-
Notifications
You must be signed in to change notification settings - Fork 10
refactor(torchtitan): rollback Titan to 99c0cb2(20250907) and stabilize trainer UTs #262
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Merged
Conversation
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
…pported kwargs (early_stop)
…or compatibility with Titan commit 99c0cb2
Enhanced patch_titan_train_spec to support multi-level nested model overrides
(e.g., "model.moe_args.num_experts" or {"model": {"moe_args": {"num_experts": 16}}}).
Added recursive attribute assignment with dataclass/dict awareness and improved
error messages and logging for better traceability.
…AMD-AIG-AIMA/Primus into fix/torchtitan/patch-checkpoint-wrapper
…TorchTitan trainer Re-enabled DeepSeek-V3 16B and 671B unit tests in . Added explicit CLI overrides to disable PrimusTurbo () and grouped expert matmul () for consistent test behavior across CI environments. Also updated other TorchTitan trainer tests to include PrimusTurbo flag for clarity and reproducibility.
…on module Condensed multi-line torch.split and apply_rotary_emb calls into single-line expressions for improved readability and consistency with surrounding code. No functional change.
…on module Condensed multi-line torch.split and apply_rotary_emb calls into single-line expressions for improved readability and consistency with surrounding code. No functional change.
…on module Condensed multi-line torch.split and apply_rotary_emb calls into single-line expressions for improved readability and consistency with surrounding code. No functional change.
…able DeepSeek-V3 UTs
wenxie-amd
approved these changes
Oct 31, 2025
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
This PR rolls back the integrated TorchTitan backend to 99c0cb2(20250907), restoring compatibility with the current ROCm 7.0 stack and Primus-Turbo extension.
The update also refines TorchTitan trainer unit tests and attention modules to ensure stable end-to-end behavior under the new baseline.