biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[LoRA][1/N]Remove LoRA extra vocab (#28382)

Browse Source 
Signed-off-by: Jee Jee Li <pandaleefree@gmail.com>
This commit is contained in:
Jee Jee Li
2025-11-12 03:06:21 +08:00
committed by GitHub
parent 8c32c6e4b4
commit 9d1c474704
65 changed files with 197 additions and 754 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
vllm/model_executor/models/whisper.py
View File

@@ -890,7 +890,7 @@ class WhisperForConditionalGeneration(
self.dtype = vllm_config.model_config.dtype
self.model = WhisperModel(vllm_config=vllm_config, prefix=prefix)
self.unpadded_vocab_size = config.vocab_size
self.proj_out = ParallelLMHead(
config.vocab_size,
config.d_model,
@@ -899,9 +899,7 @@ class WhisperForConditionalGeneration(
)
self.proj_out = self.proj_out.tie_weights(self.model.decoder.embed_tokens)
logit_scale = getattr(config, "logit_scale", 1.0)
self.logits_processor = LogitsProcessor(
self.unpadded_vocab_size, config.vocab_size, logit_scale
)
self.logits_processor = LogitsProcessor(config.vocab_size, scale=logit_scale)
def forward(
self,