Instructions to use unsloth/Qwen2.5-VL-32B-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-32B-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-32B-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, AutoModelForImageTextToText

processor = AutoProcessor.from_pretrained("unsloth/Qwen2.5-VL-32B-Instruct-unsloth-bnb-4bit")
model = AutoModelForImageTextToText.from_pretrained("unsloth/Qwen2.5-VL-32B-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?"}
        ]
    },
]
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

vLLM

How to use unsloth/Qwen2.5-VL-32B-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-32B-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-32B-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-32B-Instruct-unsloth-bnb-4bit

SGLang

How to use unsloth/Qwen2.5-VL-32B-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-32B-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-32B-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-32B-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-32B-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 Studio new

How to use unsloth/Qwen2.5-VL-32B-Instruct-unsloth-bnb-4bit with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for unsloth/Qwen2.5-VL-32B-Instruct-unsloth-bnb-4bit to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for unsloth/Qwen2.5-VL-32B-Instruct-unsloth-bnb-4bit to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for unsloth/Qwen2.5-VL-32B-Instruct-unsloth-bnb-4bit to start chatting

Load model with FastModel

pip install unsloth
from unsloth import FastModel
model, tokenizer = FastModel.from_pretrained(
    model_name="unsloth/Qwen2.5-VL-32B-Instruct-unsloth-bnb-4bit",
    max_seq_length=2048,
)

Docker Model Runner
How to use unsloth/Qwen2.5-VL-32B-Instruct-unsloth-bnb-4bit with Docker Model Runner:
```
docker model run hf.co/unsloth/Qwen2.5-VL-32B-Instruct-unsloth-bnb-4bit
```

Doesn't work with vLLM

by Eryk-Chmielewski - opened Mar 27, 2025

Discussion

Eryk-Chmielewski

Mar 27, 2025

Hi, how to run this model using vLLM?

v1 engine return error that bitsandbytes is not supported in v1

v0 returns error
docker run --runtime nvidia --gpus all -d --name vllm-Qwen-VL-32b-v2 --restart unless-stopped -v ~/.cache/huggingface:/root/.cache/huggingface -e VLLM_USE_V1=0 -p 8000:8000 vllm/vllm-openai:v0.8.2 --model unsloth/Qwen2.5-VL-32B-Instruct-unsloth-bnb-4bit --dtype auto --served-model-name llm --max-model-len 27740 --max-num-batched-tokens 27740

