Instructions to use lkhl/VideoLLaMA3-2B-Image-HF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use lkhl/VideoLLaMA3-2B-Image-HF with Transformers:

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

pipe = pipeline("image-text-to-text", model="lkhl/VideoLLaMA3-2B-Image-HF")
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("lkhl/VideoLLaMA3-2B-Image-HF")
model = AutoModelForMultimodalLM.from_pretrained("lkhl/VideoLLaMA3-2B-Image-HF", 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 lkhl/VideoLLaMA3-2B-Image-HF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "lkhl/VideoLLaMA3-2B-Image-HF"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "lkhl/VideoLLaMA3-2B-Image-HF",
		"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/lkhl/VideoLLaMA3-2B-Image-HF

SGLang

How to use lkhl/VideoLLaMA3-2B-Image-HF 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 "lkhl/VideoLLaMA3-2B-Image-HF" \
    --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": "lkhl/VideoLLaMA3-2B-Image-HF",
		"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 "lkhl/VideoLLaMA3-2B-Image-HF" \
        --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": "lkhl/VideoLLaMA3-2B-Image-HF",
		"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 lkhl/VideoLLaMA3-2B-Image-HF with Docker Model Runner:
```
docker model run hf.co/lkhl/VideoLLaMA3-2B-Image-HF
```

lkhl commited on Aug 29, 2025

Commit

e4996c0

verified ·

1 Parent(s): 36a63f8

Upload processor

Browse files

Files changed (3) hide show

chat_template.jinja +5 -7
preprocessor_config.json +8 -2
video_preprocessor_config.json +10 -6

chat_template.jinja CHANGED Viewed

@@ -17,15 +17,13 @@ You are VideoLLaMA3 created by Alibaba DAMO Academy, a helpful assistant to help
         {{- message['content'] -}}
     {% else %}
         {% for content in message['content'] %}
-            {% if content is string %}
-                {{- content -}}
-            {% elif content['type'] == 'text' or 'text' in content %}
                 {{- content['text'] -}}
-            {% elif content['type'] == 'image' or 'image' in content %}
-                {{- image_token + '
 ' -}}
-            {% elif content['type'] == 'video' or 'video' in content %}
-                {{- video_token + '
 ' -}}
             {% endif %}
         {% endfor %}

         {{- message['content'] -}}
     {% else %}
         {% for content in message['content'] %}
+            {% if content['type'] == 'text' %}
                 {{- content['text'] -}}
+            {% elif content['type'] == 'image' %}
+                {{- '<|image_pad|>' + '
 ' -}}
+            {% elif content['type'] == 'video' %}
+                {{- '<|video_pad|>' + '
 ' -}}
             {% endif %}
         {% endfor %}

preprocessor_config.json CHANGED Viewed

@@ -8,18 +8,24 @@
     0.5,
     0.5
   ],
-  "image_merge_size": 1,
   "image_processor_type": "Videollama3ImageProcessor",
   "image_std": [
     0.5,
     0.5,
     0.5
   ],
   "max_tokens": 16384,
   "min_tokens": 16,
   "patch_size": 14,
   "processor_class": "Videollama3Processor",
   "resample": 3,
   "rescale_factor": 0.00392156862745098,
-  "video_merge_size": 1
 }

     0.5,
     0.5
   ],
   "image_processor_type": "Videollama3ImageProcessor",
   "image_std": [
     0.5,
     0.5,
     0.5
   ],
+  "max_pixels": 3211264,
   "max_tokens": 16384,
+  "merge_size": 1,
+  "min_pixels": 3136,
   "min_tokens": 16,
   "patch_size": 14,
   "processor_class": "Videollama3Processor",
   "resample": 3,
   "rescale_factor": 0.00392156862745098,
+  "size": {
+    "longest_edge": 1003520,
+    "shortest_edge": 3136
+  },
+  "temporal_patch_size": 1
 }

video_preprocessor_config.json CHANGED Viewed

@@ -9,14 +9,13 @@
   "do_pad": null,
   "do_rescale": true,
   "do_resize": true,
-  "do_sample_frames": true,
   "fps": null,
   "image_mean": [
     0.5,
     0.5,
     0.5
   ],
-  "image_merge_size": null,
   "image_std": [
     0.5,
     0.5,
@@ -24,16 +23,21 @@
   ],
   "input_data_format": null,
   "max_frames": 180,
-  "max_tokens": 16384,
-  "min_tokens": 16,
   "num_frames": null,
   "patch_size": 14,
   "processor_class": "Videollama3Processor",
   "resample": 3,
   "rescale_factor": 0.00392156862745098,
-  "size": null,
   "size_divisor": null,
-  "video_merge_size": 2,
   "video_metadata": null,
   "video_processor_type": "Videollama3VideoProcessor"
 }

   "do_pad": null,
   "do_rescale": true,
   "do_resize": true,
+  "do_sample_frames": false,
   "fps": null,
   "image_mean": [
     0.5,
     0.5,
     0.5
   ],
   "image_std": [
     0.5,
     0.5,
   ],
   "input_data_format": null,
   "max_frames": 180,
+  "max_pixels": 12845056,
+  "merge_size": 2,
+  "min_frames": 4,
+  "min_pixels": 12544,
   "num_frames": null,
   "patch_size": 14,
   "processor_class": "Videollama3Processor",
   "resample": 3,
   "rescale_factor": 0.00392156862745098,
+  "size": {
+    "longest_edge": 12845056,
+    "shortest_edge": 12544
+  },
   "size_divisor": null,
+  "temporal_patch_size": 1,
   "video_metadata": null,
   "video_processor_type": "Videollama3VideoProcessor"
 }