Instructions to use Undi95/MXLewd-L2-20B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Undi95/MXLewd-L2-20B with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="Undi95/MXLewd-L2-20B")# Load model directly from transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("Undi95/MXLewd-L2-20B") model = AutoModelForCausalLM.from_pretrained("Undi95/MXLewd-L2-20B") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use Undi95/MXLewd-L2-20B with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "Undi95/MXLewd-L2-20B" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Undi95/MXLewd-L2-20B", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/Undi95/MXLewd-L2-20B
- SGLang
How to use Undi95/MXLewd-L2-20B 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/MXLewd-L2-20B" \ --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/MXLewd-L2-20B", "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/MXLewd-L2-20B" \ --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/MXLewd-L2-20B", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }' - Docker Model Runner
How to use Undi95/MXLewd-L2-20B with Docker Model Runner:
docker model run hf.co/Undi95/MXLewd-L2-20B
Plans for 70b?
This one works better than any of the others I've tested, including the others here. It gives the least generic responses and seems to avoid repetition even at full context. Any plans for a 70b version or any plans to find someone who can train it?
@Delcos Have you tried the Emerhyst 20B one yet (https://huggingface.co/Undi95/Emerhyst-20B)? If yes, what's your impression of it comparing to this one?
I don't know if it's possible kek, I also don't have the power to try/doing that sadly.
@Delcos Have you tried the Emerhyst 20B one yet (https://huggingface.co/Undi95/Emerhyst-20B)? If yes, what's your impression of it comparing to this one?
This one just seems to get actual conversation better, including Emerhyst. It seems to be able to stay away from generic answers better. As an example I have a character that doesn't speak like the generic AI at all, and out of the 4 here including Em MX was the only one to never give an answer with a word like "assist".
Repetition not being as big of an issue isn't something I can't confirm or know if it's the way I prompt it, or more down to the model.
I don't know if it's possible kek, I also don't have the power to try/doing that sadly.
I forgot this was part of a merge, but if you need GPUs runpod if pretty good for training larger models. Not great for actually using them though.