)
IBM Analog Hardware Acceleration Kit 0.6.0
[0.6.0] - 2022/05/16
Added
- Set weights can be used to re-apply the weight scaling omega. (#360)
- Out scaling factors can be learnt even if weight scaling omega was set to 0. (#360)
- Reverse up / down option for LinearStepDevice. (#361)
- Generic Analog RNN classes (LSTM, RNN, GRU) uni or bidirectional. (#358)
- Added new PiecewiseStepDevice where the update-step response function can be arbitrarily defined by the user in a piece-wise linear manner. It can be conveniently used to fit any experimental device data. (#356)
- Several enhancements to the public documentations: added a new section for hw-aware training, refreshed the reference API doc, and added the newly supported LSTM layers and the mapped conv layers. (#374)
Fixed
- Legacy checkpoint load with alpha scaling. (#360)
- Re-application of weight scaling omega when loading checkpoints. (#360)
- Write noise was not correctly applied for CUDA if dw_min_std=0. (#356)
Changed
IBM Analog Hardware Acceleration Kit 0.5.1
Added
- Load model state dict into a new model with modified
RPUConfig
. (#276) - Visualization for noise models for analog inference hardware simulation. (#278)
- State independent inference noise model. (# 284)
- Transfer LR parameter for
MixedPrecisionCompound
. (#283) - The bias term can now be handled either by the analog or digital domain by controlling the
digital_bias
layer parameter. (#307) - PCM short-term weight noise. (#312)
- IR-drop simulation across columns during analog mat-vec. (#312)
- Transposed-read for
TransferCompound
. (#312) BufferedTranferCompound
and TTv2 presets. (#318)- Stochastic rounding for
MixedPrecisionCompound
. (#318) - Decay with arbitrary decay point (to reset bias). (#319)
- Linear layer
AnalogLinearMapped
which maps a large weight matrix onto multiple analog tiles. (#302) - Convolution layers
AnalogConvNdMapped
which maps large weight matrix onto multiple tiles if necessary. (#331) - In the new mapping field of
RPUConfig
the max tile input and output sizes can be configured for the*Mapped
layers. (#331) - Notebooks directory with several notebook examples (#333, #334)
- Analog information summary function. (#302)
- The alpha weight scaling factor can now be defined as learnable parameter by switching
learn_out_scaling_alpha
in therpu_config.mapping
parameters. (#353)
Fixed
- Removed GPU warning during destruction when using multiple GPUs. (#277)
- Fixed issue in transfer counter for mixed precision in case of GPU. (#283)
- Map location keyword for load / save observed. (#293)
- Fixed issue with CUDA buffer allocation when batch size changed. (#294)
- Fixed missing load statedict for
AnalogSequential
. (#295) - Fixed issue with hierarchical hidden parameter settings. (#313)
- Fixed serious issue that loaded model would not update analog gradients. (#302)
- Fixed cuda import in examples. (#320)
Changed
- The inference noise models are now located in
aihwkit.inference
. (#281) - Analog state dict structure `has changed (shared weight are not saved). (#293)
- Some of the parameter names of the
TransferCompound
have changed. (#312) - New fast learning rate parameter for
TransferCompound
, SGD learning rate then is applied on the slow matrix (#312). - The fixed_value of
WeightClipParameter
is now applied for all clipping types if set larger than zero. (#318) - The use of generators for analog tiles of an
AnalogModuleBase
. (#302) - Digital bias is now accessible through
MappingParameter
. (#331) - The aihwkit documentation. New content around analog AI concepts, training presets, analog AI optimizers, new references, and examples. (#348)
- The
weight_scaling_omega
can now be defined in therpu_config.mapping
. (#353)
Deprecated
- The module
aihwkit.simulator.noise_models
has been depreciated in favor ofaihwkit.inference
. (#281)
IBM Analog Hardware Acceleration Kit 0.4.0
Added
- A number of new config presets added to the library, namely
EcRamMOPreset
,
EcRamMO2Preset
,EcRamMO4Preset
,TikiTakaEcRamMOPreset
,
MixedPrecisionEcRamMOPreset
. These can be used for tile configuration
(rpu_config
). They specify a particular device and optimizer choice. (#207) - Weight refresh mechanism for
OneSidedUnitCell
to counteract saturation, by
differential read, reset, and re-write. (#209) - Complex cycle-to-cycle noise for
ExpStepDevice
. (#226) - Added the following presets:
PCMPresetDevice
(uni-directional),
PCMPresetUnitCell
(a pair of uni-directional devices with periodical
refresh) and aMixedPrecisionPCMPreset
for using the mixed precision
optimizer with a PCM pair. (#226) AnalogLinear
layer now accepts multi-dimensional inputs in the same
way as PyTorch'sLinear
layer does. (#227)- A new
AnalogLSTM
module: a recurrent neural network that uses
AnalogLinear
. (#240) - Return of weight gradients for
InferenceTile
(only), so that the gradient
can be handled with any PyTorch optimizer. (#241) - Added a generic analog optimizer
AnalogOptimizer
that allows extending
any existing optimizer with analog-specific features. (#242) - Conversion tools for converting torch models into a model having analog
layers. (#265)
Changed
- Renamed the
DifferenceUnitCell
toOneSidedUnitCell
which more properly
reflects its function. (#209) - The
BaseTile
subclass that is instantiated in the analog layers is now
retrieved from the newRPUConfig.tile_class
attribute, facilitating the
use of custom tiles. (#218) - The default parameter for the
dataset
constructor used byBasicTraining
is now thetrain=bool
argument. If using a dataset that requires other
arguments or transforms, they can now be specified via overriding
get_dataset_arguments()
andget_dataset_transform()
. (#225) AnalogContext
is introduced, along with tile registration function to
handle arbitrary optimizers, so that re-grouping param groups becomes
unnecessary. (#241)- The
AnalogSGD
optimizer is now implemented based on the generic analog
optimizer, and its base module isaihwkit.optim.analog_optimizer
. (#242) - The default refresh rate is changed to once per mini-batch for
PCMPreset
(as opposed to once per mat-vec). (#243)
Deprecated
- Deprecated the
CudaAnalogTile
andCudaInferenceTile
and
CudaFloatingPointTile
. Now theAnalogTile
can be either on cuda or on cpu
(determined by thetile
and thedevice
attribute) similar to a torch
Tensor
. In particular, call ofcuda()
does not change theAnalogTile
to
CudaAnalogTile
anymore, but only changes the instance in thetile
field,
which makes in-place calls tocuda()
possible. (#257)
Removed
- Removed
weight
andbias
of analog layers from the module parameters as
these parameters are handled internally for analog tiles. (#241)
Fixed
- Fixed autograd functionality for recurrent neural networks. (#240)
- N-D support for
AnalogLinear
. (#227) - Fixed an issue in the
Experiments
that was causing the epoch training loss
to be higher than the epoch validation loss. (#238) - Fixed "Wrong device ordinal" errors for CUDA which resulted from a known
issue of using CUB together with pytorch. (#250) - Renamed persistent weight hidden parameter field to
persistent_weights
.
(#251) - Analog tiles now always move correctly to CUDA when
model.cuda()
ormodel.to(device)
is used. (#252, #257) - Added an error message when wrong tile class is used for loading an analog
state dict. (#262) - Fixed
MixedPrecisionCompound
being bypassed with floating point compute.
(#263)
IBM Analog Hardware Acceleration Kit 0.3.0
Added
- New analog devices:
- A new abstract device (
MixedPrecisionCompound
) implementing an SGD
optimizer that computes the rank update in digital (assuming digital
high precision storage) and then transfers the matrix sequentially to
the analog device, instead of using the default fully parallel pulsed
update. (#159) - A new device model class
PowStepDevice
that implements a power-exponent
type of non-linearity based on the Fusi & Abott synapse model. (#192) - New parameterization of the
SoftBoundsDevice
, called
SoftBoundsPmaxDevice
. (#191)
- A new abstract device (
- Analog devices and tiles improvements:
- Option to choose deterministic pulse trains for the rank-1 update of
analog devices during training. (#99) - More noise types for hardware-aware training for inference
(polynomial). (#99) - Additional bound management schemes (worst case, average max, shift).
(#99) - Cycle-to-cycle output referred analog multiply-and-accumulate weight
noise that resembles the conductance dependent PCM read noise
statistics. (#99) - C++ backend improvements (slice backward/forward/update, direct
update). (#99) - Option to excluded bias row for hardware-aware training noise. (#99)
- Option to automatically scale the digital weights into the full range of
the simulated crossbar by applying a fixed output global factor in
digital. (#129) - Optional power-law drift during analog training. (#158)
- Cleaner setting of
dw_min
using device granularity. (#200)
- Option to choose deterministic pulse trains for the rank-1 update of
- PyTorch interface improvements:
- New modules added:
- A library of device presets that are calibrated to real hardware data,
namelyReRamESPresetDevice
,ReRamSBPresetDevice
,ECRamPresetDevice
,
CapacitorPresetDevice
, and device presets that are based on models in the
literature, e.g.GokmenVlasovPresetDevice
andIdealizedPresetDevice
.
They can be used defining the device field in theRPUConfig
. (#144) - A library of config presets, such as
ReRamESPreset
,Capacitor2Preset
,
TikiTakaReRamESPreset
, and many more. These can be used for tile
configuration (rpu_config
). They specify a particular device and optimizer
choice. (#144) - Utilities for visualization the pulse response properties of a given
device configuration. (#146) - A new
aihwkit.experiments
module has been added that allows creating and
running specific high-level use cases (for example, neural network training)
conveniently. (#171, #172) - A
CloudRunner
class has been added that allows executing experiments in
the cloud. (#184)
- A library of device presets that are calibrated to real hardware data,
Changed
- The minimal PyTorch version has been bumped to
1.7+
. Please recompile your
library and update the dependencies accordingly. (#176) - Default value for TransferCompound for
transfer_every=0
(#174).
Fixed
- Issue of number of loop estimations for realistic reads. (#192)
- Fixed small issues that resulted in warnings for windows compilation. (#99)
- Faulty backward noise management error message removed for perfect backward
and CUDA. (#99) - Fixed segfault when using diffusion or reset with vector unit cells for
CUDA. (#129) - Fixed random states mismatch in IoManager that could cause crashed in same
network size and batch size cases for CUDA, in particular for
TransferCompound
. (#132) - Fixed wrong update for
TransferCompound
in case oftransfer_every
smaller
than the batch size. (#132, #174) - Period in the modulus of
TransferCompound
could become zero which
caused a floating point exception. (#174) - Ceil instead of round for very small transfers in
TransferCompound
(to avoid zero transfer for extreme settings). (#174)
Removed
- The legacy
NumpyAnalogTile
andNumpyFloatingPointTile
tiles have been
finally removed. The regular, tensor-poweredaihwkit.simulator.tiles
tiles
contain all their functionality and numerous additions. (#122)