app.py

from peft import PeftModel
from transformers import LLaMATokenizer, LLaMAForCausalLM, GenerationConfig
import torch
n_gpus = torch.cuda.device_count()
max_memory = {i: max_memory for i in range(n_gpus)}

print(f'Max memory : {max_memory}')

tokenizer = LLaMATokenizer.from_pretrained("decapoda-research/llama-7b-hf")
max_memory = '40GB'

model = LLaMAForCausalLM.from_pretrained(
    "decapoda-research/llama-7b-hf",
    load_in_8bit=True,
    device_map="auto",max_memory=max_memory
)

model = PeftModel.from_pretrained(model, "tloen/alpaca-lora-7b")

def generate_prompt(instruction, input=None):
    if input:
        return f"""Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request. Answer step by step.

### Instruction:
{instruction}

### Input:
{input}

### Response:"""
    else:
        return f"""Below is an instruction that describes a task. Write a response that appropriately completes the request. Answer step by step.

### Instruction:
{instruction}

### Response:"""

generation_config = GenerationConfig(
    temperature=0.1,
    top_p=0.75,
    num_beams=4,
)

def evaluate(instruction, input=None):
    prompt = generate_prompt(instruction, input)
    inputs = tokenizer(prompt, return_tensors="pt")
    input_ids = inputs["input_ids"].cuda()
    generation_output = model.generate(
        input_ids=input_ids,
        generation_config=generation_config,
        return_dict_in_generate=True,
        output_scores=True,
        max_new_tokens=256
    )
    for s in generation_output.sequences:
        output = tokenizer.decode(s)
        print("Response:", output.split("### Response:")[1].strip())

import gradio as gr
from peft import PeftModel
from transformers import LLaMATokenizer, LLaMAForCausalLM, GenerationConfig

import gradio as gr

def evaluate1(instruction):
    prompt = generate_prompt(instruction)
    inputs = tokenizer(prompt, return_tensors="pt")
    input_ids = inputs["input_ids"].cuda()
    generation_output = model.generate(
        input_ids=input_ids,
        generation_config=generation_config,
        return_dict_in_generate=True,
        output_scores=True,
        max_new_tokens=256
    )
    for s in generation_output.sequences:
        output = tokenizer.decode(s)
        return output.split("### Response:")[1].strip()

inputs = gr.inputs.Textbox(lines=5, label="Instruction")
outputs = gr.outputs.Textbox(label="Response")
title = "LLaMA-7B Language Model"
description = "This is a LLaMA-7B language model fine-tuned on various text datasets to generate text for a given task. It was trained on PyTorch by and is capable of generating high-quality, coherent text that is similar to human writing. The model is highly versatile and can be used for a variety of tasks, including text completion, summarization, and translation."
copyright = "Copyright Bhaskar Tripathi (2023)"

gr.Interface(evaluate1, inputs, outputs, title=title, description=description, footer=copyright, flag=False).launch()