Instructions to use aleada/Llama-3.2-11B-Vision-Instruct-W4A16 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use aleada/Llama-3.2-11B-Vision-Instruct-W4A16 with Transformers:

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

pipe = pipeline("image-text-to-text", model="aleada/Llama-3.2-11B-Vision-Instruct-W4A16")
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("aleada/Llama-3.2-11B-Vision-Instruct-W4A16")
model = AutoModelForMultimodalLM.from_pretrained("aleada/Llama-3.2-11B-Vision-Instruct-W4A16", 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 aleada/Llama-3.2-11B-Vision-Instruct-W4A16 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "aleada/Llama-3.2-11B-Vision-Instruct-W4A16"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "aleada/Llama-3.2-11B-Vision-Instruct-W4A16",
		"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/aleada/Llama-3.2-11B-Vision-Instruct-W4A16

SGLang

How to use aleada/Llama-3.2-11B-Vision-Instruct-W4A16 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 "aleada/Llama-3.2-11B-Vision-Instruct-W4A16" \
    --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": "aleada/Llama-3.2-11B-Vision-Instruct-W4A16",
		"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 "aleada/Llama-3.2-11B-Vision-Instruct-W4A16" \
        --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": "aleada/Llama-3.2-11B-Vision-Instruct-W4A16",
		"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 aleada/Llama-3.2-11B-Vision-Instruct-W4A16 with Docker Model Runner:
```
docker model run hf.co/aleada/Llama-3.2-11B-Vision-Instruct-W4A16
```

