Model: Support Qwen 2.5 vl (#3258)

python/sglang/srt/configs/__init__.py

@@ -1,12 +1,15 @@
 from sglang.srt.configs.chatglm import ChatGLMConfig
 from sglang.srt.configs.dbrx import DbrxConfig
 from sglang.srt.configs.exaone import ExaoneConfig
-from sglang.srt.configs.qwen2vl import Qwen2VLConfig, Qwen2VLVisionConfig
+from sglang.srt.configs.qwen2_5_vl_config import (
+    Qwen2_5_VLConfig,
+    Qwen2_5_VLVisionConfig,
+)
 
 __all__ = [
     "ExaoneConfig",
-    "Qwen2VLConfig",
-    "Qwen2VLVisionConfig",
     "ChatGLMConfig",
     "DbrxConfig",
+    "Qwen2_5_VLConfig",
+    "Qwen2_5_VLVisionConfig",
 ]

python/sglang/srt/configs/model_config.py

@@ -403,6 +403,7 @@ def is_multimodal_model(model_architectures: List[str]):
         or "LlavaVidForCausalLM" in model_architectures
         or "MllamaForConditionalGeneration" in model_architectures
         or "Qwen2VLForConditionalGeneration" in model_architectures
+        or "Qwen2_5_VLForConditionalGeneration" in model_architectures
         or "MiniCPMV" in model_architectures
     ):
         return True

python/sglang/srt/configs/qwen2vl.py

@@ -1,130 +0,0 @@
-# coding=utf-8
-# Copyright 2024 The Qwen team, Alibaba Group and the HuggingFace Inc. team.
-# All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-"""Qwen2VL model configuration"""
-
-import os
-from typing import Union
-
-from transformers import PretrainedConfig
-
-
-class Qwen2VLVisionConfig(PretrainedConfig):
-    model_type = "qwen2_vl"
-
-    def __init__(
-        self,
-        depth=32,
-        embed_dim=1280,
-        hidden_size=3584,
-        hidden_act="quick_gelu",
-        mlp_ratio=4,
-        num_heads=16,
-        in_channels=3,
-        patch_size=14,
-        spatial_merge_size=2,
-        temporal_patch_size=2,
-        **kwargs,
-    ):
-        super().__init__(**kwargs)
-
-        self.depth = depth
-        self.embed_dim = embed_dim
-        self.hidden_size = hidden_size
-        self.hidden_act = hidden_act
-        self.mlp_ratio = mlp_ratio
-        self.num_heads = num_heads
-        self.in_channels = in_channels
-        self.patch_size = patch_size
-        self.spatial_merge_size = spatial_merge_size
-        self.temporal_patch_size = temporal_patch_size
-
-    @classmethod
-    def from_pretrained(
-        cls, pretrained_model_name_or_path: Union[str, os.PathLike], **kwargs
-    ) -> "PretrainedConfig":
-        cls._set_token_in_kwargs(kwargs)
-
-        config_dict, kwargs = cls.get_config_dict(
-            pretrained_model_name_or_path, **kwargs
-        )
-
-        if config_dict.get("model_type") == "qwen2_vl":
-            config_dict = config_dict["vision_config"]
-
-        return cls.from_dict(config_dict, **kwargs)
-
-
-class Qwen2VLConfig(PretrainedConfig):
-    model_type = "qwen2_vl"
-
-    def __init__(
-        self,
-        vocab_size=152064,
-        hidden_size=8192,
-        intermediate_size=29568,
-        num_hidden_layers=80,
-        num_attention_heads=64,
-        num_key_value_heads=8,
-        hidden_act="silu",
-        max_position_embeddings=32768,
-        initializer_range=0.02,
-        rms_norm_eps=1e-05,
-        use_cache=True,
-        tie_word_embeddings=False,
-        rope_theta=1000000.0,
-        use_sliding_window=False,
-        sliding_window=4096,
-        max_window_layers=80,
-        attention_dropout=0.0,
-        vision_config=None,
-        rope_scaling=None,
-        **kwargs,
-    ):
-        if isinstance(vision_config, dict):
-            self.vision_config = Qwen2VLVisionConfig(**vision_config)
-        elif vision_config is None:
-            self.vision_config = Qwen2VLVisionConfig()
-
-        self.vocab_size = vocab_size
-        self.max_position_embeddings = max_position_embeddings
-        self.hidden_size = hidden_size
-        self.intermediate_size = intermediate_size
-        self.num_hidden_layers = num_hidden_layers
-        self.num_attention_heads = num_attention_heads
-        self.use_sliding_window = use_sliding_window
-        self.sliding_window = sliding_window
-        self.max_window_layers = max_window_layers
-
-        # for backward compatibility
-        if num_key_value_heads is None:
-            num_key_value_heads = num_attention_heads
-
-        self.num_key_value_heads = num_key_value_heads
-        self.hidden_act = hidden_act
-        self.initializer_range = initializer_range
-        self.rms_norm_eps = rms_norm_eps
-        self.use_cache = use_cache
-        self.rope_theta = rope_theta
-        self.attention_dropout = attention_dropout
-        self.rope_scaling = rope_scaling
-
-        # NOTE(HandH1998): This is necessary for configuring the `rope_type`` of qwen2vl models after removing dependencies on vllm.
-        if self.rope_scaling is not None and "type" in self.rope_scaling:
-            if self.rope_scaling["type"] == "mrope":
-                self.rope_scaling["type"] = "default"
-            self.rope_scaling["rope_type"] = self.rope_scaling["type"]
-
-        super().__init__(tie_word_embeddings=tie_word_embeddings, **kwargs)