# import gradio as gr

# gr.load("models/manohar02/Llama-2-7b-quantize").launch()


import gradio as gr
from langchain import HuggingFacePipeline, PromptTemplate, LLMChain
from transformers import AutoTokenizer
import transformers
import torch

# Define the Hugging Face model
model = "manohar02/Llama-2-7b-quantize"

# Define the Hugging Face pipeline
pipeline = transformers.pipeline(
    "text-generation",  # task
    model=model,
    torch_dtype=torch.bfloat16,
    max_length=20000,
    do_sample=True,
    top_k=10,
    num_return_sequences=1,
    eos_token_id=AutoTokenizer.from_pretrained(model).eos_token_id
)

llm = HuggingFacePipeline(pipeline=pipeline, model_kwargs={'temperature': 0})

# Define the template for summarization
template = """
Write a concise summary of the following text delimited by triple backquotes.
'''{text}'''
SUMMARY:
"""

prompt = PromptTemplate(template=template, input_variables=["text"])

llm_chain = LLMChain(prompt=prompt, llm=llm)

# Function to generate summary
def generate_summary(text):
    summary = llm_chain.run(text)
    return summary.split('SUMMARY:')[-1].strip()

# Create a Gradio interface
iface = gr.Interface(fn=generate_summary, inputs="text", outputs="text", title="LLaMA2 Summarizer")
iface.launch()