File size: 741 Bytes
eb91edf |
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 |
import streamlit as st
import torch
from transformers import GPT2LMHeadModel, GPT2Tokenizer
model_path = "finetuned_model/"
model_name = "sberbank-ai/rugpt3small_based_on_gpt2"
tokenizer = GPT2Tokenizer.from_pretrained(model_name)
model = GPT2LMHeadModel.from_pretrained(model_path)
promt = st.text_input("Ask a question")
generate = st.button("Generate")
if generate:
if not promt:
st.write("42")
promt = tokenizer.encode(promt, return_tensors="pt")
model.eval()
with torch.no_grad():
out = model.generate(
promt,
do_sample=True,
num_beams=2,
temperature=1.5,
top_p=0.9,
)
out = list(map(tokenizer.decode, out))[0]
st.write(out)
|