Skip to content

Scatter embeddings for sequence parallelism in standalone LM forwards #5628

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.

Sign up for GitHub

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

Open
kevalmorabia97 wants to merge 1 commit into
base: main
Choose a base branch
from
+8 −0
Open

Scatter embeddings for sequence parallelism in standalone LM forwards #5628

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 8 additions & 0 deletions megatron/core/models/gpt/gpt_model.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -332,6 +332,14 @@ def _preprocess(
f"input_ids shape {input_ids.shape}"
)
decoder_input = self.embedding(input_ids=input_ids, position_ids=position_ids)
if self.config.sequence_parallel and not self.embedding.scatter_to_sequence_parallel:
# Some models build the embedding to not scatter for sequence parallelism (e.g.
# VLM language models, whose outer multimodal model scatters the merged
# vision/audio+text embeddings itself). When such a model is run standalone
# (input_ids, no external decoder_input) -- e.g. language-model-only
# distillation/PTQ -- scatter here so the decoder sees a sequence-parallel
# sharded sequence and the output-side gather does not double the sequence.
decoder_input = tensor_parallel.scatter_to_sequence_parallel_region(decoder_input)
if padding_mask is not None and self.config.sequence_parallel:
padding_mask = (
tensor_parallel.scatter_to_sequence_parallel_region(
Expand Down
Loading