(prototype) Delete the four old SimCLR copies - #2039
Conversation
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 39228a7bb5
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| > SimCLR has moved to [benchmarks/simclr/](../simclr), which carries one row per | ||
| > dataset and is gated against `examples/simclr.py`. The rest of this tree follows | ||
| > as each method is ported. |
There was a problem hiding this comment.
Remove SimCLR from the old benchmark inventory
After moving SimCLR out of this tree, the Available Benchmarks table still advertises it under resnet50/. Users following that inventory and passing --methods simclr to this README's resnet50/main.py command will get a KeyError, because the SimCLR file was deleted and METHODS no longer contains that key. Update the inventory alongside this move.
Useful? React with 👍 / 👎.
7 of 7 in a stack. Base: #2038.
Removes the four SimCLR implementations that PR 6 replaces:
plus their generated notebooks and the
METHODSentry in the ResNet-50 runner.dcl.pyanddclw.pyimportSimCLRTransformdirectly rather than the deleted module, so they keep working.docs/source/examples/simclr.rstloses its three tabs for oneliteralincludeof the new example, and gains a section on reproducing a published number.The README's 63.2 linear and 73.9 kNN came from the file being deleted here, on the run logged as
imagenet_resnet50_simclr_2023-06-22_09-11-13. The number stays in the table with a footnote saying which file produced it and that the new benchmark has not reproduced it.benchmarks/simclr/README.mdlists the three things that differ: batch size and learning-rate scaling, the online probe's inputs and optimiser, and the kNN probe's lifecycle.Testing:
make generate-example-notebooksproduces no diff, and nothing indocs/orREADME.mdstill points at a deleted path.