feat(deepseek_v41): defer activation-recompute decision to the V4.1 b… - #205
Merged
tastelikefeet merged 1 commit intoSep 22, 2026
Conversation
…ackbone Engram's _validate_unsupported_features rejects any activation recomputation, a guard that predates the V4.1 hybrid backbone. That backbone now owns the recompute policy in TransformerConfig._validate_dsv41_config and threads input_ids through _checkpointed_forward for the replayed Engram forward. Add _RecomputeDeferredView, a read-only view that masks only recompute_granularity, and override DeepseekV41EngramConfig._validate_unsupported_features to validate through it. This lets full/selective recompute work for V4.1+Engram while keeping every other Engram unsupported-feature check live and without mutating the shared transformer config.
hjh0119
approved these changes
Sep 21, 2026
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
3 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
…ackbone
Engram's _validate_unsupported_features rejects any activation recomputation, a guard that predates the V4.1 hybrid backbone. That backbone now owns the recompute policy in TransformerConfig._validate_dsv41_config and threads input_ids through _checkpointed_forward for the replayed Engram forward.
Add _RecomputeDeferredView, a read-only view that masks only recompute_granularity, and override DeepseekV41EngramConfig._validate_unsupported_features to validate through it. This lets full/selective recompute work for V4.1+Engram while keeping every other Engram unsupported-feature check live and without mutating the shared transformer config.