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

[4/N] Initialize MM components in context managers (M-P) (#32663)

Browse Source 
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung
2026-01-20 22:06:32 +08:00
committed by GitHub
parent bb9172030e
commit fda3f03eb2
24 changed files with 290 additions and 353 deletions
Download Patch File Download Diff File Expand all files Collapse all files

26
vllm/model_executor/models/molmo.py
View File

@@ -1441,15 +1441,20 @@ class MolmoForCausalLM(
self.multimodal_config = multimodal_config
vision_config = VisionBackboneConfig()
self.vision_backbone = MolmoVisionBackbone(
config,
vision_config,
quant_config,
prefix=maybe_prefix(prefix, "vision_backbone"),
)
self.model = MolmoModel(
vllm_config=vllm_config, prefix=maybe_prefix(prefix, "model")
)
with self._mark_tower_model(vllm_config, "image"):
self.vision_backbone = MolmoVisionBackbone(
config,
vision_config,
quant_config,
prefix=maybe_prefix(prefix, "vision_backbone"),
)
with self._mark_language_model(vllm_config):
self.model = MolmoModel(
vllm_config=vllm_config, prefix=maybe_prefix(prefix, "model")
)
self.img_patch_id = None
if self.config.weight_tying:
@@ -1525,9 +1530,6 @@ class MolmoForCausalLM(
results.append(feats[is_valid][order])
return results
def get_language_model(self) -> torch.nn.Module:
return self.model
def embed_multimodal(self, **kwargs: object) -> MultiModalEmbeddings:
image_input = self._parse_and_validate_image_input(**kwargs)
if image_input is None: