Spaces:

pyamath
/

chatcot

Sleeping

App Files Files Community

chatcot / app1.py

pyamath

Rename app.py to app1.py

8a9b515 7 months ago

raw history blame contribute delete

No virus

3.5 kB

	import streamlit as st
	import numpy as np
	import MeCab
	from transformers import BertJapaneseTokenizer, BertModel
	from sentence_transformers import SentenceTransformer
	from sentence_transformers import models
	import torch
	from torch.nn.functional import cosine_similarity

	# 日本語対応パッケージのインストール

	st.title("質問箱")

	# 定数定義
	USER_NAME = "user"
	ASSISTANT_NAME = "assistant"
	MORIAGE_YAKU_NAME = "moriage_yaku"
	MORIAGE_YAKU2_NAME = "moriage_yaku2"
	MODEL_NAME = 'cl-tohoku/bert-base-japanese-whole-word-masking'

	tokenizer = BertJapaneseTokenizer.from_pretrained(MODEL_NAME)
	model = BertModel.from_pretrained(MODEL_NAME)

	def sentence_to_vector(model, tokenizer, sentence):

	# 文を単語に区切って数字にラベル化
	tokens = tokenizer(sentence)["input_ids"]

	# BERTモデルの処理のためtensor型に変換
	input = torch.tensor(tokens).reshape(1,-1)

	# BERTモデルに入力し文のベクトルを取得
	with torch.no_grad():
	outputs = model(input, output_hidden_states=True)
	last_hidden_state = outputs.last_hidden_state[0]
	averaged_hidden_state = last_hidden_state.sum(dim=0) / len(last_hidden_state)

	return averaged_hidden_state

	def calc_similarity(sentence1, sentence2):
	st.write(sentence1, sentence2)
	sentence_vector1 = sentence_to_vector(model, tokenizer, sentence1)
	sentence_vector2 = sentence_to_vector(model, tokenizer, sentence2)

	score = torch.nn.functional.cosine_similarity(sentence_vector1, sentence_vector2, dim=0).detach().numpy().copy()
	st.write("類似度：", score,sentence_vector1,sentence_vector2)
	# チャットログを保存したセッション情報を初期化
	if "chat_log" not in st.session_state:
	st.session_state.chat_log = []

	# ユーザーのアバターを設定
	# img_moriyage_yaku2 = np.array(Image.open("moriage_yaku2.jpeg"))
	avator_img_dict = {
	MORIAGE_YAKU_NAME: "🎉",
	# MORIAGE_YAKU2_NAME: img_moriyage_yaku2,
	}

	wakati = MeCab.Tagger("-Owakati")

	user_msg = st.chat_input("質問、要望等あれば入力してください")
	if user_msg:
	a = wakati.parse(user_msg).split()
	sentence1 = "吾輩は猫である"
	sentence2 = "私は猫です"
	calc_similarity(sentence1, sentence2)
	# 以前のチャットログを表示
	for chat in st.session_state.chat_log:
	avator = avator_img_dict.get(chat["name"], None)
	with st.chat_message(chat["name"], avatar=avator):
	st.write(chat["msg"])

	# 最新のメッセージを表示
	assistant_msg = "もう一度入力してください"
	moriage_yaku_msg = "アンコール！アンコール！"
	moriage_yaku2_msg = "そっれ、アンコール！アンコール！"
	with st.chat_message(USER_NAME):
	st.write(user_msg)
	with st.chat_message(ASSISTANT_NAME):
	st.write(a)
	with st.chat_message(MORIAGE_YAKU_NAME, avatar=avator_img_dict[MORIAGE_YAKU_NAME]):
	st.write(moriage_yaku_msg)
	# with st.chat_message(
	# MORIAGE_YAKU2_NAME,
	# avatar=avator_img_dict[MORIAGE_YAKU2_NAME],
	# ):
	# st.write(moriage_yaku2_msg)

	# セッションにチャットログを追加
	st.session_state.chat_log.append({"name": USER_NAME, "msg": user_msg})
	st.session_state.chat_log.append({"name": ASSISTANT_NAME, "msg": user_msg})
	st.session_state.chat_log.append({"name": MORIAGE_YAKU_NAME, "msg": user_msg})
	# st.session_state.chat_log.append({"name": MORIAGE_YAKU2_NAME, "msg": user_msg})