Instructions to use reeducator/vicuna-13b-cocktail with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use reeducator/vicuna-13b-cocktail with Transformers:

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

pipe = pipeline("text-generation", model="reeducator/vicuna-13b-cocktail")

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("reeducator/vicuna-13b-cocktail")
model = AutoModelForCausalLM.from_pretrained("reeducator/vicuna-13b-cocktail", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use reeducator/vicuna-13b-cocktail with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/reeducator/vicuna-13b-cocktail

SGLang

How to use reeducator/vicuna-13b-cocktail 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 "reeducator/vicuna-13b-cocktail" \
    --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": "reeducator/vicuna-13b-cocktail",
		"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 "reeducator/vicuna-13b-cocktail" \
        --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": "reeducator/vicuna-13b-cocktail",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use reeducator/vicuna-13b-cocktail with Docker Model Runner:
```
docker model run hf.co/reeducator/vicuna-13b-cocktail
```

The Stickiness Problem

by deleted - opened May 10, 2023

Discussion

deleted

May 10, 2023

https://huggingface.co/ehartford/WizardLM-7B-Uncensored/discussions/10

@reeducator Just linking this thread from Henk about overtraining. Some extra reading direction for model stickiness assuming you haven't been messing with dials and knobs already around these things.

reeducator

Owner May 10, 2023

Thanks @gozfarb. Our hyperparams are still default from the vicuna branch. Learning rate is 2e-5 with cosine scheduling, i.e. the LR goes orders of magnitude lower towards the end of the training. We might keep it as it is for now, unless someone points out that the model we have here suffers from similar issues. For Vicuna I can't lower it much, since the training already takes quite some time, but if necessary, there's still room for longer training periods.

deleted changed discussion title from The Stickiness Promblem to The Stickiness Problem May 13, 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