Skip to content

[https://nvbugs/6402058][fix] [None][fix] Remove stale prepare_inputs event sync to unhang DS-V3 EP4 MTP #15849

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
trtllm-agent wants to merge 1 commit into
base: main
Choose a base branch
from
+8 −1
Open

[https://nvbugs/6402058][fix] [None][fix] Remove stale prepare_inputs event sync to unhang DS-V3 EP4 MTP #15849

Show file tree
Hide file tree
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 tensorrt_llm/_torch/modules/mla.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1391,6 +1391,14 @@ def _bmm_bf16_out(self, a, b_no_transpose, b_transposed, output):
if self.use_cute_dsl_bf16_bmm and is_sm_100f():
torch.ops.trtllm.cute_dsl_bf16_bmm_blackwell(a, b_no_transpose, output)
else:
if get_sm_version() in (120, 121) and not a.is_contiguous():

@brnguyen2 brnguyen2 Aug 11, 2026

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The guard densifies only the LHS, but at the call sites the output is also a non-contiguous transposed view (attn_output.transpose(0, 1), mla.py:1578 and mla.py:1700), and b_transposed is a transposed view too. If cuBLAS's kernel heuristic keys off those strides as well, a contiguous a may not be sufficient outside the shapes that were bisected. Worth stating in the comment which operand layouts were actually swept, or asserting/densifying defensively on the output as well.

# `a` is a head-major transpose of a [tokens, heads, dim] buffer, so its
# batch stride is the token count rather than the tile extent. cuBLAS picks
# a TMA-based nvjet kernel for that layout on SM120/121, and
# cuTensorMapEncodeTiled cannot describe it -- the kernel then faults with
# an MMU page fault (surfaced as CUBLAS_STATUS_INTERNAL_ERROR or a later
# illegal memory access). Densify so a non-TMA kernel is selected.
a = a.contiguous()

@yuxianq yuxianq Aug 10, 2026

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This solution is completely different from the PR title and description, please update them.

torch.ops.trtllm.bmm_out(a, b_transposed, output)

def forward_absorption_generation(
Expand Down
1 change: 0 additions & 1 deletion tests/integration/test_lists/waives.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -28,7 +28,6 @@ accuracy/test_llm_api_pytorch.py::TestDeepSeekV32::test_dsa_host_cache_offload[h
accuracy/test_llm_api_pytorch.py::TestDeepSeekV32::test_dsa_host_cache_offload[host_cache_offload_mtp3_no_adp] SKIP (https://nvbugs/6384357)
accuracy/test_llm_api_pytorch.py::TestDeepSeekV32::test_nvfp4_multi_gpus_piecewise_cuda_graph[mtp3_fp8kv_chunked] SKIP (https://nvbugs/5989920)
accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_bfloat16_4gpus[ep4-mtp_nextn=2-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=False] SKIP (https://nvbugs/6517844)
accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_bfloat16_4gpus[ep4-mtp_nextn=2-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=True] SKIP (https://nvbugs/6402058)
accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_bfloat16_4gpus[pp4-mtp_nextn=0-attention_dp=False-cuda_graph=False-overlap_scheduler=False-torch_compile=False] SKIP (https://nvbugs/6278337)
accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_bfloat16_4gpus[pp4-mtp_nextn=0-attention_dp=False-cuda_graph=False-overlap_scheduler=False-torch_compile=True] SKIP (https://nvbugs/6428057)
accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_bfloat16_4gpus[pp4-mtp_nextn=0-attention_dp=False-cuda_graph=True-overlap_scheduler=False-torch_compile=False] SKIP (https://nvbugs/6278337)
Expand Down
Loading