Instructions to use QuixiAI/WizardLM-13B-Uncensored with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use QuixiAI/WizardLM-13B-Uncensored with Transformers:

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

pipe = pipeline("text-generation", model="QuixiAI/WizardLM-13B-Uncensored")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("QuixiAI/WizardLM-13B-Uncensored")
model = AutoModelForCausalLM.from_pretrained("QuixiAI/WizardLM-13B-Uncensored")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use QuixiAI/WizardLM-13B-Uncensored with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "QuixiAI/WizardLM-13B-Uncensored"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "QuixiAI/WizardLM-13B-Uncensored",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/QuixiAI/WizardLM-13B-Uncensored

SGLang

How to use QuixiAI/WizardLM-13B-Uncensored 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 "QuixiAI/WizardLM-13B-Uncensored" \
    --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": "QuixiAI/WizardLM-13B-Uncensored",
		"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 "QuixiAI/WizardLM-13B-Uncensored" \
        --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": "QuixiAI/WizardLM-13B-Uncensored",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use QuixiAI/WizardLM-13B-Uncensored with Docker Model Runner:
```
docker model run hf.co/QuixiAI/WizardLM-13B-Uncensored
```

What setting must be used? Model fails to load in oobabooga.

by YuriGagarine - opened May 11, 2023

Discussion

YuriGagarine

May 11, 2023

The Oobabooga webui downloads the model fine but then only loads it with errors. What settings must be used? (I've tied wbits:4, groupsize: 128, llama)

Traceback (most recent call last):
File “D:\oobabooga_windows\text-generation-webui\server.py”, line 59, in load_model_wrapper
shared.model, shared.tokenizer = load_model(shared.model_name)
File “D:\oobabooga_windows\text-generation-webui\modules\models.py”, line 159, in load_model
model = load_quantized(model_name)
File “D:\oobabooga_windows\text-generation-webui\modules\GPTQ_loader.py”, line 170, in load_quantized
exit()
File “D:\oobabooga_windows\installer_files\env\lib_sitebuiltins.py”, line 26, in call
raise SystemExit(code)
SystemExit: None

CyberJohn

May 11, 2023

I have the same problem...

Wildstar50

May 11, 2023

•

edited May 11, 2023

I got it working with wbits and groupsize blank.
Model type = llama
Load in 8-bit checked.

ehartford

Quixi AI org May 11, 2023

It's not 8 bit. It's native.

ehartford

Quixi AI org May 11, 2023

I'll make a blog post tomorrow

ehartford

Quixi AI org May 11, 2023

https://erichartford.com

ehartford

Quixi AI org May 11, 2023

"GPTQ_loader.py" means you are trying to load it as quantized.
It's not quantized.

CyberJohn

May 11, 2023

"GPTQ_loader.py" means you are trying to load it as quantized.
It's not quantized.

Thank you and wildstar50 - much appreciated.

jcuffe

May 11, 2023

•

edited May 11, 2023

@ehartford do you have any advice for hosting this model using the huggingface "inference endpoints" or "spaces?"

I'm interested in experimenting with models but just wasted all my cash on an AMD card so all I have is a huggingface account and a credit limit

edit: Download and attempted to load with oobabooga and it looks like it'll only run with CUDA support :(

carlosgfhg

May 21, 2023

thanks friend "wildstars50" it was true I put it as you indicated and it works fine.

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