Spaces:
Runtime error
Runtime error
File size: 2,307 Bytes
78a8a9a 04d2512 6f2a620 04d2512 9568320 6f2a620 9568320 04d2512 6f2a620 9568320 04d2512 6f2a620 04d2512 9568320 04d2512 6f2a620 9568320 04d2512 86d9873 04d2512 9568320 04d2512 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 |
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, pipeline
# pipe = pipeline("text-generation", model="shivanikerai/TinyLlama-1.1B-Chat-v1.0-seo-optimised-title-suggestion-v1.0")
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)
# predictions = pipeline(prompt)
# output=((predictions[0]['generated_text']).split(B_out)[-1]).strip()
# return (output)
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:")
return (output_text)
gr.Interface(
generate_title_suggestions,
inputs=['text','text'],
outputs='text',
title="SEO Optimised Title Suggestion",
).launch()
|