from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig
import torch

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/llama-eap"

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

prompt = "[INST] Write a poem about tomatoes in the style of Poe.[/INST]"
inputs = tokenizer(prompt, return_tensors="pt")

outputs = model.generate(**inputs)

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

License

This project is licensed under the MIT License.

glides/llama-eap

Edgar Allen Poe LLM

Features

Usage

License

glides/
llama-eap