Instructions to use KoboldAI/GPT-NeoX-20B-Erebus with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use KoboldAI/GPT-NeoX-20B-Erebus with Transformers:

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

pipe = pipeline("text-generation", model="KoboldAI/GPT-NeoX-20B-Erebus")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("KoboldAI/GPT-NeoX-20B-Erebus")
model = AutoModelForCausalLM.from_pretrained("KoboldAI/GPT-NeoX-20B-Erebus")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use KoboldAI/GPT-NeoX-20B-Erebus with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "KoboldAI/GPT-NeoX-20B-Erebus"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "KoboldAI/GPT-NeoX-20B-Erebus",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/KoboldAI/GPT-NeoX-20B-Erebus

SGLang

How to use KoboldAI/GPT-NeoX-20B-Erebus 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 "KoboldAI/GPT-NeoX-20B-Erebus" \
    --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": "KoboldAI/GPT-NeoX-20B-Erebus",
		"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 "KoboldAI/GPT-NeoX-20B-Erebus" \
        --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": "KoboldAI/GPT-NeoX-20B-Erebus",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use KoboldAI/GPT-NeoX-20B-Erebus with Docker Model Runner:
```
docker model run hf.co/KoboldAI/GPT-NeoX-20B-Erebus
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

GPT-NeoX-20B-Erebus

Model description

This is the second generation of the original Shinen made by Mr. Seeker. The full dataset consists of 6 different sources, all surrounding the "Adult" theme. The name "Erebus" comes from the greek mythology, also named "darkness". This is in line with Shin'en, or "deep abyss". For inquiries, please contact the KoboldAI community. Warning: THIS model is NOT suitable for use by minors. The model will output X-rated content.

Training procedure

GPT-NeoX-20B-Erebus was trained on a TPUv3-256 TPU pod using a heavily modified version of Ben Wang's Mesh Transformer JAX library, the original version of which was used by EleutherAI to train their GPT-J-6B model.

Training data

The data can be divided in 6 different datasets:

Literotica (everything with 4.5/5 or higher)
Sexstories (everything with 90 or higher)
Dataset-G (private dataset of X-rated stories)
Doc's Lab (all stories)
Pike Dataset (novels with "adult" rating)
SoFurry (collection of various animals)

The dataset uses [Genre: <comma-separated list of genres>] for tagging.

Limitations and biases

Based on known problems with NLP technology, potential relevant factors include bias (gender, profession, race and religion). Warning: This model has a very strong NSFW bias!

Citation details

The GPT-NeoX-20B model weights:

@inproceedings{gpt-neox-20b,
  title={{GPT-NeoX-20B}: An Open-Source Autoregressive Language Model},
  author={Black, Sid and Biderman, Stella and Hallahan, Eric and Anthony, Quentin and Gao, Leo and Golding, Laurence and He, Horace and Leahy, Connor and McDonell, Kyle and Phang, Jason and Pieler, Michael and Prashanth, USVSN Sai and Purohit, Shivanshu and Reynolds, Laria and Tow, Jonathan and Wang, Ben and Weinbach, Samuel},
  booktitle={Proceedings of the ACL Workshop on Challenges \& Perspectives in Creating Large Language Models},
  url={https://arxiv.org/abs/2204.06745},
  year={2022}
}

The Mesh Transformer JAX library:

@misc{mesh-transformer-jax,
  author = {Wang, Ben},
  title = {{Mesh-Transformer-JAX: Model-Parallel Implementation of Transformer Language Model with JAX}},
  howpublished = {\url{https://github.com/kingoflolz/mesh-transformer-jax}},
  year = 2021,
  month = May
}

Downloads last month: 2,137

Model tree for KoboldAI/GPT-NeoX-20B-Erebus

Quantizations

3 models

Spaces using KoboldAI/GPT-NeoX-20B-Erebus 32

Paper for KoboldAI/GPT-NeoX-20B-Erebus

GPT-NeoX-20B: An Open-Source Autoregressive Language Model

Paper • 2204.06745 • Published Apr 14, 2022 • 1