Instructions to use phanerozoic/PirateTalk-13b-v1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use phanerozoic/PirateTalk-13b-v1 with Transformers:

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

pipe = pipeline("text-generation", model="phanerozoic/PirateTalk-13b-v1")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("phanerozoic/PirateTalk-13b-v1")
model = AutoModelForCausalLM.from_pretrained("phanerozoic/PirateTalk-13b-v1")

Inference
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use phanerozoic/PirateTalk-13b-v1 with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/phanerozoic/PirateTalk-13b-v1

SGLang

How to use phanerozoic/PirateTalk-13b-v1 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 "phanerozoic/PirateTalk-13b-v1" \
    --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": "phanerozoic/PirateTalk-13b-v1",
		"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 "phanerozoic/PirateTalk-13b-v1" \
        --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": "phanerozoic/PirateTalk-13b-v1",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use phanerozoic/PirateTalk-13b-v1 with Docker Model Runner:
```
docker model run hf.co/phanerozoic/PirateTalk-13b-v1
```

This repository contains the PirateTalk-13b-v1 model, an advanced derivative of the 13b Llama 2 Chat model. It has been fine-tuned on a comprehensive dataset encompassing a wide spectrum of pirate-themed content, from standard pirate lexemes to intricate elements of pirate vernacular.

Objective: The inception of Piratetalk-13b-v1 was driven by the objective to integrate a specific dialect—pirate language—into the model. Our ambition was to ensure that the model not only adopts pirate vocabulary but also the nuanced syntactic structures inherent to pirate discourse.

Model Evolution: Piratetalk-13b-v1 epitomizes our continued efforts in domain-specific model fine-tuning. While our preliminary merged model was anchored in the OpenOrca series, with PirateTalk-13b-v1, we've leveraged the lessons from that experiment and incorporated the fine-tuning directly into the Llama 2 architecture. This methodology, combined with a curated dataset, reflects our ongoing commitment to pushing the boundaries of model adaptability.

Performance Insights: Comparative evaluations indicate that PirateTalk-13b-v1 surpasses its OpenOrca-based predecessor in terms of both response accuracy and dialect consistency. The enhanced performance of PirateTalk-13b-v1 can likely be attributed to our refined dataset and optimized hyperparameter settings. It's important to emphasize that this improvement isn't a reflection of any shortcomings of the OpenOrca model but rather the advancements in our training strategies.

Technical Specifications: PirateTalk-13b-v1 underwent training at half precision (16) and is optimized for inference at this precision level.

Future Endeavors: While we acknowledge the success of PirateTalk-13b-v1 as a testament to our proof-of-concept, our exploration doesn't conclude here. We envisage extending this methodology to larger quantized models, aiming to further enhance the model's knowledge depth, practical utility, and linguistic flair in subsequent iterations.

Downloads last month: 10