Skip to main content

Tracking and Visualize after the burning PyTorch

Project description

Torch Ember

Tracking and visualize after the burning pytorch

This framework tracks the pytorch model:

  • On nn.Module level
  • Down to the metrics/ features of all tensors, includes
    • inputs/outputs of each module
    • weight/grad tensors
  • By minimal extra coding

WebUI

Other lovely features

  • Customizable metrics, with easy decorator syntax
  • Split the tracking log in the way you like, just mark(k=v,k1=v2...)
  • You can easily switch on/off the tracking:
    • Even cost of computation is tiny, torchember don't have to calculate metric for every iteration
    • Hence, you can track eg. only the last steps, only each 200 steps .etc

Installation

pip install torchember

Fast Tutorial

Step1, Track your model

Place you torch ember tracker on your model

from torchember.core import torchEmber
te = torchEmber(model)

The above can track input and output of every module,The following can track status of every module

for i in range(1000):
    ...
    loss.backward()
    optimizer.step()

    te.log_model()

Train your model as usual

Step2, Check the analysis on the WebUI

Run the service from terminal

$ torchember

The default port will be 8080

Or assign a port

$ torchember --port=4200

Visit your analysis at http://[host]:[port]

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for torchember, version 0.2.7
Filename, size File type Python version Upload date Hashes
Filename, size torchember-0.2.7-py3-none-any.whl (1.9 MB) File type Wheel Python version py3 Upload date Hashes View
Filename, size torchember-0.2.7.tar.gz (1.8 MB) File type Source Python version None Upload date Hashes View

Supported by

AWS AWS Cloud computing Datadog Datadog Monitoring DigiCert DigiCert EV certificate Facebook / Instagram Facebook / Instagram PSF Sponsor Fastly Fastly CDN Google Google Object Storage and Download Analytics Pingdom Pingdom Monitoring Salesforce Salesforce PSF Sponsor Sentry Sentry Error logging StatusPage StatusPage Status page