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

[Misc] Make MM embedding merge interface explicit in model runner (#21147)

Browse Source 
Signed-off-by: Roger Wang <hey@rogerw.me>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
This commit is contained in:
Roger Wang
2025-07-18 00:13:57 -07:00
committed by GitHub
parent 1bf65138f6
commit ba2dfbb0c2
2 changed files with 8 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
vllm/v1/worker/gpu_model_runner.py
View File

@ -1328,11 +1328,10 @@ class GPUModelRunner(LoRAModelRunnerMixin):
# embeddings), we always use embeddings (rather than token ids)
# as input to the multimodal model, even when the input is text.
input_ids = self.input_ids[:num_scheduled_tokens]
if mm_embeds:
inputs_embeds = self.model.get_input_embeddings(
input_ids, mm_embeds)
else:
inputs_embeds = self.model.get_input_embeddings(input_ids)
inputs_embeds = self.model.get_input_embeddings(
input_ids=input_ids,
multimodal_embeddings=mm_embeds or None,
)
# TODO(woosuk): Avoid the copy. Optimize.
self.inputs_embeds[:num_scheduled_tokens].copy_(inputs_embeds)
inputs_embeds = self.inputs_embeds[:num_input_tokens]

9
vllm/v1/worker/tpu_model_runner.py
View File

@ -937,11 +937,10 @@ class TPUModelRunner(LoRAModelRunnerMixin):
# NOTE(woosuk): To unify token ids and soft tokens (vision
# embeddings), we always use embeddings (rather than token ids)
# as input to the multimodal model, even when the input is text.
if mm_embeds:
inputs_embeds = self.model.get_input_embeddings(
input_ids, mm_embeds)
else:
inputs_embeds = self.model.get_input_embeddings(input_ids)
inputs_embeds = self.model.get_input_embeddings(
input_ids=input_ids,
multimodal_embeddings=mm_embeds,
)
return None, inputs_embeds
else:
# For text-only models, we use token ids as input.