Instructions to use Undi95/MLewd-L2-13B-v2-1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Undi95/MLewd-L2-13B-v2-1 with Transformers:

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

pipe = pipeline("text-generation", model="Undi95/MLewd-L2-13B-v2-1")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Undi95/MLewd-L2-13B-v2-1")
model = AutoModelForCausalLM.from_pretrained("Undi95/MLewd-L2-13B-v2-1")

Inference
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Undi95/MLewd-L2-13B-v2-1 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Undi95/MLewd-L2-13B-v2-1"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Undi95/MLewd-L2-13B-v2-1",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/Undi95/MLewd-L2-13B-v2-1

SGLang

How to use Undi95/MLewd-L2-13B-v2-1 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 "Undi95/MLewd-L2-13B-v2-1" \
    --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": "Undi95/MLewd-L2-13B-v2-1",
		"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 "Undi95/MLewd-L2-13B-v2-1" \
        --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": "Undi95/MLewd-L2-13B-v2-1",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use Undi95/MLewd-L2-13B-v2-1 with Docker Model Runner:
```
docker model run hf.co/Undi95/MLewd-L2-13B-v2-1
```

v2.2 requests

pinned

by Undi95 - opened Sep 7, 2023

Discussion

Undi95

Owner Sep 7, 2023

•

edited Sep 7, 2023

Hello, I've seen some feedback here and there, the main issue as I saw was that the model was overbaked with p*rn (while still being able to talk normally), and finish story/sexual act for the user.

My goal was first to uncensor all the shit you degen (and myself heheh) are able to think about, and it worked, thus having the secret recipe and each step of it saved, I can clearly modify what's in it.

Do you have suggestion of model with good dataset I could add in it ? Maybe a request to make it act differently ?
Post it here.

Also, there 3 variant of this model and I suggest you to try them (015 / 050). Finding the magic sauce is hard bros, but I will smash it.

In V2.2, pyg2 will be less present, pyg2 have a lot of good data in it but have a censoring side too, I tested, and in the most degen thing 2/10 reply was censored, softly but it was.
Please be patient with me, I take 6h of sleep and 18h of rocksmashing model per day to try to find the magic touch kek

Undi95 pinned discussion Sep 7, 2023

teknium

Sep 17, 2023

Dataset? Pls 🙏

Undi95

Owner Sep 17, 2023

Dataset? Pls 🙏

I don't have any. I break rock (model). Then I glue them (merge).
Kek...

I'm sorry. All modeI/lora used is in the model card anyway, try to contact the OG authors to get the dataset.

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