@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / b7df53cd42
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 16, 09:30 UTC

[Bugfix] Use "vision_model" prefix for MllamaVisionModel (#9628)

Signed-off-by: mgoin <michael@neuralmagic.com>

Michael Goin committed Oct 23, 2024 at 22:07 UTC b7df53cd42f3eab007b4f287c151960858e949df
1 file changed +2 -1
vllm/model_executor/models/mllama.py
+2 -1
@@ -1053,7 +1053,8 @@ class MllamaForConditionalGeneration(nn.Module, SupportsMultiModal):
1053 self.image_size = config.vision_config.image_size
1054
1055 self.vision_model = MllamaVisionModel(config.vision_config,
1056 - quant_config)
1056 + quant_config,
1057 + prefix="vision_model")
1058 self.language_model = MllamaForCausalLM(
1059 config.text_config,
1060 cache_config=cache_config,