Instructions to use Salesforce/instructblip-vicuna-7b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Salesforce/instructblip-vicuna-7b with Transformers:

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

pipe = pipeline("image-text-to-text", model="Salesforce/instructblip-vicuna-7b")

# Load model directly
from transformers import AutoProcessor, AutoModelForImageTextToText

processor = AutoProcessor.from_pretrained("Salesforce/instructblip-vicuna-7b")
model = AutoModelForImageTextToText.from_pretrained("Salesforce/instructblip-vicuna-7b")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Salesforce/instructblip-vicuna-7b with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/Salesforce/instructblip-vicuna-7b

SGLang

How to use Salesforce/instructblip-vicuna-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 "Salesforce/instructblip-vicuna-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": "Salesforce/instructblip-vicuna-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 "Salesforce/instructblip-vicuna-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": "Salesforce/instructblip-vicuna-7b",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use Salesforce/instructblip-vicuna-7b with Docker Model Runner:
```
docker model run hf.co/Salesforce/instructblip-vicuna-7b
```

Is v0 or v1?

by bemer12 - opened May 31, 2023

Discussion

bemer12

May 31, 2023

Thanks for your effort!

nielsr

May 31, 2023

Hi,

What do you mean by v0 or v1?

kjerk

Jun 7, 2023

Hi,

What do you mean by v0 or v1?

Between Vicuna v0 and v1.1 the instruction prompt structure changed as detailed here https://github.com/lm-sys/FastChat/blob/main/docs/vicuna_weights_version.md#example-prompt-weight-v11
That would mean that this model would need to be prompted one way or the other to maintain the expected instruction structure, because the instruct_blip_vicuna*.pth weights were finetuned on one or the other.

According to the instructblip project page, the Vicuna 1.1 weights are meant to be prepared, detailed here https://github.com/salesforce/LAVIS/tree/main/projects/instructblip#prepare-vicuna-weights
So I would guess that is canonically correct, can't verify 100% that's the source of these weights, but it's a good guess.

nielsr

Jun 7, 2023

Oh thanks for notifying! It seems to be Vicuna 1.1.

I followed the guide here: https://github.com/lm-sys/FastChat#model-weights.

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