Update main.py

main.py

@@ -33,8 +33,11 @@ from sentence_transformers import SentenceTransformer, util
 from textblob import TextBlob
 import nltk
 
-nltk.download('punkt_tab')
-nltk.download('words')
+data_dir = 'nltk_data'
+
+# Set the NLTK data path to the local directory
+nltk.data.path.append(data_dir)
+
 
 ''''''''''''''''''''''''' Skeletal Structure for the Models '''''''''''''''''''''''''''
 
@@ -89,20 +92,24 @@ class DistilBertForRegression(DistilBertPreTrainedModel):
 device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
 
 
-print("Downloading  pronunciation tokenizer...")
-pronunciation_tokenizer = Wav2Vec2Tokenizer.from_pretrained("facebook/wav2vec2-base-960h")
-print("Downloading  pronunciation model...")
-pronunciation_model = Wav2Vec2ForCTC.from_pretrained("facebook/wav2vec2-base-960h")
+# Load the Pronunciation model and tokenizer from the local directory
+pronunciation_model_dir = 'pronunciation_model'
+fluency_model_dir = 'fluency_model'
+
+print("Loading pronunciation tokenizer from local directory...")
+pronunciation_tokenizer = Wav2Vec2Tokenizer.from_pretrained(pronunciation_model_dir)
 
-pronunciation_model.to(device)
+print("Loading pronunciation model from local directory...")
+pronunciation_model = Wav2Vec2ForCTC.from_pretrained(pronunciation_model_dir)
 
-''''''''''''''''''' Loading the Fluency Model and Tokenizer '''''''''''''
+# Load the Fluency model and tokenizer from the local directory
+print("Loading fluency tokenizer from local directory...")
+fluency_tokenizer = DistilBertTokenizer.from_pretrained(fluency_model_dir)
 
-print("Downloading  fluency tokenizer...")
-fluency_tokenizer = DistilBertTokenizer.from_pretrained('distilbert-base-uncased')
-print("Downloading fluency model...")
-fluency_model = DistilBertForRegression.from_pretrained("Kartikeyssj2/Fluency_Scoring_V2")
-print("Download completed.")
+print("Loading fluency model from local directory...")
+fluency_model = DistilBertForSequenceClassification.from_pretrained(fluency_model_dir)
+
+print("Models loaded successfully.")
 
 fluency_model.to(device)
 # device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
@@ -128,7 +135,17 @@ linreg_pronunciation = load_pickle_file("pronunciation_model_biasing.pkl")
 
 '''''''''''''''''''''' Load the Content Relevance and Scoring Model '''''''''''''''
 
-content_relevance_model = SentenceTransformer('sentence-transformers/msmarco-distilbert-cos-v5')
+
+model_dir = 'content_relevance_model'
+
+# Load the SentenceTransformer model from the local directory
+print("Loading SentenceTransformer model from local directory...")
+model = SentenceTransformer(model_dir)
+
+print("Model loaded successfully.")
+
+
+
 
 print(linreg_fluency)
 print(linreg_pronunciation)
@@ -143,9 +160,20 @@ import torch
 
 ''''''''''''''''''''''' IMAGE CAPTIONING MODEL '''''''''''''''''
 
-image_captioning_processor = BlipProcessor.from_pretrained("noamrot/FuseCap")
+# Define the directories where the models and processors are saved
+processor_dir = 'blip_processor'
+model_dir = 'blip_model'
+
+# Load the BlipProcessor from the local directory
+print("Loading BlipProcessor from local directory...")
+image_captioning_processor = BlipProcessor.from_pretrained(processor_dir)
+print("BlipProcessor loaded successfully.")
 
-image_captioning_model = BlipForConditionalGeneration.from_pretrained("noamrot/FuseCap").to(device)
+# Load the BlipForConditionalGeneration model from the local directory
+print("Loading BlipForConditionalGeneration model from local directory...")
+image_captioning_model = BlipForConditionalGeneration.from_pretrained(model_dir)
+image_captioning_model.to(device)  # Move model to the appropriate device
+print("BlipForConditionalGeneration model loaded successfully.")