Llama-3.2-11B-Vision-Instruct-W4A16 / config.json

aleada

Upload config.json

e7f41b9 verified about 2 months ago

Raw

History Blame Contribute Delete

19.6 kB

	{
	"architectures": [
	"MllamaForConditionalGeneration"
	],
	"dtype": "bfloat16",
	"image_token_index": 128256,
	"model_type": "mllama",
	"quantization_config": {
	"config_groups": {
	"group_0": {
	"format": "pack-quantized",
	"input_activations": null,
	"output_activations": null,
	"targets": [
	"Linear"
	],
	"weights": {
	"actorder": "static",
	"block_structure": null,
	"dynamic": false,
	"group_size": 128,
	"num_bits": 4,
	"observer": "memoryless_minmax",
	"observer_kwargs": {},
	"scale_dtype": null,
	"strategy": "group",
	"symmetric": true,
	"type": "int",
	"zp_dtype": null
	}
	}
	},
	"format": "pack-quantized",
	"global_compression_ratio": null,
	"ignore": [
	"model.vision_model.transformer.layers.0.self_attn.q_proj",
	"model.vision_model.transformer.layers.0.self_attn.k_proj",
	"model.vision_model.transformer.layers.0.self_attn.v_proj",
	"model.vision_model.transformer.layers.0.self_attn.o_proj",
	"model.vision_model.transformer.layers.0.mlp.fc1",
	"model.vision_model.transformer.layers.0.mlp.fc2",
	"model.vision_model.transformer.layers.1.self_attn.q_proj",
	"model.vision_model.transformer.layers.1.self_attn.k_proj",
	"model.vision_model.transformer.layers.1.self_attn.v_proj",
	"model.vision_model.transformer.layers.1.self_attn.o_proj",
	"model.vision_model.transformer.layers.1.mlp.fc1",
	"model.vision_model.transformer.layers.1.mlp.fc2",
	"model.vision_model.transformer.layers.2.self_attn.q_proj",
	"model.vision_model.transformer.layers.2.self_attn.k_proj",
	"model.vision_model.transformer.layers.2.self_attn.v_proj",
	"model.vision_model.transformer.layers.2.self_attn.o_proj",
	"model.vision_model.transformer.layers.2.mlp.fc1",
	"model.vision_model.transformer.layers.2.mlp.fc2",
	"model.vision_model.transformer.layers.3.self_attn.q_proj",
	"model.vision_model.transformer.layers.3.self_attn.k_proj",
	"model.vision_model.transformer.layers.3.self_attn.v_proj",
	"model.vision_model.transformer.layers.3.self_attn.o_proj",
	"model.vision_model.transformer.layers.3.mlp.fc1",
	"model.vision_model.transformer.layers.3.mlp.fc2",
	"model.vision_model.transformer.layers.4.self_attn.q_proj",
	"model.vision_model.transformer.layers.4.self_attn.k_proj",
	"model.vision_model.transformer.layers.4.self_attn.v_proj",
	"model.vision_model.transformer.layers.4.self_attn.o_proj",
	"model.vision_model.transformer.layers.4.mlp.fc1",
	"model.vision_model.transformer.layers.4.mlp.fc2",
	"model.vision_model.transformer.layers.5.self_attn.q_proj",
	"model.vision_model.transformer.layers.5.self_attn.k_proj",
	"model.vision_model.transformer.layers.5.self_attn.v_proj",
	"model.vision_model.transformer.layers.5.self_attn.o_proj",
	"model.vision_model.transformer.layers.5.mlp.fc1",
	"model.vision_model.transformer.layers.5.mlp.fc2",
	"model.vision_model.transformer.layers.6.self_attn.q_proj",
	"model.vision_model.transformer.layers.6.self_attn.k_proj",
	"model.vision_model.transformer.layers.6.self_attn.v_proj",
	"model.vision_model.transformer.layers.6.self_attn.o_proj",
	"model.vision_model.transformer.layers.6.mlp.fc1",
	"model.vision_model.transformer.layers.6.mlp.fc2",
	"model.vision_model.transformer.layers.7.self_attn.q_proj",
	"model.vision_model.transformer.layers.7.self_attn.k_proj",
	"model.vision_model.transformer.layers.7.self_attn.v_proj",
	"model.vision_model.transformer.layers.7.self_attn.o_proj",
	"model.vision_model.transformer.layers.7.mlp.fc1",
