feat: support LoRA for DeepSeek-OCR(Language Model part) (#31569)

Signed-off-by: zhima771 <15836938703@163.com>
Co-authored-by: Cyrus Leung <tlleungac@connect.ust.hk>
Co-authored-by: Jee Jee Li <pandaleefree@gmail.com>
This commit is contained in:
zhima771
2026-01-01 19:32:11 -08:00
committed by GitHub
co-authored by Cyrus Leung Jee Jee Li
parent 5cc4876630
commit 27864a851c
2 changed files with 14 additions and 2 deletions
+1 -1
View File
@@ -677,7 +677,7 @@ These models primarily accept the [`LLM.generate`](./generative_models.md#llmgen
| `ChameleonForConditionalGeneration` | Chameleon | T + I | `facebook/chameleon-7b`, etc. | | ✅︎ |
| `Cohere2VisionForConditionalGeneration` | Command A Vision | T + I<sup>+</sup> | `CohereLabs/command-a-vision-07-2025`, etc. | | ✅︎ |
| `DeepseekVLV2ForCausalLM`<sup>^</sup> | DeepSeek-VL2 | T + I<sup>+</sup> | `deepseek-ai/deepseek-vl2-tiny`, `deepseek-ai/deepseek-vl2-small`, `deepseek-ai/deepseek-vl2`, etc. | | ✅︎ |
| `DeepseekOCRForCausalLM` | DeepSeek-OCR | T + I<sup>+</sup> | `deepseek-ai/DeepSeek-OCR`, etc. | | ✅︎ |
| `DeepseekOCRForCausalLM` | DeepSeek-OCR | T + I<sup>+</sup> | `deepseek-ai/DeepSeek-OCR`, etc. | ✅︎ | ✅︎ |
| `Ernie4_5_VLMoeForConditionalGeneration` | Ernie4.5-VL | T + I<sup>+</sup>/ V<sup>+</sup> | `baidu/ERNIE-4.5-VL-28B-A3B-PT`, `baidu/ERNIE-4.5-VL-424B-A47B-PT` | | ✅︎ |
| `FuyuForCausalLM` | Fuyu | T + I | `adept/fuyu-8b`, etc. | | ✅︎ |
| `Gemma3ForConditionalGeneration` | Gemma 3 | T + I<sup>E+</sup> | `google/gemma-3-4b-it`, `google/gemma-3-27b-it`, etc. | ✅︎ | ✅︎ |
+13 -1
View File
@@ -14,9 +14,11 @@ from vllm.config import VllmConfig
from vllm.config.multimodal import BaseDummyOptions
from vllm.model_executor.models.interfaces import (
MultiModalEmbeddings,
SupportsLoRA,
SupportsMultiModal,
SupportsPP,
)
from vllm.model_executor.models.module_mapping import MultiModelKeys
from vllm.model_executor.models.utils import (
AutoWeightsLoader,
WeightsMapper,
@@ -343,7 +345,7 @@ class DeepseekOCRMultiModalProcessor(
info=DeepseekOCRProcessingInfo,
dummy_inputs=DeepseekOCRDummyInputsBuilder,
)
class DeepseekOCRForCausalLM(nn.Module, SupportsMultiModal, SupportsPP):
class DeepseekOCRForCausalLM(nn.Module, SupportsMultiModal, SupportsPP, SupportsLoRA):
hf_to_vllm_mapper = WeightsMapper(
orig_to_new_prefix={
# map prefix for language backbone
@@ -589,3 +591,13 @@ class DeepseekOCRForCausalLM(nn.Module, SupportsMultiModal, SupportsPP):
loader = AutoWeightsLoader(self)
autoloaded_weights = loader.load_weights(weights, mapper=self.hf_to_vllm_mapper)
return autoloaded_weights
def get_mm_mapping(self) -> MultiModelKeys:
"""
Get the module prefix in multimodal models
"""
return MultiModelKeys.from_string_field(
language_model="language_model",
connector="projector",
tower_model=["sam_model", "vision_model"],
)