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 15, 23:30 UTC
Set `torch_dtype` in `TransformersModel` (#13088)
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Harry Mellor committed
Feb 11, 2025 at 15:51 UTC
ad9776353e6b00d019415e94fd17c78ad4575ff7
1 file changed
+1
vllm/model_executor/models/transformers.py
+1
@@ -143,6 +143,7 @@ class TransformersModel(nn.Module):
143
self.model: PreTrainedModel = AutoModel.from_config(
144
self.config,
145
attn_implementation="vllm",
146
+ torch_dtype=vllm_config.model_config.dtype,
147
trust_remote_code=vllm_config.model_config.trust_remote_code,
148
)
149
prefix = self.model.base_model_prefix