Skip to content

[None][fix] Enable cuda_scaled_mm fast path for FP8 linear on SM121 #15928

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
souvikDevloper wants to merge 5 commits into
base: main
Choose a base branch
from
+3 −4
Open

[None][fix] Enable cuda_scaled_mm fast path for FP8 linear on SM121 #15928

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
581fb47
[None][fix] Enable cuda_scaled_mm fast path for FP8 linear on SM121
Souvikalp Jul 3, 2026
b2beab9
[None][fix] Use membership check for cuda core capability allowlist
Souvikalp Jul 3, 2026
ba74883
Merge branch 'main' into fix/sm121-cuda-scaled-mm
souvikDevloper Jul 3, 2026
2a7cab2
Merge branch 'main' into fix/sm121-cuda-scaled-mm
souvikDevloper Jul 4, 2026
2a21185
Merge branch 'main' into fix/sm121-cuda-scaled-mm
souvikDevloper Jul 4, 2026
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
2 changes: 1 addition & 1 deletion tensorrt_llm/_torch/auto_deploy/custom_ops/quantization/quant.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -109,7 +109,7 @@ def _trtllm_fp8_prequant_linear_core(
enable_cuda_core = False
if torch.cuda.is_available():
capability = torch.cuda.get_device_capability(0)
enable_cuda_core = capability == (8, 9) or capability == (12, 0)
enable_cuda_core = capability in ((8, 9), (12, 0), (12, 1))

if x.shape[0] <= 8 and enable_cuda_core:
output = torch.ops.trtllm.cuda_scaled_mm(
Expand Down
5 changes: 2 additions & 3 deletions tensorrt_llm/_torch/modules/linear.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3273,9 +3273,8 @@ def __init__(
if torch.cuda.is_available():
capability = torch.cuda.get_device_capability(
torch.device('cuda:0'))
# enable cuda core for sm89 and sm120
self.enable_cuda_core = (capability[0] == 8 and capability[1] == 9) \
or (capability[0] == 12 and capability[1] == 0)
# enable cuda core for sm89, sm120 and sm121
self.enable_cuda_core = capability in ((8, 9), (12, 0), (12, 1))

if not skip_create_weights_in_init:
self.create_weights()
Expand Down