Instructions to use yujiepan/llama-3.2-vision-tiny-random with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use yujiepan/llama-3.2-vision-tiny-random with Transformers:

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

pipe = pipeline("text-generation", model="yujiepan/llama-3.2-vision-tiny-random")
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("yujiepan/llama-3.2-vision-tiny-random")
model = AutoModelForMultimodalLM.from_pretrained("yujiepan/llama-3.2-vision-tiny-random", 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 yujiepan/llama-3.2-vision-tiny-random with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "yujiepan/llama-3.2-vision-tiny-random"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "yujiepan/llama-3.2-vision-tiny-random",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/yujiepan/llama-3.2-vision-tiny-random

SGLang

How to use yujiepan/llama-3.2-vision-tiny-random 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 "yujiepan/llama-3.2-vision-tiny-random" \
    --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": "yujiepan/llama-3.2-vision-tiny-random",
		"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 "yujiepan/llama-3.2-vision-tiny-random" \
        --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": "yujiepan/llama-3.2-vision-tiny-random",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use yujiepan/llama-3.2-vision-tiny-random with Docker Model Runner:
```
docker model run hf.co/yujiepan/llama-3.2-vision-tiny-random
```

llama-3.2-vision-tiny-random / README.md

yujiepan

Upload folder using huggingface_hub

ab2317d verified almost 2 years ago

preview code

Raw

History Blame Contribute Delete

2.72 kB

metadata

library_name: transformers
pipeline_tag: text-generation
inference: true
widget:
  - text: Hello!
    example_title: Hello world
    group: Python

This model is for debugging. It is randomly initialized using the config from meta-llama/Llama-3.2-90B-Vision-Instruct but with smaller size.

Codes:

import os

import accelerate
import requests
import torch
import transformers
from huggingface_hub import create_repo, upload_folder
from PIL import Image
from transformers import AutoProcessor, MllamaForConditionalGeneration
from transformers.models.mllama import MllamaConfig

model_id = 'meta-llama/Llama-3.2-90B-Vision-Instruct'
repo_id = 'yujiepan/llama-3.2-vision-tiny-random'
save_path = f'/tmp/{repo_id}'

os.system(f'rm -rf {save_path}')

config = transformers.AutoConfig.from_pretrained(
    model_id,
    trust_remote_code=True,
)
config.text_config.hidden_size = 8
config.text_config.intermediate_size = 16
config.text_config.num_attention_heads = 2
config.text_config.num_key_value_heads = 1
config.text_config.num_hidden_layers = 2
config.text_config.cross_attention_layers = [1]

config.vision_config.attention_heads = 2
config.vision_config.hidden_size = 8
config.vision_config.intermediate_size = 16
config.vision_config.intermediate_layers_indices = [0]
config.vision_config.num_global_layers = 2
config.vision_config.num_hidden_layers = 2
config.vision_config.vision_output_dim = 16


transformers.set_seed(42)
model = MllamaForConditionalGeneration(config)
model.generation_config = transformers.GenerationConfig.from_pretrained(
    model_id)
model = model.to(torch.bfloat16)

transformers.set_seed(42)
with torch.no_grad():
    for p in model.parameters():
        torch.nn.init.normal_(p)

model.save_pretrained(save_path)

processor = AutoProcessor.from_pretrained(model_id)
processor.save_pretrained(save_path)

url = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/0052a70beed5bf71b92610a43a52df6d286cd5f3/diffusers/rabbit.jpg"
image = Image.open(requests.get(url, stream=True).raw)

messages = [
    {"role": "user", "content": [
        {"type": "image"},
        {"type": "text", "text": "If I had to write a haiku for this one, it would be: "}
    ]}
]
input_text = processor.apply_chat_template(
    messages, add_generation_prompt=True)
inputs = processor(image, input_text, return_tensors="pt").to(model.device)

output = model.generate(**inputs, max_new_tokens=30)
print(processor.decode(output[0]))

os.system(f'ls -alh {save_path}')
# os.system(f'rm -rf {save_path}/model.safetensors')
# create_repo(repo_id, exist_ok=True)
# upload_folder(repo_id=repo_id, folder_path=save_path)