Instructions to use DewEfresh/pixtral-12b-8bit with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use DewEfresh/pixtral-12b-8bit with Transformers:

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

pipe = pipeline("image-text-to-text", model="DewEfresh/pixtral-12b-8bit")

# Load model directly
from transformers import AutoProcessor, AutoModelForImageTextToText

processor = AutoProcessor.from_pretrained("DewEfresh/pixtral-12b-8bit")
model = AutoModelForImageTextToText.from_pretrained("DewEfresh/pixtral-12b-8bit")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use DewEfresh/pixtral-12b-8bit with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "DewEfresh/pixtral-12b-8bit"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "DewEfresh/pixtral-12b-8bit",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/DewEfresh/pixtral-12b-8bit

SGLang

How to use DewEfresh/pixtral-12b-8bit 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 "DewEfresh/pixtral-12b-8bit" \
    --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": "DewEfresh/pixtral-12b-8bit",
		"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 "DewEfresh/pixtral-12b-8bit" \
        --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": "DewEfresh/pixtral-12b-8bit",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use DewEfresh/pixtral-12b-8bit with Docker Model Runner:
```
docker model run hf.co/DewEfresh/pixtral-12b-8bit
```

Input type (torch.cuda.FloatTensor) and weight type (torch.cuda.HalfTensor) should be the same

by treehugg3 - opened Apr 10, 2025

Discussion

treehugg3

Apr 10, 2025

•

edited Apr 10, 2025

I get this error when I load the model using the latest transformers: Input type (torch.cuda.FloatTensor) and weight type (torch.cuda.HalfTensor) should be the same

I am just loading it using

model = LlavaForConditionalGeneration.from_pretrained(vlm_path,
    use_safetensors=True)

I get this error on this quantized version but not the full pixtral-12b repo. I did however copy the tokenizer_config.json from that repo into this one in order to use the tokenizer.

edit: It fails even with the example in #1.
edit2: I got it to work on this ANCIENT transformers fork branch mentioned here: https://huggingface.co/DewEfresh/pixtral-12b-8bit/discussions/1#66f1a38916c5478fa68c05d6

Unfortunately it is broken on the modern transformers package :(

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