Skip to content

[https://nvbugs/5636986][fix] Fix DeepGemmMoe get_buffer calls #8939

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
merged 2 commits into from
Nov 7, 2025
Merged

[https://nvbugs/5636986][fix] Fix DeepGemmMoe get_buffer calls #8939

Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
d8fec4e
[https://nvbugs/5636986][fix] Fix DeepGemmMoe get_buffer calls
VALLIS-NERIA Nov 5, 2025
58208de
Remove duplicate SKIP entries in waives.txt
VALLIS-NERIA Nov 7, 2025
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
10 changes: 6 additions & 4 deletions tensorrt_llm/_torch/modules/fused_moe/ops/moe_op_deepgemm.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -89,22 +89,24 @@ def _get_deepgemm_workspace(self, module: 'MoE', m_max: int,
# Workspace for FP8 activations
capture_graph = torch.cuda.is_current_stream_capturing()
workspace["workspace_0"] = DeepGemmMoEOp.buffers.get_buffer(
(expert_size_per_partition * m_max * fp8_dim),
[expert_size_per_partition, m_max, fp8_dim],
dtype=torch.float8_e4m3fn,
buffer_name='workspace_0',
reserve_buffer=capture_graph)

# Workspace for intermediate results
workspace["workspace_1"] = DeepGemmMoEOp.buffers.get_buffer(
(expert_size_per_partition * m_max *
max(intermediate_size * 2, hidden_size)),
[
expert_size_per_partition, m_max,
max(intermediate_size * 2, hidden_size)
],
dtype=torch.bfloat16,
buffer_name='workspace_1',
reserve_buffer=capture_graph)

# Workspace for scaling factors
workspace["workspace_sf"] = DeepGemmMoEOp.buffers.get_buffer(
expert_size_per_partition * (scale_k_padded // 4) * m_padded,
[expert_size_per_partition, (scale_k_padded // 4), m_padded],
dtype=torch.int32,
buffer_name='workspace_sf',
reserve_buffer=capture_graph)
Expand Down
Loading