Skip to content

[https://nvbugs/5385987][fix] Fix Qwen2 quantization issue by pinning transformers version #6673

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 5 commits into from
Aug 12, 2025
Merged

[https://nvbugs/5385987][fix] Fix Qwen2 quantization issue by pinning transformers version #6673

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
fd3d428
Pin transformers to 4.51 for modelopt qunatization support for qwen
chang-l Aug 6, 2025
2dbb7f7
Unwaive test
chang-l Aug 6, 2025
523af97
Fix format
chang-l Aug 6, 2025
6908912
Rebase to latest TOT and fix compat issue
chang-l Aug 7, 2025
c2479e6
add comment
chang-l Aug 11, 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
1 change: 1 addition & 0 deletions examples/models/core/multimodal/requirements-qwen2vl.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,2 +1,3 @@
accelerate
qwen-vl-utils==0.0.8 # 0.0.9 has bug https://github.com/QwenLM/Qwen2-VL/pull/673, rollback until a newer version is released
transformers==4.51.0 # nvbugs/5385987
12 changes: 10 additions & 2 deletions tensorrt_llm/tools/multimodal_builder.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1324,6 +1324,7 @@ def rot_pos_emb(grid_thw, rotary_pos_emb_func):


def build_qwen2_vl_engine(args):
import transformers
from qwen_vl_utils import process_vision_info
from transformers import AutoProcessor, Qwen2VLForConditionalGeneration
from transformers.models.qwen2_vl.configuration_qwen2_vl import \
Expand Down Expand Up @@ -1391,8 +1392,15 @@ def build_qwen2_vl_engine(args):
class VisionAttentionOpt(VisionAttention):

def __init__(self, config: Qwen2VLVisionConfig):
super().__init__(config)
self.head_dim = config.embed_dim // config.num_heads
# Fallback for compatibility with older transformers versions (for certain nvbugs/tests)
if transformers.__version__ >= '4.53.0':
super().__init__(config)
self.head_dim = config.embed_dim // config.num_heads
else:
num_heads = config.num_heads
dim = config.embed_dim
super().__init__(dim, num_heads)
self.head_dim = dim // num_heads

def forward(self,
hidden_states: torch.Tensor,
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 @@ -235,7 +235,6 @@ stress_test/stress_test.py::test_run_stress_test[llama-v3-8b-instruct-hf_tp1-str
accuracy/test_llm_api_pytorch.py::TestGemma3_1BInstruct::test_auto_dtype SKIP (https://nvbugs/5375620)
test_e2e.py::test_ptp_quickstart_advanced_8gpus[Llama3.1-405B-FP8-llama-3.1-model/Llama-3.1-405B-Instruct-FP8] SKIP (https://nvbugs/5380570)
test_e2e.py::test_ptp_quickstart_advanced_8gpus[Nemotron-Ultra-253B-nemotron-nas/Llama-3_1-Nemotron-Ultra-253B-v1] SKIP (https://nvbugs/5380570)
examples/test_multimodal.py::test_llm_fp8_multimodal_general[fp8-fp8-cnn_dailymail-Qwen2-VL-7B-Instruct-pp:1-tp:1-bfloat16-bs:1-cpp_e2e:False] SKIP (https://nvbugs/5385987)
examples/test_multimodal.py::test_llm_multimodal_general[Phi-4-multimodal-instruct-pp:1-tp:1-float16-bs:1-cpp_e2e:False-nb:1] SKIP (https://nvbugs/5385992)
examples/test_multimodal.py::test_llm_multimodal_general[kosmos-2-pp:1-tp:1-float16-bs:1-cpp_e2e:False-nb:1] SKIP (https://nvbugs/5387422)
examples/test_multimodal.py::test_llm_multimodal_general[fuyu-8b-pp:1-tp:1-float16-bs:1-cpp_e2e:False-nb:1] SKIP (https://nvbugs/5387424)
Expand Down