[Model] Remove unnecessary get_language_model (#37545)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung
2026-03-19 20:00:36 +08:00
committed by GitHub
parent c7bc12c20f
commit 7a6ebcbfcf
7 changed files with 93 additions and 95 deletions
+23 -22
View File
@@ -163,29 +163,30 @@ class LightOnOCRForConditionalGeneration(Mistral3ForConditionalGeneration):
self.config = config
self.multimodal_config = multimodal_config
self.vision_tower = init_vision_tower_for_llava(
config,
quant_config=quant_config,
require_post_norm=False,
prefix=maybe_prefix(prefix, "vision_tower"),
)
with self._mark_tower_model(vllm_config, "image"):
self.vision_tower = init_vision_tower_for_llava(
config,
quant_config=quant_config,
require_post_norm=False,
prefix=maybe_prefix(prefix, "vision_tower"),
)
self.multi_modal_projector = Mistral3MultiModalProjector(
vision_hidden_size=config.vision_config.hidden_size,
text_hidden_size=config.text_config.hidden_size,
projector_hidden_act=config.projector_hidden_act,
spatial_merge_size=config.spatial_merge_size,
patch_size=config.vision_config.patch_size,
multimodal_projector_bias=config.multimodal_projector_bias,
quant_config=quant_config,
prefix=maybe_prefix(prefix, "multi_modal_projector"),
)
self.multi_modal_projector = Mistral3MultiModalProjector(
vision_hidden_size=config.vision_config.hidden_size,
text_hidden_size=config.text_config.hidden_size,
projector_hidden_act=config.projector_hidden_act,
spatial_merge_size=config.spatial_merge_size,
patch_size=config.vision_config.patch_size,
multimodal_projector_bias=config.multimodal_projector_bias,
quant_config=quant_config,
prefix=maybe_prefix(prefix, "multi_modal_projector"),
)
self.language_model = init_vllm_registered_model(
vllm_config=vllm_config,
hf_config=config.text_config,
prefix=maybe_prefix(prefix, "language_model"),
)
with self._mark_language_model(vllm_config):
self.language_model = init_vllm_registered_model(
vllm_config=vllm_config,
hf_config=config.text_config,
prefix=maybe_prefix(prefix, "language_model"),
)
self.make_empty_intermediate_tensors = (
self.language_model.make_empty_intermediate_tensors