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 6, 10:45 UTC
[LoRA][BugFix] Fix skipped LoRA adapters for Mistral3 (#36928)
Signed-off-by: Woosuk Kwon <woosuk@inferact.ai>
Woosuk Kwon committed
Mar 18, 2026 at 15:34 UTC
5bc1da147fb02957c57ba1c6284e16ed578363ea
1 file changed
+3
vllm/model_executor/models/mistral3.py
+3
@@ -429,6 +429,9 @@ class Mistral3ForConditionalGeneration(
429
"model.vision_tower.": "vision_tower.",
430
"model.multi_modal_projector.": "multi_modal_projector.",
431
"lm_head.": "language_model.lm_head.",
432
+ # Some PEFT LoRAs are trained against the text submodule directly
433
+ # and produce names like `base_model.model.model.layers.*`.
434
+ "model.": "language_model.model.",
435
}
436
)
437