app.py

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

peft_model_id = "Bsbell21/MarketMailAI"
config = PeftConfig.from_pretrained(peft_model_id)
model = AutoModelForCausalLM.from_pretrained(
    config.base_model_name_or_path,
    return_dict=True,
    device_map="auto"
)
#tokenizer = AutoTokenizer.from_pretrained(config.base_model_name_or_path)

mixtral_tokenizer = AutoTokenizer.from_pretrained(peft_model_id)

# Load the Lora model
model = PeftModel.from_pretrained(model, peft_model_id)

def input_from_text(product, description):
  return f"<s>[INST]Below is a product and description, please write a marketing email for this product.\n\n### Product:\n{product}\n### Description:\n{description}\n\n### Marketing Email:[/INST]"

def make_inference(product, description):
  inputs = mixtral_tokenizer(input_from_text(product, description), return_tensors="pt")

  outputs = merged_model.generate(
      **inputs,
      max_new_tokens=150,
      generation_kwargs={"repetition_penalty" : 1.7}
      )
  # print(mixtral_tokenizer.decode(outputs[0], skip_special_tokens=True))
  result = mixtral_tokenizer.decode(outputs[0], skip_special_tokens=True).split("[/INST]")[1]
  return result
'''
def make_inference(product_name, product_description):
    batch = tokenizer(
        f"### Product and Description:\n{product_name}: {product_description}\n\n### Ad:",
        return_tensors="pt",
    )
    
    batch = {key: value.to('cuda:0') for key, value in batch.items()}
    with torch.cuda.amp.autocast():
        output_tokens = model.generate(**batch, max_new_tokens=50)
    return tokenizer.decode(output_tokens[0], skip_special_tokens=True)
'''
if __name__ == "__main__":
    # make a gradio interface
    import gradio as gr

    gr.Interface(
        make_inference,
        [
            gr.Textbox(lines=2, label="Product Name"),
            gr.Textbox(lines=5, label="Product Description"),
        ],
        gr.Textbox(label="Ad"),
        title="GenerAd-AI",
        description="GenerAd-AI is a generative model that generates ads for products.",
    ).launch()