kanshan/TensorRT-LLMs
1
0
Fork 0
mirror of https://github.com/NVIDIA/TensorRT-LLM.git synced 2026-01-14 06:27:45 +08:00
Code Issues Actions 1 Packages Projects Releases Wiki Activity

fix 5505835

Browse Source 
Signed-off-by: Xiwen Yu <13230610+VALLIS-NERIA@users.noreply.github.com>
This commit is contained in:
Xiwen Yu 2025-09-08 14:52:00 +08:00
parent d4d9e778a1
commit 019b1db438
2 changed files with 2 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
tensorrt_llm/_torch/auto_deploy/models/patches/llama4.py
View File

@ -104,7 +104,7 @@ def _forward_with_cond(
return inputs_embeds
def _no_vision_branch(inputs_embeds, pixel_values, input_ids):
return inputs_embeds
return inputs_embeds.clone()
# decide by whether there is any non-zero pixel_values
has_image: torch.Tensor = torch.any(pixel_values != 0)

3
tests/unittest/_torch/auto_deploy/unit/singlegpu/models/test_llama4_vlm_patch.py
View File

@ -1,4 +1,3 @@
import pytest
import torch
from _model_test_utils import get_small_model_config
from build_and_run_ad import ExperimentConfig
@ -9,7 +8,7 @@ from tensorrt_llm._torch.auto_deploy.export import apply_export_patches, torch_e
from tensorrt_llm._torch.auto_deploy.transformations._graph import move_to_device
@pytest.mark.skip(reason="https://nvbugspro.nvidia.com/bug/5505835")
# @pytest.mark.skip(reason="https://nvbugspro.nvidia.com/bug/5505835")
def test_build_run_llama4_vlm():
atol = 1e-3
rtol = 1e-3