app.py

import gradio as gr

# gr.load("models/shivanikerai/TinyLlama-1.1B-Chat-v1.0-seo-optimised-title-suggestion-v1.0").launch()
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("shivanikerai/TinyLlama-1.1B-Chat-v1.0-seo-optimised-title-suggestion-v1.0")
model = AutoModelForCausalLM.from_pretrained("shivanikerai/TinyLlama-1.1B-Chat-v1.0-seo-optimised-title-suggestion-v1.0")
def generate_title_suggestions(keywords, product_info):
    # Define the roles and markers
    B_SYS, E_SYS = "<<SYS>>", "<</SYS>>"
    B_INST, E_INST = "[INST]", "[/INST]"
    B_in, E_in = "[Product Details]", "[/Product Details]"
    B_out, E_out = "[Suggested Titles]", "[/Suggested Titles]"

    # Format your prompt template
    prompt = f"""{B_INST} {B_SYS} You are a helpful, respectful and honest assistant for ecommerce product title creation. {E_SYS}\nCreate a SEO optimized e-commerce product title for the keywords:{keywords.strip()}\n{B_in}{product_info}{E_in}\n{E_INST}\n\n{B_out}"""


    print("Prompt:")
    print(prompt)

    encoding = tokenizer(prompt, return_tensors="pt").to("cuda:0")
    output = model.generate(input_ids=encoding.input_ids,
                            attention_mask=encoding.attention_mask,
                            max_new_tokens=1024,
                            do_sample=True,
                            temperature=0.01,
                            eos_token_id=tokenizer.eos_token_id,
                            top_k=0)

    print()

    # Subtract the length of input_ids from output to get only the model's response
    output_text = tokenizer.decode(output[0, len(encoding.input_ids[0]):], skip_special_tokens=False)
    output_text = re.sub('\n+', '\n', output_text)  # remove excessive newline characters

    print("Generated Assistant Response:")
    print(output_text)
    gr.Interface(
    generate_title_suggestions,
    inputs='text',
    outputs='text',
    title="Title Suggestion",
).launch()