update

app.py

@@ -12,34 +12,67 @@ app = FastAPI()
 
 # ---------------------------------------------------------------------------------------------------------------------
 import gradio as gr
-import os
-import torch
-import json
 import tqdm
-from time import sleep
+import torch
 import numpy as np
-import requests
-from rank_bm25 import BM25Okapi
-from bs4 import BeautifulSoup
+from time import sleep
 from datetime import datetime
+import threading
+import gc
+import os
+import json
+import pytorch_lightning as pl
+from urllib.parse import urlparse
+from accelerate import Accelerator
+import spaces
 
 from transformers import BartTokenizer, BartForConditionalGeneration
 from transformers import BloomTokenizerFast, BloomForCausalLM, BertTokenizer, BertForSequenceClassification
 from transformers import RobertaTokenizer, RobertaForSequenceClassification
-import pytorch_lightning as pl
 
-from urllib.parse import urlparse
+from rank_bm25 import BM25Okapi
+# import bm25s
+# import Stemmer  # optional: for stemming
 from html2lines import url2lines
 from googleapiclient.discovery import build
 from averitec.models.DualEncoderModule import DualEncoderModule
 from averitec.models.SequenceClassificationModule import SequenceClassificationModule
 from averitec.models.JustificationGenerationModule import JustificationGenerationModule
+from averitec.data.sample_claims import CLAIMS_Type
+
+# ---------------------------------------------------------------------------
+# load .env
+from utils import create_user_id
+user_id = create_user_id()
+
+from azure.storage.fileshare import ShareServiceClient
+try:
+    from dotenv import load_dotenv
+    load_dotenv()
+except Exception as e:
+    pass
+
+# ---------------------------------------------------------------------------
+# os.environ["TOKENIZERS_PARALLELISM"] = "false"
+account_url = os.environ["AZURE_ACCOUNT_URL"]
+credential = {
+    "account_key":  os.environ['AZURE_ACCOUNT_KEY'],
+    "account_name": os.environ['AZURE_ACCOUNT_NAME']
+}
+
+file_share_name = "averitec"
+azure_service = ShareServiceClient(account_url=account_url, credential=credential)
+azure_share_client = azure_service.get_share_client(file_share_name)
 
 # ---------------------------------------------------------------------------------------------------------------------
+import requests
+from bs4 import BeautifulSoup
 import wikipediaapi
 wiki_wiki = wikipediaapi.Wikipedia('AVeriTeC (zd302@cam.ac.uk)', 'en')
 
 import nltk
+nltk.download('averaged_perceptron_tagger_eng')
+nltk.download('averaged_perceptron_tagger')
 nltk.download('punkt')
 nltk.download('punkt_tab')
 from nltk import pos_tag, word_tokenize, sent_tokenize
@@ -125,30 +158,10 @@ tokenized_corpus1, prompt_corpus1 = generate_step2_reference_corpus(reference_fi
 prompt_bm25 = BM25Okapi(tokenized_corpus1)
 
 # ---------------------------------------------------------------------------------------------------------------------
-# ---------------------------------------------------------------------------
-# load .env
-from utils import create_user_id
-user_id = create_user_id()
-
-from azure.storage.fileshare import ShareServiceClient
-try:
-    from dotenv import load_dotenv
-    load_dotenv()
-except Exception as e:
-    pass
-
-account_url = os.environ["AZURE_ACCOUNT_URL"]
-credential = {
-    "account_key":  os.environ['AZURE_ACCOUNT_KEY'],
-    "account_name": os.environ['AZURE_ACCOUNT_NAME']
-}
-
-file_share_name = "averitec"
-azure_service = ShareServiceClient(account_url=account_url, credential=credential)
-azure_share_client = azure_service.get_share_client(file_share_name)
 
 # ---------- Setting ----------
 # ---------- Load Veracity and Justification prediction model ----------
+print("Loading models ...")
 LABEL = [
     "Supported",
     "Refuted",