INFO 03-27 14:06:32 [weight_utils.py:265] Using model weights format ['*.safetensors']
Loading safetensors checkpoint shards: 0% Completed | 0/5 [00:00<?, ?it/s]
ERROR 03-27 14:06:33 [engine.py:448] 'layers.54.mlp.down_proj.weight.absmax'
ERROR 03-27 14:06:33 [engine.py:448] Traceback (most recent call last):
ERROR 03-27 14:06:33 [engine.py:448] File "/usr/local/lib/python3.12/dist-packages/vllm/engine/multiprocessing/engine.py", line 436, in run_mp_engine
ERROR 03-27 14:06:33 [engine.py:448] engine = MQLLMEngine.from_vllm_config(
ERROR 03-27 14:06:33 [engine.py:448] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 03-27 14:06:33 [engine.py:448] File "/usr/local/lib/python3.12/dist-packages/vllm/engine/multiprocessing/engine.py", line 128, in from_vllm_config
ERROR 03-27 14:06:33 [engine.py:448] return cls(
ERROR 03-27 14:06:33 [engine.py:448] ^^^^
ERROR 03-27 14:06:33 [engine.py:448] File "/usr/local/lib/python3.12/dist-packages/vllm/engine/multiprocessing/engine.py", line 82, in init
ERROR 03-27 14:06:33 [engine.py:448] self.engine = LLMEngine(*args, **kwargs)
ERROR 03-27 14:06:33 [engine.py:448] ^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 03-27 14:06:33 [engine.py:448] File "/usr/local/lib/python3.12/dist-packages/vllm/engine/llm_engine.py", line 280, in init
ERROR 03-27 14:06:33 [engine.py:448] self.model_executor = executor_class(vllm_config=vllm_config, )
ERROR 03-27 14:06:33 [engine.py:448] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 03-27 14:06:33 [engine.py:448] File "/usr/local/lib/python3.12/dist-packages/vllm/executor/executor_base.py", line 52, in init
ERROR 03-27 14:06:33 [engine.py:448] self._init_executor()
ERROR 03-27 14:06:33 [engine.py:448] File "/usr/local/lib/python3.12/dist-packages/vllm/executor/uniproc_executor.py", line 47, in _init_executor
ERROR 03-27 14:06:33 [engine.py:448] self.collective_rpc("load_model")
ERROR 03-27 14:06:33 [engine.py:448] File "/usr/local/lib/python3.12/dist-packages/vllm/executor/uniproc_executor.py", line 56, in collective_rpc
ERROR 03-27 14:06:33 [engine.py:448] answer = run_method(self.driver_worker, method, args, kwargs)
ERROR 03-27 14:06:33 [engine.py:448] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 03-27 14:06:33 [engine.py:448] File "/usr/local/lib/python3.12/dist-packages/vllm/utils.py", line 2255, in run_method
ERROR 03-27 14:06:33 [engine.py:448] return func(*args, **kwargs)
ERROR 03-27 14:06:33 [engine.py:448] ^^^^^^^^^^^^^^^^^^^^^
ERROR 03-27 14:06:33 [engine.py:448] File "/usr/local/lib/python3.12/dist-packages/vllm/worker/worker.py", line 183, in load_model
ERROR 03-27 14:06:33 [engine.py:448] self.model_runner.load_model()
ERROR 03-27 14:06:33 [engine.py:448] File "/usr/local/lib/python3.12/dist-packages/vllm/worker/model_runner.py", line 1113, in load_model
ERROR 03-27 14:06:33 [engine.py:448] self.model = get_model(vllm_config=self.vllm_config)
ERROR 03-27 14:06:33 [engine.py:448] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 03-27 14:06:33 [engine.py:448] File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/model_loader/init.py", line 14, in get_model
Process SpawnProcess-1:
ERROR 03-27 14:06:33 [engine.py:448] return loader.load_model(vllm_config=vllm_config)
ERROR 03-27 14:06:33 [engine.py:448] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 03-27 14:06:33 [engine.py:448] File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/model_loader/loader.py", line 444, in load_model
ERROR 03-27 14:06:33 [engine.py:448] loaded_weights = model.load_weights(
ERROR 03-27 14:06:33 [engine.py:448] ^^^^^^^^^^^^^^^^^^^
ERROR 03-27 14:06:33 [engine.py:448] File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/qwen2_5_vl.py", line 1113, in load_weights
ERROR 03-27 14:06:33 [engine.py:448] return loader.load_weights(weights, mapper=self.hf_to_vllm_mapper)
ERROR 03-27 14:06:33 [engine.py:448] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 03-27 14:06:33 [engine.py:448] File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/utils.py", line 235, in load_weights
ERROR 03-27 14:06:33 [engine.py:448] autoloaded_weights = set(self._load_module("", self.module, weights))
ERROR 03-27 14:06:33 [engine.py:448] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 03-27 14:06:33 [engine.py:448] File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/utils.py", line 196, in _load_module
ERROR 03-27 14:06:33 [engine.py:448] yield from self._load_module(prefix,
ERROR 03-27 14:06:33 [engine.py:448] File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/utils.py", line 173, in _load_module
ERROR 03-27 14:06:33 [engine.py:448] loaded_params = module_load_weights(weights)
ERROR 03-27 14:06:33 [engine.py:448] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 03-27 14:06:33 [engine.py:448] File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/qwen2.py", line 490, in load_weights
ERROR 03-27 14:06:33 [engine.py:448] return loader.load_weights(weights)
ERROR 03-27 14:06:33 [engine.py:448] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 03-27 14:06:33 [engine.py:448] File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/utils.py", line 235, in load_weights
ERROR 03-27 14:06:33 [engine.py:448] autoloaded_weights = set(self._load_module("", self.module, weights))
ERROR 03-27 14:06:33 [engine.py:448] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 03-27 14:06:33 [engine.py:448] File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/utils.py", line 196, in _load_module
ERROR 03-27 14:06:33 [engine.py:448] yield from self._load_module(prefix,
ERROR 03-27 14:06:33 [engine.py:448] File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/utils.py", line 173, in _load_module
ERROR 03-27 14:06:33 [engine.py:448] loaded_params = module_load_weights(weights)
ERROR 03-27 14:06:33 [engine.py:448] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 03-27 14:06:33 [engine.py:448] File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/qwen2.py", line 400, in load_weights
ERROR 03-27 14:06:33 [engine.py:448] param = params_dict[name]
ERROR 03-27 14:06:33 [engine.py:448] ~~~~~~~~~~~^^^^^^
ERROR 03-27 14:06:33 [engine.py:448] KeyError: 'layers.54.mlp.down_proj.weight.absmax'

b0yan9

Mar 28, 2025

Maybe try to add these parameters? --dtype bfloat16 --load_format bitsandbytes --quantization bitsandbytes

lktinhtemp

Mar 29, 2025

Maybe try to add these parameters? --dtype bfloat16 --load_format bitsandbytes --quantization bitsandbytes

not work

Tendy

Apr 2, 2025

same issue.

Mdubbya

Apr 8, 2025

getting issue here too.

