Chranos ebdc6fed03 fix: pass lm_head to LogitsProcessor instead of calling forward() ...

In vLLM v0.6.2, ParallelLMHead.forward() raises RuntimeError since
its weights should be used through LogitsProcessor.linear_method.apply().
Pass lm_head as first arg to LogitsProcessor which handles the
hidden_states -> logits projection internally.

2026-02-06 14:21:14 +08:00

..

__pycache__

add qwen3

2026-02-04 17:22:39 +08:00

guided_decoding

add qwen3

2026-02-04 17:22:39 +08:00

layers

testing dynamic register

2026-02-05 18:36:03 +08:00

model_loader

add dynamic register

2026-02-05 15:53:43 +08:00

models

fix: pass lm_head to LogitsProcessor instead of calling forward()

2026-02-06 14:21:14 +08:00

__init__.py

add qwen3

2026-02-04 17:22:39 +08:00

custom_op.py

add qwen3

2026-02-04 17:22:39 +08:00

parameter.py

add qwen3

2026-02-04 17:22:39 +08:00

pooling_metadata.py

add qwen3

2026-02-04 17:22:39 +08:00

sampling_metadata.py

add qwen3

2026-02-04 17:22:39 +08:00

utils.py

add qwen3

2026-02-04 17:22:39 +08:00