Instructions to use ethzanalytics/gpt-j-8bit-KILT_WoW_10k_steps with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ethzanalytics/gpt-j-8bit-KILT_WoW_10k_steps with Transformers:

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

pipe = pipeline("text-generation", model="ethzanalytics/gpt-j-8bit-KILT_WoW_10k_steps")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("ethzanalytics/gpt-j-8bit-KILT_WoW_10k_steps")
model = AutoModelForCausalLM.from_pretrained("ethzanalytics/gpt-j-8bit-KILT_WoW_10k_steps")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use ethzanalytics/gpt-j-8bit-KILT_WoW_10k_steps with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ethzanalytics/gpt-j-8bit-KILT_WoW_10k_steps"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ethzanalytics/gpt-j-8bit-KILT_WoW_10k_steps",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/ethzanalytics/gpt-j-8bit-KILT_WoW_10k_steps

SGLang

How to use ethzanalytics/gpt-j-8bit-KILT_WoW_10k_steps 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 "ethzanalytics/gpt-j-8bit-KILT_WoW_10k_steps" \
    --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": "ethzanalytics/gpt-j-8bit-KILT_WoW_10k_steps",
		"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 "ethzanalytics/gpt-j-8bit-KILT_WoW_10k_steps" \
        --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": "ethzanalytics/gpt-j-8bit-KILT_WoW_10k_steps",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use ethzanalytics/gpt-j-8bit-KILT_WoW_10k_steps with Docker Model Runner:
```
docker model run hf.co/ethzanalytics/gpt-j-8bit-KILT_WoW_10k_steps
```

ethzanalytics/gpt-j-8bit-KILT_WoW_10k_steps

This is a version of hivemind/gpt-j-6B-8bit fine-tuned on the Wizard of Wikipedia dataset for 10k steps (just under an epoch) on an A100. it can be used as a chatbot. It is designed to be used with ai-msgbot to take advantage of the prompt engineering.

Usage

NOTE: this needs to be loaded via the special patching technique outlined in the hivemind model card (as with all 8bit models)

Examples of how to load the model correctly are already in place in the notebook linked above. A .py of said notebook was uploaded to the repo for reference - link here

Training

For details, please see this wandb report for both the daily-dialogues version and the WoW version.

Downloads last month: 15

Dataset used to train ethzanalytics/gpt-j-8bit-KILT_WoW_10k_steps

Paper for ethzanalytics/gpt-j-8bit-KILT_WoW_10k_steps

Wizard of Wikipedia: Knowledge-Powered Conversational agents

Paper • 1811.01241 • Published Nov 3, 2018