bhuvneshsaini/merged_model

Model Card for Model ID

This is an adapter for Meta's Llama 2 7B fine-tuned for translating Italian text into English.

Model Details

Model Description

• Developed by: Bhuvnesh Saini
• Model type: LoRA Adapter for Llama 2 7B
• Language(s) (NLP): French, Italian, English
• License: MIT license

Uses

This adapter must be loaded on top of Llama 2 7B. It has been fine-tuned with QLoRA. For optimal results, the base model must be loaded with the exact same configuration used during fine-tuning. You can use the following code to load the model:

from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig
import torch
from peft import PeftModel

base_model = "meta-llama/Llama-2-7b-hf"
compute_dtype = getattr(torch, "float16")
bnb_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_quant_type="nf4",
    bnb_4bit_compute_dtype=compute_dtype,
    bnb_4bit_use_double_quant=True,
)
model = AutoModelForCausalLM.from_pretrained(
        original_model_directory, device_map={"": 0}, quantization_config=bnb_config
)
tokenizer = AutoTokenizer.from_pretrained(base_model, use_fast=True)
model = PeftModel.from_pretrained(model, "kaitchup/Llama-2-7b-mt-Italian-to-English")

Then, run the model as follows:

my_text = "" #put your text to translate here

prompt = my_text+" ###>"

tokenized_input = tokenizer(prompt, return_tensors="pt")
input_ids = tokenized_input["input_ids"].cuda()

generation_output = model.generate(
        input_ids=input_ids,
        num_beams=10,
        return_dict_in_generate=True,
        output_scores=True,
        max_new_tokens=130

)
for seq in generation_output.sequences:
    output = tokenizer.decode(seq, skip_special_tokens=True)
    print(output.split("###>")[1].strip())