[Fix] Prevent FSDPModule from auto-converting position_embedding to bf16 in DecoderLayer #1462
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
HAOCHENYE
added a commit
that referenced
this pull request
Jan 29, 2026
…f16 in DecoderLayer When FSDPModule is applied to DecoderLayer, it automatically converts position_embedding to bf16, which causes the forward pass output to differ from HuggingFace models beyond the acceptable threshold. This leads to test failures for Qwen VL. The impact on text-only components is currently uncertain. ghstack-source-id: 7bff952 Pull-Request: #1462
This was referenced Jan 28, 2026
Open
HAOCHENYE
added a commit
to HAOCHENYE/xtuner
that referenced
this pull request
Jan 29, 2026
…f16 in DecoderLayer When FSDPModule is applied to DecoderLayer, it automatically converts position_embedding to bf16, which causes the forward pass output to differ from HuggingFace models beyond the acceptable threshold. This leads to test failures for Qwen VL. The impact on text-only components is currently uncertain. ghstack-source-id: 7bff952 Pull-Request: InternLM#1462
HAOCHENYE
added a commit
that referenced
this pull request
Jan 29, 2026
…f16 in DecoderLayer When FSDPModule is applied to DecoderLayer, it automatically converts position_embedding to bf16, which causes the forward pass output to differ from HuggingFace models beyond the acceptable threshold. This leads to test failures for Qwen VL. The impact on text-only components is currently uncertain. ghstack-source-id: 8007bdb Pull-Request: #1462
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Outdated suggestions cannot be applied.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.
Stack from ghstack (oldest at bottom):
fullyshard in BaseModel` #1460chunk_losswhen backward with the frozelm_head#1458XTunerBaseModelConfig#1456DataloaderConfig#1454clean_param_name#1452SequenceContext#1451float8_handlerinfully_shared#1450When FSDPModule is applied to DecoderLayer, it automatically converts position_embedding to bf16, which causes the forward pass output to differ from HuggingFace models beyond the acceptable threshold. This leads to test failures for Qwen VL. The impact on text-only components is currently uncertain.