Instructions to use sunjuice/Molmo2-8B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use sunjuice/Molmo2-8B with Transformers:

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

pipe = pipeline("image-text-to-text", model="sunjuice/Molmo2-8B", trust_remote_code=True)
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 AutoModelForImageTextToText
model = AutoModelForImageTextToText.from_pretrained("sunjuice/Molmo2-8B", trust_remote_code=True, device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use sunjuice/Molmo2-8B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "sunjuice/Molmo2-8B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "sunjuice/Molmo2-8B",
		"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/sunjuice/Molmo2-8B

SGLang

How to use sunjuice/Molmo2-8B 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 "sunjuice/Molmo2-8B" \
    --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": "sunjuice/Molmo2-8B",
		"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 "sunjuice/Molmo2-8B" \
        --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": "sunjuice/Molmo2-8B",
		"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 sunjuice/Molmo2-8B with Docker Model Runner:
```
docker model run hf.co/sunjuice/Molmo2-8B
```

sunjuice commited on Apr 12

Commit

43189e0

1 Parent(s): c969eb1

moving past keyvalue dynamically

Browse files

Files changed (1) hide show

modeling_molmo2.py +15 -24

modeling_molmo2.py CHANGED Viewed

@@ -1054,31 +1054,22 @@ class Molmo2TextModel(Molmo2PreTrainedModel):
                 position_embeddings_i = position_embeddings
             block_device = next(decoder_block.parameters()).device
-            if hidden_states.device != block_device:
-                hidden_states = hidden_states.to(block_device)
-            if causal_mask_mapping is not None and causal_mask_mapping.device != block_device:
-                causal_mask_mapping = causal_mask_mapping.to(block_device)
-            if position_ids is not None and position_ids.device != block_device:
-                position_ids = position_ids.to(block_device)
-            if position_embeddings_i is not None and position_embeddings_i.device != block_device:
-                position_embeddings_i = position_embeddings_i.to(block_device)
-            if cache_position is not None and cache_position.device != block_device:
-                cache_position = cache_position.to(block_device)
-            def move_past_key_values(past_key_values, device):
-                if past_key_values is None:
-                    return None
-                return tuple(
-                    tuple(p.to(device) for p in layer)
-                    for layer in past_key_values
-                )
-            past_key_values = move_past_key_values(past_key_values, block_device)
-            position_embeddings_i = position_embeddings_i.to(block_device)
             layer_outputs = decoder_block(
                 hidden_states,

                 position_embeddings_i = position_embeddings
             block_device = next(decoder_block.parameters()).device
+            def move_if_needed(x, device):
+                if x is None:
+                    return x
+                if isinstance(x, tuple):
+                    return tuple(move_if_needed(xx, device) for xx in x)
+                if hasattr(x, "device") and x.device != device:
+                    return x.to(device)
+                return x
+            hidden_states = move_if_needed(hidden_states, block_device)
+            causal_mask_mapping = move_if_needed(causal_mask_mapping, block_device)
+            position_ids = move_if_needed(position_ids, block_device)
+            position_embeddings_i = move_if_needed(position_embeddings_i, block_device)
+            cache_position = move_if_needed(cache_position, block_device)
+            past_key_values = move_if_needed(past_key_values, block_device)
             layer_outputs = decoder_block(
                 hidden_states,