Instructions to use anymodality/llava-v1.5-7b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use anymodality/llava-v1.5-7b with Transformers:

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

pipe = pipeline("text-generation", model="anymodality/llava-v1.5-7b")

# Load model directly
from transformers import AutoProcessor, AutoModelForCausalLM

processor = AutoProcessor.from_pretrained("anymodality/llava-v1.5-7b")
model = AutoModelForCausalLM.from_pretrained("anymodality/llava-v1.5-7b")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use anymodality/llava-v1.5-7b with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "anymodality/llava-v1.5-7b"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "anymodality/llava-v1.5-7b",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/anymodality/llava-v1.5-7b

SGLang

How to use anymodality/llava-v1.5-7b 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 "anymodality/llava-v1.5-7b" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "anymodality/llava-v1.5-7b",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "anymodality/llava-v1.5-7b" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "anymodality/llava-v1.5-7b",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use anymodality/llava-v1.5-7b with Docker Model Runner:
```
docker model run hf.co/anymodality/llava-v1.5-7b
```

Should Deploy Happen o̶n̶ from a GPU?

by MetaSkills - opened Nov 1, 2023

Discussion

MetaSkills

Nov 1, 2023

This might showcase my ignorance on how much the SageMaker tools are doing with the Deep Learning Containers. But thought I would ask.

liltom-eth

AnyModality org Nov 4, 2023

@MetaSkills Yeah, the deployment happens on a GPU when GPU instance is specified. The SageMaker gives users pre-built Pytorch Deep Learning Containers and we customize the container with some package defined in requirements.txt, and use inference.py to define the endpoint interface.
Is that your question?

MetaSkills

Nov 4, 2023

Thanks, but not really.

Was more or less thinking about the machine running the deployment. I never spun up SageMaker to play with the Notebook because I could never get LFS working with any SageMaker setup. Yum not working, etc, etc.

So, I ended up pulling apart the notebook into little bash scripts and running them from a Python Devcontainer locally, in my M2 Mac. And everything works. But the whole python code for the SageMaker deploy is still a mystery to me in how the final DLC is built and where that happens. My guess is remotely someplace in the pipeline. Hence the question. Totally something I can keep digging and learn myself too.

MetaSkills changed discussion title from Should Deploy Happen on a GPU? to Should Deploy Happen o̶n̶ from a GPU? Nov 4, 2023

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment