microsoft/nni

RuntimeError: Has not supported replacing the module: `PixelShuffle`

Open

#4,301 opened on Nov 10, 2021

View on GitHub
 (4 comments) (0 reactions) (1 assignee)Python (13,504 stars) (1,830 forks)batch import
enhancementexternal contributorgood first issue

Description

Describe the issue:

Environment:

  • NNI version:Release 2.3 (github master branch)
  • Training service (local|remote|pai|aml|etc):local
  • Client OS:Ubuntu 18.04
  • Server OS (for remote mode only):
  • Python version: Python 3.7.0
  • PyTorch/TensorFlow version: 1.7.1
  • Is conda/virtualenv/venv used?: virutalenv
  • Is running in Docker?: no

For super resolution models, nn.PixelShuffle(scale) is widely used. When I tried to compress and speed up a super resolution model with nn.PixelShuffle( ), an error was met: RuntimeError: Has not supported replacing the module: PixelShuffle . Is there a way that I can fix or work around it. Thanks very much!

Contributor guide

Tech stack
pythonpytorch
Domain
machine learning
Issue type
feature
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
3
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
1-3 hours
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
blocked
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
clear
Prerequisites
Familiarity with PyTorchUnderstanding of NN model compressionBasic knowledge of NNI
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
40
Research direction
Investigate how NNI currently handles module replacement for similar operations (e.g., Conv2d, BatchNorm). Look at the implementation in NNI's model compression module, likely in `nni/compression/pytorch/` directory. The error indicates that `PixelShuffle` is not in the supported list of modules. You would need to add support by registering `PixelShuffle` as a replaceable module, similar to other modules. Check existing test cases for module replacement to ensure consistency.