[Misc] Qwen2.5 VL support LoRA (#13261)

docs/source/models/supported_models.md

@@ -854,7 +854,7 @@ See [this page](#generative-models) for more information on how to use generativ
   * Qwen2.5-VL
   * T + I<sup>E+</sup> + V<sup>E+</sup>
   * `Qwen/Qwen2.5-VL-3B-Instruct`, `Qwen/Qwen2.5-VL-72B-Instruct`, etc.
-  *
+  * ✅︎
   * ✅︎
   * ✅︎
 - * `UltravoxModel`

tests/lora/conftest.py

@@ -237,6 +237,11 @@ def qwen2vl_lora_files():
     return snapshot_download(repo_id="jeeejeee/qwen2-vl-lora-pokemon")
 
 
+@pytest.fixture(scope="session")
+def qwen25vl_lora_files():
+    return snapshot_download(repo_id="jeeejeee/qwen25-vl-lora-pokemon")
+
+
 @pytest.fixture(scope="session")
 def tinyllama_lora_files():
     return snapshot_download(repo_id="jashing/tinyllama-colorist-lora")

tests/lora/test_qwen2vl.py

@@ -1,15 +1,37 @@
 # SPDX-License-Identifier: Apache-2.0
-
+from dataclasses import dataclass
-from typing import List
+from typing import Dict, List, Optional
 
 import pytest
+from packaging.version import Version
+from transformers import __version__ as TRANSFORMERS_VERSION
 
 import vllm
 from vllm.assets.image import ImageAsset
 from vllm.lora.request import LoRARequest
 from vllm.platforms import current_platform
 
-MODEL_PATH = "Qwen/Qwen2-VL-2B-Instruct"
+
+@dataclass
+class TestConfig:
+    model_path: str
+    lora_path: str
+    max_num_seqs: int = 2
+    max_loras: int = 2
+    max_lora_rank: int = 16
+    max_model_len: int = 4096
+    mm_processor_kwargs: Optional[Dict[str, int]] = None
+
+    def __post_init__(self):
+        if self.mm_processor_kwargs is None:
+            self.mm_processor_kwargs = {
+                "min_pixels": 28 * 28,
+                "max_pixels": 1280 * 28 * 28,
+            }
+
+
+class Qwen2VLTester:
+    """Test helper for Qwen2 VL models with LoRA"""
 
     PROMPT_TEMPLATE = (
         "<|im_start|>system\nYou are a helpful assistant.<|im_end|>"
@@ -17,67 +39,105 @@ PROMPT_TEMPLATE = (
         "What is in the image?<|im_end|>\n"
         "<|im_start|>assistant\n")
 
-IMAGE_ASSETS = [
+    def __init__(self, config: TestConfig):
+        self.config = config
+        self.llm = self._initialize_llm()
+
+    def _initialize_llm(self) -> vllm.LLM:
+        """Initialize the LLM with given configuration"""
+        return vllm.LLM(
+            model=self.config.model_path,
+            max_num_seqs=self.config.max_num_seqs,
+            enable_lora=True,
+            max_loras=self.config.max_loras,
+            max_lora_rank=self.config.max_lora_rank,
+            trust_remote_code=True,
+            mm_processor_kwargs=self.config.mm_processor_kwargs,
+            max_model_len=self.config.max_model_len,
+        )
+
+    def run_test(self,
+                 images: List[ImageAsset],
+                 expected_outputs: List[str],
+                 lora_id: Optional[int] = None,
+                 temperature: float = 0,
+                 max_tokens: int = 5) -> List[str]:
+
+        sampling_params = vllm.SamplingParams(
+            temperature=temperature,
+            max_tokens=max_tokens,
+        )
+        inputs = [{
+            "prompt": self.PROMPT_TEMPLATE,
+            "multi_modal_data": {
+                "image": asset.pil_image
+            },
+        } for asset in images]
+
+        lora_request = LoRARequest(str(lora_id), lora_id,
+                                   self.config.lora_path)
+        outputs = self.llm.generate(inputs,
+                                    sampling_params,
+                                    lora_request=lora_request)
+        generated_texts = [
+            output.outputs[0].text.strip() for output in outputs
+        ]
+
+        # Validate outputs
+        for generated, expected in zip(generated_texts, expected_outputs):
+            assert expected.startswith(
+                generated), f"Generated text {generated} doesn't "
+            f"match expected pattern {expected}"
+
+        return generated_texts
+
+
+TEST_IMAGES = [
     ImageAsset("stop_sign"),
     ImageAsset("cherry_blossom"),
 ]
 