2025-04-08 12:36:26 Traceback (most recent call last):
2025-04-08 12:36:26 File "/usr/local/lib/python3.12/dist-packages/vllm/engine/multiprocessing/engine.py", line 436, in run_mp_engine
2025-04-08 12:36:26 engine = MQLLMEngine.from_vllm_config(
2025-04-08 12:36:26 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2025-04-08 12:36:26 File "/usr/local/lib/python3.12/dist-packages/vllm/engine/multiprocessing/engine.py", line 128, in from_vllm_config
2025-04-08 12:36:26 return cls(
2025-04-08 12:36:26 ^^^^
2025-04-08 12:36:26 File "/usr/local/lib/python3.12/dist-packages/vllm/engine/multiprocessing/engine.py", line 82, in init
2025-04-08 12:36:26 self.engine = LLMEngine(*args, **kwargs)
2025-04-08 12:36:26 ^^^^^^^^^^^^^^^^^^^^^^^^^^
2025-04-08 12:36:26 File "/usr/local/lib/python3.12/dist-packages/vllm/engine/llm_engine.py", line 281, in init
2025-04-08 12:36:26 self.model_executor = executor_class(vllm_config=vllm_config, )
2025-04-08 12:36:26 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2025-04-08 12:36:26 File "/usr/local/lib/python3.12/dist-packages/vllm/executor/executor_base.py", line 52, in init
2025-04-08 12:36:26 self._init_executor()
2025-04-08 12:36:26 File "/usr/local/lib/python3.12/dist-packages/vllm/executor/uniproc_executor.py", line 47, in _init_executor
2025-04-08 12:36:26 self.collective_rpc("load_model")
2025-04-08 12:36:26 File "/usr/local/lib/python3.12/dist-packages/vllm/executor/uniproc_executor.py", line 56, in collective_rpc
2025-04-08 12:36:26 answer = run_method(self.driver_worker, method, args, kwargs)
2025-04-08 12:36:26 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2025-04-08 12:36:26 File "/usr/local/lib/python3.12/dist-packages/vllm/utils.py", line 2347, in run_method
2025-04-08 12:36:26 return func(*args, **kwargs)
2025-04-08 12:36:26 ^^^^^^^^^^^^^^^^^^^^^
2025-04-08 12:36:26 File "/usr/local/lib/python3.12/dist-packages/vllm/worker/worker.py", line 183, in load_model
2025-04-08 12:36:26 self.model_runner.load_model()
2025-04-08 12:36:26 File "/usr/local/lib/python3.12/dist-packages/vllm/worker/model_runner.py", line 1113, in load_model
2025-04-08 12:36:26 self.model = get_model(vllm_config=self.vllm_config)
2025-04-08 12:36:26 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2025-04-08 12:36:26 File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/model_loader/init.py", line 14, in get_model
2025-04-08 12:36:26 return loader.load_model(vllm_config=vllm_config)
2025-04-08 12:36:26 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2025-04-08 12:36:26 File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/model_loader/loader.py", line 1280, in load_model
2025-04-08 12:36:26 self._load_weights(model_config, model)
2025-04-08 12:36:26 File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/model_loader/loader.py", line 1188, in _load_weights
2025-04-08 12:36:26 loaded_weights = model.load_weights(qweight_iterator)
2025-04-08 12:36:26 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2025-04-08 12:36:26 File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/qwen2_5_vl.py", line 1113, in load_weights
2025-04-08 12:36:26 return loader.load_weights(weights, mapper=self.hf_to_vllm_mapper)
2025-04-08 12:36:26 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2025-04-08 12:36:26 File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/utils.py", line 261, in load_weights
2025-04-08 12:36:26 autoloaded_weights = set(self._load_module("", self.module, weights))
2025-04-08 12:36:26 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2025-04-08 12:36:26 File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/utils.py", line 222, in _load_module
2025-04-08 12:36:26 yield from self._load_module(prefix,
2025-04-08 12:36:26 File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/utils.py", line 195, in _load_module
2025-04-08 12:36:26 loaded_params = module_load_weights(weights)
2025-04-08 12:36:26 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2025-04-08 12:36:26 File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/qwen2.py", line 490, in load_weights
2025-04-08 12:36:26 return loader.load_weights(weights)
2025-04-08 12:36:26 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2025-04-08 12:36:26 File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/utils.py", line 261, in load_weights
2025-04-08 12:36:26 autoloaded_weights = set(self._load_module("", self.module, weights))
2025-04-08 12:36:26 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2025-04-08 12:36:26 File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/utils.py", line 222, in _load_module
2025-04-08 12:36:26 yield from self._load_module(prefix,
2025-04-08 12:36:26 File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/utils.py", line 195, in _load_module
2025-04-08 12:36:26 loaded_params = module_load_weights(weights)
2025-04-08 12:36:26 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2025-04-08 12:36:26 File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/qwen2.py", line 403, in load_weights
2025-04-08 12:36:26 weight_loader(param, loaded_weight)
2025-04-08 12:36:26 File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/layers/linear.py", line 1228, in weight_loader
2025-04-08 12:36:26 assert param_data.shape == loaded_weight.shape
2025-04-08 12:36:26 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Eryk-Chmielewski

Apr 8, 2025

I hope it will be resolved
https://github.com/vllm-project/vllm/issues/16166

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment