@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / 3e1ad4435f

Fork 0

A high-throughput and memory-efficient inference and serving engine for LLMs

Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git

Synced Nov 4, 08:45 UTC

[Bug] Fix `tests/compile/test_config.py` AttributeError: 'NoneType' object has no attribute 'dtype' (#41288)

Signed-off-by: yewentao256 <zhyanwentao@126.com>

Wentao Ye committed May 4, 2026 at 12:22 UTC 3e1ad4435f7c205dcbd5b14d1a529cc328b22ce8

1 file changed +1 -1

vllm/config/vllm.py

+1 -1

@@ -1617,7 +1617,7 @@ class VllmConfig:

1617 max_size = rocm_aiter_ops.get_aiter_allreduce_max_size()

1618 else:

1619 max_size = compilation_config.pass_config.flashinfer_max_size(tp_size)

1620 - if max_size is not None:

1620 + if max_size is not None and self.model_config is not None:

1621 assert isinstance(self.model_config.dtype, torch.dtype)

1622 max_token_num = max_size // (

1623 self.model_config.get_hidden_size()