-# After fine-tuning with LoRA, all generated content should start begin `A`.
+EXPECTED_OUTPUTS = [
-EXPECTED_OUTPUT = [
     "A red stop sign stands prominently in the foreground, with a traditional Chinese gate and a black SUV in the background, illustrating a blend of modern and cultural elements.",  # noqa: E501
     "A majestic skyscraper stands tall, partially obscured by a vibrant canopy of cherry blossoms, against a clear blue sky.",  # noqa: E501
 ]
 
-
+QWEN2VL_MODEL_PATH = "Qwen/Qwen2-VL-2B-Instruct"
-def do_sample(llm: vllm.LLM, lora_path: str, lora_id: int) -> List[str]:
+QWEN25VL_MODEL_PATH = "Qwen/Qwen2.5-VL-3B-Instruct"
-    sampling_params = vllm.SamplingParams(
-        temperature=0,
-        max_tokens=5,
-    )
-
-    inputs = [{
-        "prompt": PROMPT_TEMPLATE,
-        "multi_modal_data": {
-            "image": asset.pil_image
-        },
-    } for asset in IMAGE_ASSETS]
-
-    outputs = llm.generate(
-        inputs,
-        sampling_params,
-        lora_request=LoRARequest(str(lora_id), lora_id, lora_path)
-        if lora_id else None,
-    )
-    # Print the outputs.
-    generated_texts: List[str] = []
-    for output in outputs:
-        generated_text = output.outputs[0].text.strip()
-        generated_texts.append(generated_text)
-        print(f"Generated text: {generated_text!r}")
-    return generated_texts
 
 
 @pytest.mark.xfail(
     current_platform.is_rocm(),
     reason="Qwen2-VL dependency xformers incompatible with ROCm")
 def test_qwen2vl_lora(qwen2vl_lora_files):
-    llm = vllm.LLM(
+    """Test Qwen 2.0 VL model with LoRA"""
-        MODEL_PATH,
+    config = TestConfig(model_path=QWEN2VL_MODEL_PATH,
-        max_num_seqs=2,
+                        lora_path=qwen2vl_lora_files)
-        enable_lora=True,
+    tester = Qwen2VLTester(config)
-        max_loras=2,
-        max_lora_rank=16,
-        trust_remote_code=True,
-        mm_processor_kwargs={
-            "min_pixels": 28 * 28,
-            "max_pixels": 1280 * 28 * 28,
-        },
-        max_model_len=4096,
-    )
-    output1 = do_sample(llm, qwen2vl_lora_files, lora_id=1)
-    for i in range(len(EXPECTED_OUTPUT)):
-        assert EXPECTED_OUTPUT[i].startswith(output1[i])
 
-    output2 = do_sample(llm, qwen2vl_lora_files, lora_id=2)
+    # Test with different LoRA IDs
-    for i in range(len(EXPECTED_OUTPUT)):
+    for lora_id in [1, 2]:
-        assert EXPECTED_OUTPUT[i].startswith(output2[i])
+        tester.run_test(TEST_IMAGES,
+                        expected_outputs=EXPECTED_OUTPUTS,
+                        lora_id=lora_id)
+
+
+@pytest.mark.xfail(
+    current_platform.is_rocm(),
+    reason="Qwen2.5-VL dependency xformers incompatible with ROCm",
+)
+@pytest.mark.skipif(
+    Version(TRANSFORMERS_VERSION) < Version("4.49.0"),
+    reason="Qwen2.5-VL require transformers version no lower than 4.49.0",
+)
+def test_qwen25vl_lora(qwen25vl_lora_files):
+    """Test Qwen 2.5 VL model with LoRA"""
+    config = TestConfig(model_path=QWEN25VL_MODEL_PATH,
+                        lora_path=qwen25vl_lora_files)
+    tester = Qwen2VLTester(config)
+
+    # Test with different LoRA IDs
+    for lora_id in [1, 2]:
+        tester.run_test(TEST_IMAGES,
+                        expected_outputs=EXPECTED_OUTPUTS,
+                        lora_id=lora_id)

vllm/model_executor/models/qwen2_5_vl.py

@@ -734,16 +734,17 @@ class Qwen2_5_VLForConditionalGeneration(nn.Module, SupportsMultiModal,
             "up_proj",
         ],
     }
-    # LoRA specific attributes, TODO: double check
+    # LoRA specific attributes
     supported_lora_modules = [
+        # language model
         "qkv_proj",
         "o_proj",
         "gate_up_proj",
-        "down_proj",
+        "down_proj",  # Same name with vision encoder
-        "gate_proj"
-        "up_proj",
         # vision tower
         "qkv",
+        "gate_proj",
+        "up_proj",
         "attn.proj",  # Distinguish patch_embed.proj
         "fc1",
         "fc2",
@@ -751,6 +752,7 @@ class Qwen2_5_VLForConditionalGeneration(nn.Module, SupportsMultiModal,
         "mlp.0",
         "mlp.2"
     ]
+
     embedding_modules = {}
     embedding_padding_modules = []