	"model.vision_model.transformer.layers.7.mlp.fc2",
	"model.vision_model.transformer.layers.8.self_attn.q_proj",
	"model.vision_model.transformer.layers.8.self_attn.k_proj",
	"model.vision_model.transformer.layers.8.self_attn.v_proj",
	"model.vision_model.transformer.layers.8.self_attn.o_proj",
	"model.vision_model.transformer.layers.8.mlp.fc1",
	"model.vision_model.transformer.layers.8.mlp.fc2",
	"model.vision_model.transformer.layers.9.self_attn.q_proj",
	"model.vision_model.transformer.layers.9.self_attn.k_proj",
	"model.vision_model.transformer.layers.9.self_attn.v_proj",
	"model.vision_model.transformer.layers.9.self_attn.o_proj",
	"model.vision_model.transformer.layers.9.mlp.fc1",
	"model.vision_model.transformer.layers.9.mlp.fc2",
	"model.vision_model.transformer.layers.10.self_attn.q_proj",
	"model.vision_model.transformer.layers.10.self_attn.k_proj",
	"model.vision_model.transformer.layers.10.self_attn.v_proj",
	"model.vision_model.transformer.layers.10.self_attn.o_proj",
	"model.vision_model.transformer.layers.10.mlp.fc1",
	"model.vision_model.transformer.layers.10.mlp.fc2",
	"model.vision_model.transformer.layers.11.self_attn.q_proj",
	"model.vision_model.transformer.layers.11.self_attn.k_proj",
	"model.vision_model.transformer.layers.11.self_attn.v_proj",
	"model.vision_model.transformer.layers.11.self_attn.o_proj",
	"model.vision_model.transformer.layers.11.mlp.fc1",
	"model.vision_model.transformer.layers.11.mlp.fc2",
	"model.vision_model.transformer.layers.12.self_attn.q_proj",
	"model.vision_model.transformer.layers.12.self_attn.k_proj",
	"model.vision_model.transformer.layers.12.self_attn.v_proj",
	"model.vision_model.transformer.layers.12.self_attn.o_proj",
	"model.vision_model.transformer.layers.12.mlp.fc1",
	"model.vision_model.transformer.layers.12.mlp.fc2",
	"model.vision_model.transformer.layers.13.self_attn.q_proj",
	"model.vision_model.transformer.layers.13.self_attn.k_proj",
	"model.vision_model.transformer.layers.13.self_attn.v_proj",
	"model.vision_model.transformer.layers.13.self_attn.o_proj",
	"model.vision_model.transformer.layers.13.mlp.fc1",
	"model.vision_model.transformer.layers.13.mlp.fc2",
	"model.vision_model.transformer.layers.14.self_attn.q_proj",
	"model.vision_model.transformer.layers.14.self_attn.k_proj",
	"model.vision_model.transformer.layers.14.self_attn.v_proj",
	"model.vision_model.transformer.layers.14.self_attn.o_proj",
	"model.vision_model.transformer.layers.14.mlp.fc1",
	"model.vision_model.transformer.layers.14.mlp.fc2",
	"model.vision_model.transformer.layers.15.self_attn.q_proj",
	"model.vision_model.transformer.layers.15.self_attn.k_proj",
	"model.vision_model.transformer.layers.15.self_attn.v_proj",
	"model.vision_model.transformer.layers.15.self_attn.o_proj",
	"model.vision_model.transformer.layers.15.mlp.fc1",
	"model.vision_model.transformer.layers.15.mlp.fc2",
	"model.vision_model.transformer.layers.16.self_attn.q_proj",
	"model.vision_model.transformer.layers.16.self_attn.k_proj",
	"model.vision_model.transformer.layers.16.self_attn.v_proj",
	"model.vision_model.transformer.layers.16.self_attn.o_proj",
	"model.vision_model.transformer.layers.16.mlp.fc1",
	"model.vision_model.transformer.layers.16.mlp.fc2",
	"model.vision_model.transformer.layers.17.self_attn.q_proj",
	"model.vision_model.transformer.layers.17.self_attn.k_proj",
	"model.vision_model.transformer.layers.17.self_attn.v_proj",
	"model.vision_model.transformer.layers.17.self_attn.o_proj",
	"model.vision_model.transformer.layers.17.mlp.fc1",
