Add CPU-only classifier pipeline and Windows-native setup guide#1
Open
ernesttyx wants to merge 1 commit into
Open
Add CPU-only classifier pipeline and Windows-native setup guide#1ernesttyx wants to merge 1 commit into
ernesttyx wants to merge 1 commit into
Conversation
- experiments/run_cpu.py: trains PHIStruct MLP using precomputed SaProt embeddings; no GPU, no Foldseek, no WSL required; works on Windows natively. Covers all four similarity bands (<=40/60/80/100%). Flags: --similarity, --output, --seed. - reproduce_on_colab.ipynb: end-to-end reproduction notebook for Google Colab (T4 GPU). Runs init.sh, downloads PDB structures, generates SaProt embeddings from scratch, trains classifier, compares output to committed pickles. - README.md: new section documenting the CPU/Windows path with download links. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
experiments/run_cpu.py- standalone script that trains and evaluates the PHIStruct MLP classifier using the team's precomputed SaProt embeddings, with no GPU, no Foldseek, and no WSL required. Works on Windows natively with justpip install pandas scikit-learn imbalanced-learn. Covers all four similarity bands (40/60/80/100%) and supports--similarity,--output, and--seedflags.reproduce_on_colab.ipynb- end-to-end independent reproduction notebook for Google Colab (T4 GPU). Runsinit.sh, downloads the team's ColabFold-predicted PDB structures, generates SaProt embeddings from scratch via Foldseek + SaProt, trains the classifier, and compares the freshly generated macro F1 against the committed pickle outputs as an independent reproducibility check.README.md- new section "Running Without a GPU (Windows Native / Any CPU)" documenting therun_cpu.pypath with exact Google Drive download links and usage examples.Motivation
I tried to run the full pipeline on Windows (no GPU) and hit the wall described in the README:
init.shrequires a Linux/macOS Foldseek binary and a CUDA PyTorch wheel. The precomputed embeddings path (Option 1 in notebook 5.0) works perfectly on any OS with just scikit-learn, but it is buried in the notebook prerequisites and requires reading through several notebooks to piece together. This PR makes that path a first-class, single-command entry point.The Colab notebook is a separate contribution: an independent reproduction of the full pipeline (embeddings from scratch, classifier training, compare to committed outputs) on a fresh Linux+GPU environment.
Test plan
run_cpu.pytested on Windows 11 (no GPU) with the two prerequisite downloadsreproduce_on_colab.ipynbdesigned for Colab T4 GPU; currently running