Skip to content

[TRTLLM-14054][fix] LMHead: pass true full dims to Linear for quantized TP>1 #16352

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

Merged
QiJune merged 1 commit into from
Jul 14, 2026
Merged

[TRTLLM-14054][fix] LMHead: pass true full dims to Linear for quantized TP>1 #16352

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
14 changes: 12 additions & 2 deletions tensorrt_llm/_torch/modules/embedding.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -58,9 +58,19 @@ def __init__(
else:
self.padding_size = 0

# Linear re-shards only the dim selected by tensor_parallel_mode, so
# only that dim may be passed as local*tp (to carry the ceil-div
# padding); the other dim must be the true full size — Linear keeps it
# as-is, and quant_method.create_weights sizes the (packed) quantized
# weight and scales from it.
in_features_full = (local_in_features * tp_size if tensor_parallel_mode
== TensorParallelMode.ROW else embedding_dim)
out_features_full = (local_out_features *
tp_size if tensor_parallel_mode
== TensorParallelMode.COLUMN else num_embeddings)
super().__init__(
local_in_features * tp_size,
local_out_features * tp_size,
in_features_full,
out_features_full,
bias=False,
dtype=dtype,
mapping=mapping,
Expand Down
Loading