	"model.vision_model.transformer.layers.17.mlp.fc2",
	"model.vision_model.transformer.layers.18.self_attn.q_proj",
	"model.vision_model.transformer.layers.18.self_attn.k_proj",
	"model.vision_model.transformer.layers.18.self_attn.v_proj",
	"model.vision_model.transformer.layers.18.self_attn.o_proj",
	"model.vision_model.transformer.layers.18.mlp.fc1",
	"model.vision_model.transformer.layers.18.mlp.fc2",
	"model.vision_model.transformer.layers.19.self_attn.q_proj",
	"model.vision_model.transformer.layers.19.self_attn.k_proj",
	"model.vision_model.transformer.layers.19.self_attn.v_proj",
	"model.vision_model.transformer.layers.19.self_attn.o_proj",
	"model.vision_model.transformer.layers.19.mlp.fc1",
	"model.vision_model.transformer.layers.19.mlp.fc2",
	"model.vision_model.transformer.layers.20.self_attn.q_proj",
	"model.vision_model.transformer.layers.20.self_attn.k_proj",
	"model.vision_model.transformer.layers.20.self_attn.v_proj",
	"model.vision_model.transformer.layers.20.self_attn.o_proj",
	"model.vision_model.transformer.layers.20.mlp.fc1",
	"model.vision_model.transformer.layers.20.mlp.fc2",
	"model.vision_model.transformer.layers.21.self_attn.q_proj",
	"model.vision_model.transformer.layers.21.self_attn.k_proj",
	"model.vision_model.transformer.layers.21.self_attn.v_proj",
	"model.vision_model.transformer.layers.21.self_attn.o_proj",
	"model.vision_model.transformer.layers.21.mlp.fc1",
	"model.vision_model.transformer.layers.21.mlp.fc2",
	"model.vision_model.transformer.layers.22.self_attn.q_proj",
	"model.vision_model.transformer.layers.22.self_attn.k_proj",
	"model.vision_model.transformer.layers.22.self_attn.v_proj",
	"model.vision_model.transformer.layers.22.self_attn.o_proj",
	"model.vision_model.transformer.layers.22.mlp.fc1",
	"model.vision_model.transformer.layers.22.mlp.fc2",
	"model.vision_model.transformer.layers.23.self_attn.q_proj",
	"model.vision_model.transformer.layers.23.self_attn.k_proj",
	"model.vision_model.transformer.layers.23.self_attn.v_proj",
	"model.vision_model.transformer.layers.23.self_attn.o_proj",
	"model.vision_model.transformer.layers.23.mlp.fc1",
	"model.vision_model.transformer.layers.23.mlp.fc2",
	"model.vision_model.transformer.layers.24.self_attn.q_proj",
	"model.vision_model.transformer.layers.24.self_attn.k_proj",
	"model.vision_model.transformer.layers.24.self_attn.v_proj",
	"model.vision_model.transformer.layers.24.self_attn.o_proj",
	"model.vision_model.transformer.layers.24.mlp.fc1",
	"model.vision_model.transformer.layers.24.mlp.fc2",
	"model.vision_model.transformer.layers.25.self_attn.q_proj",
	"model.vision_model.transformer.layers.25.self_attn.k_proj",
	"model.vision_model.transformer.layers.25.self_attn.v_proj",
	"model.vision_model.transformer.layers.25.self_attn.o_proj",
	"model.vision_model.transformer.layers.25.mlp.fc1",
	"model.vision_model.transformer.layers.25.mlp.fc2",
	"model.vision_model.transformer.layers.26.self_attn.q_proj",
	"model.vision_model.transformer.layers.26.self_attn.k_proj",
	"model.vision_model.transformer.layers.26.self_attn.v_proj",
	"model.vision_model.transformer.layers.26.self_attn.o_proj",
	"model.vision_model.transformer.layers.26.mlp.fc1",
	"model.vision_model.transformer.layers.26.mlp.fc2",
	"model.vision_model.transformer.layers.27.self_attn.q_proj",
	"model.vision_model.transformer.layers.27.self_attn.k_proj",
	"model.vision_model.transformer.layers.27.self_attn.v_proj",
	"model.vision_model.transformer.layers.27.self_attn.o_proj",
	"model.vision_model.transformer.layers.27.mlp.fc1",
	"model.vision_model.transformer.layers.27.mlp.fc2",
