Add HydraGNN AMD inference scripts and fix overlay build pattern#31
Merged
Conversation
- Add build_overlay_amd.sh: builds ext3 overlay on \$TMPDIR (node-local disk) then copies the finished image to NFS as one sequential write, avoiding the ext3-on-NFS per-file FUSE overhead. Uses Looong01/pyg-rocm- build release 15 wheels (torch-scatter, torch-sparse, etc.) for ROCm 7.1+. - Add sbatch_infer_amd.sh: SLURM job script for inference with Apptainer. - Fix run_inference.sh: GFM checkpoints are saved with DDP (module.* keys) but inference runs without DDP wrapping — strip the module. prefix rather than adding it. Removes Docker-era cd /workspace/HydraGNN and replaces hydragnn.run_prediction() (requires Dataset config key absent from GFM configs) with direct create_model_config + torch.load + load_state_dict. - Apply same \$TMPDIR overlay pattern to ORBIT-2/examples/build_overlay_amd.sh. - Propagate overlay and inference lessons to Cursor and Claude skills. Validated end-to-end on AMD MI355X (ROCm 7.2.2, torch 2.10) with mlupopa/HydraGNN_Predictive_GFM_2024 gfm_0.229 checkpoint. Co-Authored-By: Claude Sonnet 4 <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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
copies finished image to NFS as one sequential write — avoids ext3-on-NFS
FUSE per-file overhead. Uses Looong01/pyg-rocm-build r15 wheels for ROCm 7.1+.
inference runs without DDP — strip the prefix, don't add it. Remove Docker-era
cd /workspace/HydraGNN. Replace hydragnn.run_prediction() (requires Dataset
config key absent from GFM configs) with direct create_model_config + torch.load.
Validated on AMD MI355X (ROCm 7.2.2, torch 2.10, lux partition)