__init__ of qwen2vl

Author: rebel-thkim

src/optimum/rbln/__init__.py

@@ -158,6 +158,8 @@
         "RBLNQwen2VisionTransformerPretrainedModelConfig",
         "RBLNQwen2VLForConditionalGeneration",
         "RBLNQwen2VLForConditionalGenerationConfig",
+        "RBLNQwen2VLModel",
+        "RBLNQwen2VLModelConfig",
         "RBLNResNetForImageClassification",
         "RBLNResNetForImageClassificationConfig",
         "RBLNRobertaForMaskedLM",
@@ -450,6 +452,8 @@
         RBLNQwen2VisionTransformerPretrainedModelConfig,
         RBLNQwen2VLForConditionalGeneration,
         RBLNQwen2VLForConditionalGenerationConfig,
+        RBLNQwen2VLModel,
+        RBLNQwen2VLModelConfig,
         RBLNQwen3ForCausalLM,
         RBLNQwen3ForCausalLMConfig,
         RBLNQwen3Model,

src/optimum/rbln/transformers/__init__.py

@@ -140,6 +140,8 @@
         "RBLNQwen2VisionTransformerPretrainedModelConfig",
         "RBLNQwen2VLForConditionalGeneration",
         "RBLNQwen2VLForConditionalGenerationConfig",
+        "RBLNQwen2VLModel",
+        "RBLNQwen2VLModelConfig",
         "RBLNQwen2Model",
         "RBLNQwen2ModelConfig",
         "RBLNQwen2ForCausalLM",
@@ -302,6 +304,8 @@
         RBLNQwen2VisionTransformerPretrainedModelConfig,
         RBLNQwen2VLForConditionalGeneration,
         RBLNQwen2VLForConditionalGenerationConfig,
+        RBLNQwen2VLModel,
+        RBLNQwen2VLModelConfig,
         RBLNQwen3ForCausalLM,
         RBLNQwen3ForCausalLMConfig,
         RBLNQwen3Model,

src/optimum/rbln/transformers/models/__init__.py

@@ -96,6 +96,8 @@
         "RBLNQwen2VisionTransformerPretrainedModelConfig",
         "RBLNQwen2VLForConditionalGeneration",
         "RBLNQwen2VLForConditionalGenerationConfig",
+        "RBLNQwen2VLModel",
+        "RBLNQwen2VLModelConfig",
     ],
     "decoderonly": [
         "RBLNDecoderOnlyModelConfig",
@@ -305,6 +307,8 @@
         RBLNQwen2VisionTransformerPretrainedModelConfig,
         RBLNQwen2VLForConditionalGeneration,
         RBLNQwen2VLForConditionalGenerationConfig,
+        RBLNQwen2VLModel,
+        RBLNQwen2VLModelConfig,
     )
     from .qwen3 import RBLNQwen3ForCausalLM, RBLNQwen3ForCausalLMConfig, RBLNQwen3Model, RBLNQwen3ModelConfig
     from .resnet import RBLNResNetForImageClassification, RBLNResNetForImageClassificationConfig

src/optimum/rbln/transformers/models/qwen2_vl/__init__.py

@@ -15,5 +15,10 @@
 from .configuration_qwen2_vl import (
     RBLNQwen2VisionTransformerPretrainedModelConfig,
     RBLNQwen2VLForConditionalGenerationConfig,
+    RBLNQwen2VLModelConfig,
+)
+from .modeling_qwen2_vl import (
+    RBLNQwen2VisionTransformerPretrainedModel,
+    RBLNQwen2VLForConditionalGeneration,
+    RBLNQwen2VLModel,
 )
-from .modeling_qwen2_vl import RBLNQwen2VisionTransformerPretrainedModel, RBLNQwen2VLForConditionalGeneration

src/optimum/rbln/transformers/models/qwen2_vl/configuration_qwen2_vl.py

@@ -15,7 +15,7 @@
 from typing import Any, Dict, List, Optional, Union
 
 from ....configuration_utils import RBLNModelConfig
-from ..decoderonly.configuration_decoderonly import RBLNDecoderOnlyModelForCausalLMConfig
+from ..decoderonly.configuration_decoderonly import RBLNDecoderOnlyModelConfig, RBLNDecoderOnlyModelForCausalLMConfig
 
 
 class RBLNQwen2VLForConditionalGenerationConfig(RBLNDecoderOnlyModelForCausalLMConfig):
@@ -48,6 +48,16 @@ def __init__(
         self.visual = visual
 
 
+class RBLNQwen2VLModelConfig(RBLNDecoderOnlyModelConfig):
+    """
+    Configuration class for RBLNQwen2VLModel.
+    """
+
+    def __init__(self, visual: Optional[RBLNModelConfig] = None, **kwargs: Dict[str, Any]):
+        super().__init__(**kwargs)
+        self.visual = self.initialize_submodule_config(submodule_config=visual)
+
+
 class RBLNQwen2VisionTransformerPretrainedModelConfig(RBLNModelConfig):
     def __init__(self, max_seq_lens: Union[int, List[int]] = None, **kwargs: Dict[str, Any]):
         """