pytorch / pytorch/vision

Again details about how pretrained models are trained?

Open
#4,115 2 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

question
Dominant language
Python
Stars
17.9k
Forks
7.3k
Avg merge
1d 15h
Merged PRs (30d)
13

Description

  1. I use v0.5.0/references to train a resnet50 wirth defalut config. But, I got Best_val Top1=75.806%, which has a gap of 0.3% about the pretrained model. How can I to repreduce your accuracy ?
  2. I notice you said you use you recompute the batch norm statistics after training, can you show more details?

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start with v0.5.0/references/classification/train.py and compare its default ResNet-50 training configuration with the pretrained model's reported accuracy. Document the steps needed to reproduce the reported Top-1 result and explain how batch-normalization statistics are recomputed after training; the issue is done when both procedures are clear and reproducible.

Written by the indexing model from the issue text.

Assessment

Tech stack
python, pytorch
Domain
computer-vision, machine-learning
Issue type
Documentation
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
20/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.