Instructions to use rhymes-ai/Aria with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use rhymes-ai/Aria with Transformers:

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

pipe = pipeline("image-text-to-text", model="rhymes-ai/Aria")
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, AutoModelForImageTextToText

processor = AutoProcessor.from_pretrained("rhymes-ai/Aria")
model = AutoModelForImageTextToText.from_pretrained("rhymes-ai/Aria")
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

vLLM

How to use rhymes-ai/Aria with vLLM:

Install from pip and serve model

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

SGLang

How to use rhymes-ai/Aria 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 "rhymes-ai/Aria" \
    --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": "rhymes-ai/Aria",
		"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 "rhymes-ai/Aria" \
        --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": "rhymes-ai/Aria",
		"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 rhymes-ai/Aria with Docker Model Runner:
```
docker model run hf.co/rhymes-ai/Aria
```

aria-dev commited on Oct 15, 2024

Commit

006dc0f

1 Parent(s): 90fb8b4

support eager attention

Browse files

Files changed (3) hide show

config.json +2 -0
modeling_aria.py +1 -0
vision_encoder.py +0 -1

config.json CHANGED Viewed

@@ -30,8 +30,10 @@
   },
   "torch_dtype": "bfloat16",
   "transformers_version": "4.45.0",
   "vision_config": {
     "_flash_attn_2_enabled": true,
     "architectures": [
       "AriaVisionModel"
     ],

   },
   "torch_dtype": "bfloat16",
   "transformers_version": "4.45.0",
+  "_attn_implementation": "flash_attention_2",
   "vision_config": {
     "_flash_attn_2_enabled": true,
+    "_attn_implementation": "flash_attention_2",
     "architectures": [
       "AriaVisionModel"
     ],

modeling_aria.py CHANGED Viewed

@@ -133,6 +133,7 @@ class AriaForConditionalGeneration(AriaPretrainedModel):
     def __init__(self, config: AriaConfig):
         super().__init__(config)
         self.vision_tower = AriaVisionModel(config.vision_config)
         self.multi_modal_projector = build_mm_projector(config)
         self.vocab_size = config.text_config.vocab_size

     def __init__(self, config: AriaConfig):
         super().__init__(config)
+        config.vision_config._attn_implementation = config._attn_implementation
         self.vision_tower = AriaVisionModel(config.vision_config)
         self.multi_modal_projector = build_mm_projector(config)
         self.vocab_size = config.text_config.vocab_size

vision_encoder.py CHANGED Viewed

@@ -38,7 +38,6 @@ class AriaVisionConfig(SiglipVisionConfig):
         **kwargs,
     ):
         super().__init__(**kwargs)
-        self._attn_implementation = "flash_attention_2"
 class IdentityOp(torch.nn.Module):

         **kwargs,
     ):
         super().__init__(**kwargs)
 class IdentityOp(torch.nn.Module):