Spaces:
Sleeping
Sleeping
File size: 1,799 Bytes
9718b31 deb5853 9718b31 deb5853 9718b31 1468b43 9718b31 eef9168 9718b31 eef9168 9718b31 eef9168 9718b31 1468b43 9718b31 |
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 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 |
import wandb
import torch
import re
import os
import gradio
from transformers import GPT2Tokenizer,GPT2LMHeadModel
os.environ["WANDB_API_KEY"] = "d2ad0a7285379c0808ca816971d965fc242d0b5e"
wandb.login()
run = wandb.init(project="Question_Answer", job_type="model_loading", id='xeew4vz7', resume="must")
artifact = run.use_artifact('Question_Answer/final_model_QA:v0')
#artifact = run.use_artifact('enron-subgen-gpt2/model-1hhufzjv:v0')
# Download the artifact to a directory
artifact_dir = artifact.download()
MODEL_KEY = 'distilgpt2'
tokenizer= GPT2Tokenizer.from_pretrained(MODEL_KEY)
tokenizer.add_special_tokens({'pad_token':'{PAD}'})
model = GPT2LMHeadModel.from_pretrained(artifact_dir)
model.resize_token_embeddings(len(tokenizer))
def clean_text(text):
# Lowercase the text
res = re.sub(r'\d', '', text)
text = text.lower()
# Remove special characters
text = re.sub(r'\W', ' ', text)
# Remove extra white spaces
text = re.sub(r'\s+', ' ', text).strip()
return text
def generateAnswer(question):
question = question.strip()
question = "<question>" + clean_text(question) + "<answer>"
prompt = []
prompt.append(question)
tokenizer.padding_side='left'
prompts_batch_ids = tokenizer(prompt,
padding=True, truncation=True, return_tensors='pt').to(model.device)
output_ids = model.generate(
**prompts_batch_ids, max_new_tokens=50,
pad_token_id=tokenizer.pad_token_id)
outputs_batch = [seq.split('<answer>')[1] for seq in
tokenizer.batch_decode(output_ids, skip_special_tokens=True)]
print(outputs_batch)
tokenizer.padding_side='right'
return outputs_batch[0]
iface = gradio.Interface(fn=generateAnswer, inputs="text", outputs="text")
iface.launch()
|