Instructions to use FastFlowLM/Qwen3-VL-4B-Instruct-NPU2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use FastFlowLM/Qwen3-VL-4B-Instruct-NPU2 with Transformers:

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

pipe = pipeline("image-text-to-text", model="FastFlowLM/Qwen3-VL-4B-Instruct-NPU2")
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 AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("FastFlowLM/Qwen3-VL-4B-Instruct-NPU2")
model = AutoModelForCausalLM.from_pretrained("FastFlowLM/Qwen3-VL-4B-Instruct-NPU2")
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 = tokenizer.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(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use FastFlowLM/Qwen3-VL-4B-Instruct-NPU2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "FastFlowLM/Qwen3-VL-4B-Instruct-NPU2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "FastFlowLM/Qwen3-VL-4B-Instruct-NPU2",
		"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/FastFlowLM/Qwen3-VL-4B-Instruct-NPU2

SGLang

How to use FastFlowLM/Qwen3-VL-4B-Instruct-NPU2 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 "FastFlowLM/Qwen3-VL-4B-Instruct-NPU2" \
    --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": "FastFlowLM/Qwen3-VL-4B-Instruct-NPU2",
		"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 "FastFlowLM/Qwen3-VL-4B-Instruct-NPU2" \
        --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": "FastFlowLM/Qwen3-VL-4B-Instruct-NPU2",
		"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"
						}
					}
				]
			}
		]
	}'

Docker Model Runner
How to use FastFlowLM/Qwen3-VL-4B-Instruct-NPU2 with Docker Model Runner:
```
docker model run hf.co/FastFlowLM/Qwen3-VL-4B-Instruct-NPU2
```

FastFlowLM commited on Oct 29, 2025

Commit

af67e78

verified ·

1 Parent(s): abfdf5c

feat: upload vision part

Browse files

Files changed (5) hide show

.gitattributes +3 -0
config.json +37 -34
vision_attn.xclbin +3 -0
vision_mm.xclbin +3 -0
vision_weight.q4nx +3 -0

.gitattributes CHANGED Viewed

@@ -40,3 +40,6 @@ lm_head.xclbin filter=lfs diff=lfs merge=lfs -text
 mm.xclbin filter=lfs diff=lfs merge=lfs -text
 model.q4nx filter=lfs diff=lfs merge=lfs -text
 tokenizer.json filter=lfs diff=lfs merge=lfs -text

 mm.xclbin filter=lfs diff=lfs merge=lfs -text
 model.q4nx filter=lfs diff=lfs merge=lfs -text
 tokenizer.json filter=lfs diff=lfs merge=lfs -text
+vision_attn.xclbin filter=lfs diff=lfs merge=lfs -text
+vision_mm.xclbin filter=lfs diff=lfs merge=lfs -text
+vision_weight.q4nx filter=lfs diff=lfs merge=lfs -text

config.json CHANGED Viewed

@@ -1,34 +1,37 @@
-{
-    "model_type": "qwen3",
-    "vocab_size": 151936,
-    "max_position_embeddings": 40960,
-    "hidden_size": 2560,
-    "intermediate_size": 9728,
-    "num_hidden_layers": 36,
-    "num_attention_heads": 32,
-    "use_sliding_window": false,
-    "sliding_window": null,
-    "max_window_layers": 36,
-    "num_key_value_heads": 8,
-    "head_dim": 128,
-    "hidden_act": "silu",
-    "initializer_range": 0.02,
-    "rms_norm_eps": 1e-06,
-    "use_cache": true,
-    "rope_theta": 1000000,
-    "rope_scaling": null,
-    "attention_bias": false,
-    "attention_dropout": 0.0,
-  "architectures": [
-    "LlamaForCausalLM"
-  ],
-  "torch_dtype": "bfloat16",
-  "transformers_version": "4.45.0.dev0",
-  "use_cache": true,
-  "addr_qk": 9216,
-  "addr_kv": 34816,
-  "addr_l_begin_mha": 54272,
-  "addr_l_end_mha": 25600,
-  "addr_kk": 45056,
-  "flm_version": "0.9.6"
-}

+{
+    "model_type": "qwen3",
+    "vocab_size": 151936,
+    "max_position_embeddings": 40960,
+    "hidden_size": 2560,
+    "intermediate_size": 9728,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "use_sliding_window": false,
+    "sliding_window": null,
+    "max_window_layers": 36,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_act": "silu",
+    "initializer_range": 0.02,
+    "rms_norm_eps": 1e-06,
+    "use_cache": true,
+    "rope_theta": 1000000,
+    "rope_scaling": null,
+    "attention_bias": false,
+    "attention_dropout": 0.0,
+  "architectures": [
+    "LlamaForCausalLM"
+  ],
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.45.0.dev0",
+  "use_cache": true,
+  "addr_qk": 9216,
+  "addr_kv": 34816,
+  "addr_l_begin_mha": 54272,
+  "addr_l_end_mha": 25600,
+  "addr_kk": 45056,
+  "flm_version": "0.9.16",
+  "vision_model_weight": "vision_weight.q4nx",
+  "vision_mm_engine_xclbin_name": "vision_mm.xclbin",
+  "vision_mha_engine_xclbin_name":"vision_attn.xclbin"
+}

vision_attn.xclbin ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:6abbd4157ea07d048bfb1b717a9c6f72044c07d7b60169bd921a4d9a9acfdda0
+size 242043

vision_mm.xclbin ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:1a419730401233f197ee8fbd1a0ee50339d25975a244325a9ca79fa44228de6a
+size 580539

vision_weight.q4nx ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:81ec06fa8b590e8a307e11eea29231e715fc83033ac88f2daba6448b878b36c1
+size 830730632