DetaMetricsSDK

A Python SDK for interacting with my TensorBoard alternative that runs on Deta Space.

• DetaMetricsSDK
  • Installation
  • Usage
    • TensorFlow Callback
    • Base API

---

Installation

pip install detametrics

Usage

• Install in Deta Space from https://deta.space/discovery/@sam-the-programmer/detametrics

• All API classes need an URL ID passed as a parameter, which can be found in the app UI.

• They also need a Deta Space App API key, which can be gotten from...

  • Click the 3 dots on the app on your Deta Space Canvas
  • Click "Keys"
  • Add an API Key, and input it in the box on the App UI and in your Python code as seen below

TensorFlow Callback

Requires Tensorflow to be installed.

from detametrics.tf import DetaMetricsTFCallback

# use in your keras model
model = ...
model.fit(
  ..., # other params here
  callbacks=[DetaMetricsTFCallback("MY_URL_ID", "MY_API_KEY", log_batch=True)] # if log_batch is True, it logs every batch, else only every epoch
)

Base API

Get your URL id from the UI on the Deta Space app.

from detametrics import DetaMetrics

metrics = DetaMetrics("MY_URL_ID", "MY_API_KEY")
metrics.clear() # WARNING: clears all past metrics

metrics.set("GraphName", "LineName", 0.1)
metrics.set("GraphName", "LineName", 123.45)
metrics.set("GraphName", "LineName", 78.9)