pytorch
67115b22 - [quant][graphmode] Dynamic Quant Do not depend on input shapes (#39412)

Commit

4 years ago

[quant][graphmode] Dynamic Quant Do not depend on input shapes (#39412) Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/39412 This PR introduces changes to enable running the weight observer standalone in the graph It extracts the nodes from the graph that correspond to the observed weight value and adds all the related nodes to a new subgraph The subgraph is then executed using GraphFunction Test Plan: python test/test_quantization.py TestGraphMostPostTrainingStatic python test/test_quantization.py TestQuantizeDynamicScript Imported from OSS Differential Revision: D21872940 fbshipit-source-id: 55f1dcc2caef193531e2b807c8e56288b9794520

Author

supriyar

Committer

facebook-github-bot

Parents

6d13b583

Files6

test/quantization
- test_quantize.py
- test_quantize_script.py
torch
- csrc/jit/passes/quantization
  - helper.h
  - insert_quant_dequant.cpp
- quantization
  - observer.py
  - quantize_script.py

pytorch 67115b22 - [quant][graphmode] Dynamic Quant Do not depend on input shapes (#39412)

pytorch
67115b22 - [quant][graphmode] Dynamic Quant Do not depend on input shapes (#39412)