[ExecuTorch][llm] Wire --use_moe_quantized_op through export pipeline and LLM config#21126
[ExecuTorch][llm] Wire --use_moe_quantized_op through export pipeline and LLM config#21126digantdesai wants to merge 4 commits into
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/21126
Note: Links to docs will display an error until the docs builds have been completed. ❌ 1 New Failure, 1 Unrelated Failure, 1 Unclassified FailureAs of commit 91e8c0a with merge base bfed808 ( NEW FAILURE - The following job has failed:
UNCLASSIFIED FAILURE - DrCI could not classify the following job because the workflow did not run on the merge base. The failure may be pre-existing on trunk or introduced by this PR:
BROKEN TRUNK - The following job failed but were present on the merge base:👉 Rebase onto the `viable/strict` branch to avoid these failures
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
This PR needs a
|
… and LLM config Pull Request resolved: #21126 Add `--use_moe_quantized_op` to the argparse and LLM configuration paths, thread it through `_get_source_transforms`. The transform replaces eager MoE modules once through the common export pipeline and preserves shared-expert behavior. ghstack-source-id: 405736786 @exported-using-ghexport Differential Revision: [D102381996](https://our.internmc.facebook.com/intern/diff/D102381996/)
… and LLM config Pull Request resolved: #21126 Add `--use_moe_quantized_op` to the argparse and LLM configuration paths, thread it through `_get_source_transforms`. The transform replaces eager MoE modules once through the common export pipeline and preserves shared-expert behavior. ghstack-source-id: 405939022 @exported-using-ghexport Differential Revision: [D102381996](https://our.internmc.facebook.com/intern/diff/D102381996/)
… and LLM config Pull Request resolved: #21126 Add `--use_moe_quantized_op` to the argparse and LLM configuration paths, thread it through `_get_source_transforms`. The transform replaces eager MoE modules once through the common export pipeline and preserves shared-expert behavior. ghstack-source-id: 406251209 @exported-using-ghexport Differential Revision: [D102381996](https://our.internmc.facebook.com/intern/diff/D102381996/)
Stack from ghstack (oldest at bottom):
Add
--use_moe_quantized_opto the argparse and LLM configuration paths, thread it through_get_source_transforms. The transform replaces eager MoE modules once through the common export pipeline and preserves shared-expert behavior.Differential Revision: D102381996