File size: 1,236 Bytes
231b4b5 29a0778 6ef97ad 29a0778 231b4b5 29a0778 6ef97ad 4a2e7c0 9c9e3b5 4a2e7c0 231b4b5 fc6e0e6 555dc8c 0a7a5c3 231b4b5 5546b1b 231b4b5 6ef97ad 231b4b5 |
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 |
from transformers import AutoTokenizer, AutoModelForCausalLM
import torch
import gradio as gr
import random
tokenizer = AutoTokenizer.from_pretrained('TrLOX/gpt2-tdk')
model = AutoModelForCausalLM.from_pretrained('TrLOX/gpt2-tdk')
def text_generation(keywords, domain):
input_ids = tokenizer('keyword ' + keywords + ' domain ' + domain + ' title', return_tensors="pt").input_ids
torch.manual_seed(random.randint(0,18446744073709551615))
outputs = model.generate(input_ids, do_sample=True, min_length=50, max_length=250)
generated_text = tokenizer.batch_decode(outputs, skip_special_tokens=True)
title_description_arr = generated_text[0].split(' title ')[1].split('description')
title = title_description_arr[0].strip()
description = title_description_arr[1].strip()
return title + "\r\n\r\n" + description
title = "TDK GPT2"
description = "Title and description generation by keywords"
gr.Interface(
text_generation,
[gr.inputs.Textbox(default='test 1,test 2',lines=2, label="Enter keywords"), gr.inputs.Textbox(lines=2, default='test.com',label="Enter domain")],
[gr.outputs.Textbox(type="auto", label="Text Generated")],
title=title,
description=description,
theme="huggingface"
).launch() |