Draft
Conversation
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- `modal/ci.py`: Modal app that runs `run_training()` on an L4 GPU, checks val_loss against expected values. Tested locally — produces identical results to EC2 L4 (val_loss=0.364269). - `.github/workflows/gpu-e2e-modal.yml`: GHA workflow that calls `modal run` from `ubuntu-latest`. Requires `MODAL_TOKEN_ID` and `MODAL_TOKEN_SECRET` secrets. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- `modal/train.py`: Full training on Modal GPUs with `electrai-data` Volume (dataset_4, 2,885 samples). Supports GPU selection (L4/A100/H100), custom configs, WandB logging, checkpoint persistence. - `modal/populate_volume.py`: Sync S3 → Modal Volume via `boto3`. - Update spec with training docs, data provenance, secrets. Data pipeline: Globus (Della) → S3 → Modal Volume (complete). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Mirrors `gpu-benchmark.yml` but runs on Modal instead of EC2: - Uses `electrai-data` Volume (2,885 samples from dataset_4) - Configurable GPU type, model size, sample count, epochs - WandB logging with same tags/config as EC2 benchmark - Subsample support for quick benchmarks vs full dataset Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…ming - Fix `RhoData`: `strip()` + filter blank lines in filelist parsing (was treating trailing newlines as empty sample IDs → `.CHGCAR`) - Add GHA summary step to Modal benchmark workflow (config table + results, matching EC2 benchmark format) - WandB run names include dataset/samples/timestamp for local runs, GHA run number for CI runs - Print parseable `BENCHMARK_*` output for GHA step summary Co-Authored-By: Claude Opus 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
1 participant
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
Add Modal GPU infrastructure for CI, benchmarking, and training experiments.
CI (
modal/ci.py+.github/workflows/gpu-e2e-modal.yml)gpu-e2e.ymlval_loss=0.364269)Training (
modal/train.py)electrai-dataVolume withdataset_4(2,885 samples, ~205 GiB)electrai-checkpointsVolumeBenchmark (
modal/benchmark.py+.github/workflows/gpu-benchmark-modal.yml)gpu-benchmark.ymlbut runs on Modalmodal run modal/benchmark.py --gpu A100 --samples 50 --epochs 5modal-benchmarkbranch (50 samples, 5 epochs, L4)gpu-benchmark-modal.ymlusesworkflow_dispatchonly, so it won't be dispatchable until this PR merges tomainData pipeline
modal/populate_volume.py: S3 → Modal Volume syncs3://openathena/electrai/mp/chg_datasets/dataset_4/) → Modal VolumeImage construction
pyproject.tomlviapip_install_from_pyproject(no duplication)retries=0to prevent crash loops during iterationSecrets required
MODAL_TOKEN_ID/MODAL_TOKEN_SECRET— repo secrets (set)wandb-credentials— Modal secret withWANDB_API_KEY(set)aws-credentials— Modal secret forpopulate_volume.py(set, uses SSO session token)Test plan
modal run modal/ci.py— val_loss matcheslinux-gpuexpected valuesmodal run modal/train.py --epochs 2— trains on Volume data with WandBgpu-e2e-modal.ymltriggers on PR, passesmodal/populate_volume.py— 5,771 files synced to Volumemodal run modal/benchmark.py— 50 samples, 5 epochs, L4, greengpu-benchmark-modal.yml— push-triggered iteration, green