	"model.vision_model.transformer.layers.28.self_attn.q_proj",
	"model.vision_model.transformer.layers.28.self_attn.k_proj",
	"model.vision_model.transformer.layers.28.self_attn.v_proj",
	"model.vision_model.transformer.layers.28.self_attn.o_proj",
	"model.vision_model.transformer.layers.28.mlp.fc1",
	"model.vision_model.transformer.layers.28.mlp.fc2",
	"model.vision_model.transformer.layers.29.self_attn.q_proj",
	"model.vision_model.transformer.layers.29.self_attn.k_proj",
	"model.vision_model.transformer.layers.29.self_attn.v_proj",
	"model.vision_model.transformer.layers.29.self_attn.o_proj",
	"model.vision_model.transformer.layers.29.mlp.fc1",
	"model.vision_model.transformer.layers.29.mlp.fc2",
	"model.vision_model.transformer.layers.30.self_attn.q_proj",
	"model.vision_model.transformer.layers.30.self_attn.k_proj",
	"model.vision_model.transformer.layers.30.self_attn.v_proj",
	"model.vision_model.transformer.layers.30.self_attn.o_proj",
	"model.vision_model.transformer.layers.30.mlp.fc1",
	"model.vision_model.transformer.layers.30.mlp.fc2",
	"model.vision_model.transformer.layers.31.self_attn.q_proj",
	"model.vision_model.transformer.layers.31.self_attn.k_proj",
	"model.vision_model.transformer.layers.31.self_attn.v_proj",
	"model.vision_model.transformer.layers.31.self_attn.o_proj",
	"model.vision_model.transformer.layers.31.mlp.fc1",
	"model.vision_model.transformer.layers.31.mlp.fc2",
	"model.vision_model.global_transformer.layers.0.self_attn.q_proj",
	"model.vision_model.global_transformer.layers.0.self_attn.k_proj",
	"model.vision_model.global_transformer.layers.0.self_attn.v_proj",
	"model.vision_model.global_transformer.layers.0.self_attn.o_proj",
	"model.vision_model.global_transformer.layers.0.mlp.fc1",
	"model.vision_model.global_transformer.layers.0.mlp.fc2",
	"model.vision_model.global_transformer.layers.1.self_attn.q_proj",
	"model.vision_model.global_transformer.layers.1.self_attn.k_proj",
	"model.vision_model.global_transformer.layers.1.self_attn.v_proj",
	"model.vision_model.global_transformer.layers.1.self_attn.o_proj",
	"model.vision_model.global_transformer.layers.1.mlp.fc1",
	"model.vision_model.global_transformer.layers.1.mlp.fc2",
	"model.vision_model.global_transformer.layers.2.self_attn.q_proj",
	"model.vision_model.global_transformer.layers.2.self_attn.k_proj",
	"model.vision_model.global_transformer.layers.2.self_attn.v_proj",
	"model.vision_model.global_transformer.layers.2.self_attn.o_proj",
	"model.vision_model.global_transformer.layers.2.mlp.fc1",
	"model.vision_model.global_transformer.layers.2.mlp.fc2",
	"model.vision_model.global_transformer.layers.3.self_attn.q_proj",
	"model.vision_model.global_transformer.layers.3.self_attn.k_proj",
	"model.vision_model.global_transformer.layers.3.self_attn.v_proj",
	"model.vision_model.global_transformer.layers.3.self_attn.o_proj",
	"model.vision_model.global_transformer.layers.3.mlp.fc1",
	"model.vision_model.global_transformer.layers.3.mlp.fc2",
	"model.vision_model.global_transformer.layers.4.self_attn.q_proj",
	"model.vision_model.global_transformer.layers.4.self_attn.k_proj",
	"model.vision_model.global_transformer.layers.4.self_attn.v_proj",
	"model.vision_model.global_transformer.layers.4.self_attn.o_proj",
	"model.vision_model.global_transformer.layers.4.mlp.fc1",
	"model.vision_model.global_transformer.layers.4.mlp.fc2",
	"model.vision_model.global_transformer.layers.5.self_attn.q_proj",
	"model.vision_model.global_transformer.layers.5.self_attn.k_proj",
	"model.vision_model.global_transformer.layers.5.self_attn.v_proj",
