Instructions to use nyu-visionx/cambrian-phi3-3b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use nyu-visionx/cambrian-phi3-3b with Transformers:

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

pipe = pipeline("text-generation", model="nyu-visionx/cambrian-phi3-3b", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("nyu-visionx/cambrian-phi3-3b", trust_remote_code=True, dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use nyu-visionx/cambrian-phi3-3b with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "nyu-visionx/cambrian-phi3-3b"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "nyu-visionx/cambrian-phi3-3b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/nyu-visionx/cambrian-phi3-3b

SGLang

How to use nyu-visionx/cambrian-phi3-3b 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 "nyu-visionx/cambrian-phi3-3b" \
    --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": "nyu-visionx/cambrian-phi3-3b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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 "nyu-visionx/cambrian-phi3-3b" \
        --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": "nyu-visionx/cambrian-phi3-3b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use nyu-visionx/cambrian-phi3-3b with Docker Model Runner:
```
docker model run hf.co/nyu-visionx/cambrian-phi3-3b
```

cambrian-phi3-3b

File size: 2,191 Bytes

{
  "_name_or_path": "microsoft/Phi-3-mini-4k-instruct",
  "architectures": [
    "CambrianPhi3ForCausalLM"
  ],
  "attention_bias": false,
  "attention_dropout": 0.0,
  "auto_map": {
    "AutoConfig": "microsoft/Phi-3-mini-4k-instruct--configuration_phi3.Phi3Config",
    "AutoModelForCausalLM": "microsoft/Phi-3-mini-4k-instruct--modeling_phi3.Phi3ForCausalLM"
  },
  "bos_token_id": 1,
  "connector_depth": 3,
  "connector_only": false,
  "do_sample": true,
  "embd_pdrop": 0.0,
  "eos_token_id": 32007,
  "freeze_mm_mlp_adapter": false,
  "hidden_act": "silu",
  "hidden_size": 3072,
  "image_aspect_ratio": "pad",
  "image_position": 13,
  "image_token_len": 576,
  "initializer_range": 0.02,
  "intermediate_size": 8192,
  "max_position_embeddings": 4096,
  "mm_projector_lr": null,
  "mm_projector_type": "sva",
  "mm_use_im_patch_token": false,
  "mm_use_im_start_end": false,
  "mm_vision_sampler_lr": null,
  "mm_vision_select_feature": "patch",
  "mm_vision_select_layer": -2,
  "mm_vision_tower_aux_list": [
    "siglip/CLIP-ViT-SO400M-14-384",
    "openai/clip-vit-large-patch14-336",
    "facebook/dinov2-giant-res378",
    "clip-convnext-XXL-multi-stage"
  ],
  "mm_vision_tower_aux_token_len_list": [
    576,
    576,
    576,
    9216
  ],
  "mm_vision_tower_lr": null,
  "model_type": "cambrian_phi3",
  "num_attention_heads": 32,
  "num_hidden_layers": 32,
  "num_key_value_heads": 32,
  "num_of_vision_sampler_layers": 10,
  "num_query_group": 1,
  "original_max_position_embeddings": 4096,
  "pad_token_id": 32000,
  "query_num_list": [
    576
  ],
  "resid_pdrop": 0.0,
  "rms_norm_eps": 1e-05,
  "rope_scaling": null,
  "rope_theta": 10000.0,
  "sliding_window": 2047,
  "start_of_vision_sampler_layers": 0,
  "stride_of_vision_sampler_layers": 3,
  "tie_word_embeddings": false,
  "tokenizer_model_max_length": 2048,
  "tokenizer_padding_side": "right",
  "torch_dtype": "float32",
  "transformers_version": "4.39.3",
  "tune_mm_mlp_adapter": false,
  "unfreeze_mm_vision_tower": false,
  "use_cache": true,
  "use_mm_proj": true,
  "vision_hidden_size": 1024,
  "vision_tower_aux_token_len_list": [
    576,
    576,
    576,
    9216
  ],
  "vocab_size": 32064
}