metadata

language:
  - en
license: mit
tags:
  - trl
  - sft
  - sgd
base_model: TinyLlama/TinyLlama-1.1B-Chat-v1.0
datasets:
  - nroggendorff/mayo
model-index:
  - name: mayo
    results: []

Mayonnaise LLM

Mayo is a language model fine-tuned on the Mayo dataset using Supervised Fine-Tuning (SFT) and Teacher Reinforced Learning (TRL) techniques. It is based on the Mistral 7b model

Features

Utilizes SFT and TRL techniques for improved performance
Supports English language

Usage

To use the Mayo LLM, you can load the model using the Hugging Face Transformers library:

from transformers import pipeline

pipe = pipeline("text-generation", model="nroggendorff/mayo")

question = "What color is the sky?"
conv = [{"role": "user", "content": question}]

response = pipe(conv, max_new_tokens=32)[0]['generated_text'][-1]['content']
print(response)

To use the model with quantization:

from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig

bnb_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_use_double_quant=True,
    bnb_4bit_quant_type="nf4",
    bnb_4bit_compute_dtype=torch.bfloat16
)

model_id = "nroggendorff/mayo"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, quantization_config=bnb_config)

prompt = "[INST] What color is the sky? [/INST]"
inputs = tokenizer(prompt, return_tensors="pt")

outputs = model.generate(**inputs, max_new_tokens=32)

generated_text = tokenizer.batch_decode(outputs)[0]
print(generated_text)

License

This project is licensed under the MIT License.