Instructions to use unsloth/Qwen2.5-VL-3B-Instruct-unsloth-bnb-4bit with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use unsloth/Qwen2.5-VL-3B-Instruct-unsloth-bnb-4bit with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("image-text-to-text", model="unsloth/Qwen2.5-VL-3B-Instruct-unsloth-bnb-4bit")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("unsloth/Qwen2.5-VL-3B-Instruct-unsloth-bnb-4bit")
model = AutoModelForMultimodalLM.from_pretrained("unsloth/Qwen2.5-VL-3B-Instruct-unsloth-bnb-4bit", device_map="auto")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
inputs = processor.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use unsloth/Qwen2.5-VL-3B-Instruct-unsloth-bnb-4bit with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "unsloth/Qwen2.5-VL-3B-Instruct-unsloth-bnb-4bit"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "unsloth/Qwen2.5-VL-3B-Instruct-unsloth-bnb-4bit",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/unsloth/Qwen2.5-VL-3B-Instruct-unsloth-bnb-4bit

SGLang

How to use unsloth/Qwen2.5-VL-3B-Instruct-unsloth-bnb-4bit with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "unsloth/Qwen2.5-VL-3B-Instruct-unsloth-bnb-4bit" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "unsloth/Qwen2.5-VL-3B-Instruct-unsloth-bnb-4bit",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "unsloth/Qwen2.5-VL-3B-Instruct-unsloth-bnb-4bit" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "unsloth/Qwen2.5-VL-3B-Instruct-unsloth-bnb-4bit",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Unsloth Desktop
Docker Model Runner
How to use unsloth/Qwen2.5-VL-3B-Instruct-unsloth-bnb-4bit with Docker Model Runner:
```
docker model run hf.co/unsloth/Qwen2.5-VL-3B-Instruct-unsloth-bnb-4bit
```

danielhanchen commited on Feb 20, 2025

Commit

e954699

verified ·

1 Parent(s): 9e7f59c

Add files using upload-large-folder tool

Browse files

Files changed (5) hide show

config.json +3 -9
generation_config.json +1 -1
model.safetensors +2 -2
preprocessor_config.json +3 -3
tokenizer_config.json +1 -1

config.json CHANGED Viewed

@@ -50,13 +50,9 @@
       "visual.blocks.8.attn",
       "visual.blocks.8.mlp",
       "visual.blocks.9.attn",
-      "visual.blocks.9.mlp",
-      "visual.blocks.10.attn",
-      "visual.blocks.10.mlp",
       "visual.blocks.11.attn",
-      "visual.blocks.12.attn",
       "visual.blocks.12.mlp",
-      "visual.blocks.13.attn",
       "visual.blocks.13.mlp",
       "visual.blocks.14.attn",
       "visual.blocks.14.mlp",
@@ -64,11 +60,9 @@
       "visual.blocks.15.mlp",
       "visual.blocks.16.mlp",
       "visual.blocks.17.mlp",
-      "visual.blocks.18.mlp",
       "visual.blocks.21.mlp",
       "model.layers.1.mlp",
-      "model.layers.2.mlp",
-      "visual.blocks.31.mlp.down_proj"
     ],
     "llm_int8_threshold": 6.0,
     "load_in_4bit": true,
@@ -89,7 +83,7 @@
   "sliding_window": 32768,
   "tie_word_embeddings": true,
   "torch_dtype": "bfloat16",
-  "transformers_version": "4.49.0.dev0",
   "unsloth_fixed": true,
   "use_cache": true,
   "use_sliding_window": false,

       "visual.blocks.8.attn",
       "visual.blocks.8.mlp",
       "visual.blocks.9.attn",
       "visual.blocks.11.attn",
+      "visual.blocks.11.mlp",
       "visual.blocks.12.mlp",
       "visual.blocks.13.mlp",
       "visual.blocks.14.attn",
       "visual.blocks.14.mlp",
       "visual.blocks.15.mlp",
       "visual.blocks.16.mlp",
       "visual.blocks.17.mlp",
       "visual.blocks.21.mlp",
       "model.layers.1.mlp",
+      "visual.blocks.31.mlp"
     ],
     "llm_int8_threshold": 6.0,
     "load_in_4bit": true,
   "sliding_window": 32768,
   "tie_word_embeddings": true,
   "torch_dtype": "bfloat16",
+  "transformers_version": "4.50.0.dev0",
   "unsloth_fixed": true,
   "use_cache": true,
   "use_sliding_window": false,

generation_config.json CHANGED Viewed

@@ -11,5 +11,5 @@
   "temperature": 0.1,
   "top_k": 1,
   "top_p": 0.001,
-  "transformers_version": "4.49.0.dev0"
 }

   "temperature": 0.1,
   "top_k": 1,
   "top_p": 0.001,
+  "transformers_version": "4.50.0.dev0"
 }

model.safetensors CHANGED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1485b5e218a22fa049dbf50f2edb9507169919ed321a5a993da60b1cf04a74d6
-size 3190402195

 version https://git-lfs.github.com/spec/v1
+oid sha256:51a6325439277b0f7489ce22ba924841cf6a785675221d0b3df81f763e925d45
+size 3034882794

preprocessor_config.json CHANGED Viewed

@@ -8,7 +8,7 @@
     0.4578275,
     0.40821073
   ],
-  "image_processor_type": "Qwen2_5_VLImageProcessor",
   "image_std": [
     0.26862954,
     0.26130258,
@@ -22,8 +22,8 @@
   "resample": 3,
   "rescale_factor": 0.00392156862745098,
   "size": {
-    "max_pixels": 12845056,
-    "min_pixels": 3136
   },
   "temporal_patch_size": 2
 }

     0.4578275,
     0.40821073
   ],
+  "image_processor_type": "Qwen2VLImageProcessor",
   "image_std": [
     0.26862954,
     0.26130258,
   "resample": 3,
   "rescale_factor": 0.00392156862745098,
   "size": {
+    "longest_edge": 12845056,
+    "shortest_edge": 3136
   },
   "temporal_patch_size": 2
 }

tokenizer_config.json CHANGED Viewed

@@ -200,7 +200,7 @@
   "eos_token": "<|im_end|>",
   "errors": "replace",
   "extra_special_tokens": {},
-  "model_max_length": 131072,
   "pad_token": "<|vision_pad|>",
   "padding_side": "left",
   "processor_class": "Qwen2_5_VLProcessor",

   "eos_token": "<|im_end|>",
   "errors": "replace",
   "extra_special_tokens": {},
+  "model_max_length": 32768,
   "pad_token": "<|vision_pad|>",
   "padding_side": "left",
   "processor_class": "Qwen2_5_VLProcessor",