	"model.vision_model.global_transformer.layers.5.self_attn.o_proj",
	"model.vision_model.global_transformer.layers.5.mlp.fc1",
	"model.vision_model.global_transformer.layers.5.mlp.fc2",
	"model.vision_model.global_transformer.layers.6.self_attn.q_proj",
	"model.vision_model.global_transformer.layers.6.self_attn.k_proj",
	"model.vision_model.global_transformer.layers.6.self_attn.v_proj",
	"model.vision_model.global_transformer.layers.6.self_attn.o_proj",
	"model.vision_model.global_transformer.layers.6.mlp.fc1",
	"model.vision_model.global_transformer.layers.6.mlp.fc2",
	"model.vision_model.global_transformer.layers.7.self_attn.q_proj",
	"model.vision_model.global_transformer.layers.7.self_attn.k_proj",
	"model.vision_model.global_transformer.layers.7.self_attn.v_proj",
	"model.vision_model.global_transformer.layers.7.self_attn.o_proj",
	"model.vision_model.global_transformer.layers.7.mlp.fc1",
	"model.vision_model.global_transformer.layers.7.mlp.fc2",
	"model.multi_modal_projector",
	"lm_head"
	],
	"kv_cache_scheme": null,
	"quant_method": "compressed-tensors",
	"quantization_status": "compressed",
	"sparsity_config": {},
	"transform_config": {},
	"version": "0.17.1"
	},
	"text_config": {
	"_attn_implementation_autoset": false,
	"add_cross_attention": false,
	"bos_token_id": 128000,
	"cross_attention_hidden_size": null,
	"cross_attention_layers": [
	3,
	8,
	13,
	18,
	23,
	28,
	33,
	38
	],
	"decoder_start_token_id": null,
	"dropout": 0,
	"dtype": "bfloat16",
	"eos_token_id": [
	128001,
	128008,
	128009
	],
	"finetuning_task": null,
	"hidden_act": "silu",
	"hidden_size": 4096,
	"initializer_range": 0.02,
	"intermediate_size": 14336,
	"is_decoder": false,
	"max_position_embeddings": 131072,
	"model_type": "mllama_text_model",
	"num_attention_heads": 32,
	"num_hidden_layers": 40,
	"num_key_value_heads": 8,
	"pad_token_id": 128004,
	"prefix": null,
	"pruned_heads": {},
	"rms_norm_eps": 1e-05,
	"rope_parameters": {
	"factor": 8.0,
	"high_freq_factor": 4.0,
	"low_freq_factor": 1.0,
	"original_max_position_embeddings": 8192,
	"rope_theta": 500000.0,
	"rope_type": "llama3"
	},
	"sep_token_id": null,
	"task_specific_params": null,
	"tf_legacy_loss": false,
	"tie_encoder_decoder": false,
	"tie_word_embeddings": false,
	"tokenizer_class": null,
	"torchscript": false,
	"use_bfloat16": false,
	"use_cache": true,
	"vocab_size": 128256
	},
	"transformers_version": "5.10.1",
	"unsloth_fixed": true,
	"vision_config": {
	"_attn_implementation_autoset": false,
	"add_cross_attention": false,
	"attention_heads": 16,
	"bos_token_id": null,
	"cross_attention_hidden_size": null,
	"decoder_start_token_id": null,
	"dtype": "bfloat16",
	"eos_token_id": null,
	"finetuning_task": null,
	"hidden_act": "gelu",
	"hidden_size": 1280,
	"image_size": 560,
	"initializer_range": 0.02,
	"intermediate_layers_indices": [
	3,
	7,
	15,
	23,
	30
	],
	"intermediate_size": 5120,
	"is_decoder": false,
	"max_num_tiles": 4,
	"model_type": "mllama_vision_model",
	"norm_eps": 1e-05,
	"num_channels": 3,
	"num_global_layers": 8,
	"num_hidden_layers": 32,
	"pad_token_id": null,
	"patch_size": 14,
	"prefix": null,
	"pruned_heads": {},
	"sep_token_id": null,
	"supported_aspect_ratios": [
	[
	1,
	1
	],
	[
	1,
	2
	],
	[
	1,
	3
	],
	[
	1,
	4
	],
	[
	2,
	1
	],
	[
	2,
	2
	],
	[
	3,
	1
	],
	[
	4,
	1
	]
	],
	"task_specific_params": null,
	"tf_legacy_loss": false,
	"tie_encoder_decoder": false,
	"tie_word_embeddings": true,
	"tokenizer_class": null,
	"torchscript": false,
	"use_bfloat16": false,
	"vision_output_dim": 7680
	}
	}