Skip to content

[#9496][fix] AutoDeploy: remove auto-tuner from nvfp4_gemm forward #9497

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
nzmora-nvidia merged 2 commits into from
Dec 1, 2025
Merged

[#9496][fix] AutoDeploy: remove auto-tuner from nvfp4_gemm forward #9497

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
f9c8223
iAutoDeploy: remove auttuner from nvfp4_gemm forward
nzmora-nvidia Nov 26, 2025
dcac75a
Merge branch 'main' into users/nzmora/remove_nvfp4_gemm_autotuner
nzmora-nvidia Nov 30, 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
9 changes: 3 additions & 6 deletions tensorrt_llm/_torch/auto_deploy/custom_ops/quant.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,8 +7,6 @@
from flashinfer import bmm_fp8
from torch import nn

from tensorrt_llm._torch.autotuner import autotune

from ..distributed import common as dist
from ..distributed import trtllm as trtllm_dist
from .torch_libs.float8_python_api import addmm_float8_unwrapped
Expand Down Expand Up @@ -336,10 +334,9 @@ def nvfp4_linear(
x_fp4, x_sf_block = torch.ops.trtllm.fp4_quantize(
input, input_scale, TRTLLM_NVFP4_SCALING_VECTOR_SIZE, False
)
with autotune():
output = torch.ops.trtllm.nvfp4_gemm(
x_fp4, weight_fp4, x_sf_block, weight_scale, alpha, input.dtype
)
output = torch.ops.trtllm.nvfp4_gemm(
x_fp4, weight_fp4, x_sf_block, weight_scale, alpha, input.dtype
)

if bias is not None:
output = output + bias
Expand Down