Add GRU model (fingers crossed it works)

app.py

@@ -1,3 +1,6 @@
+import os
+os.environ["CUDA_VISIBLE_DEVICES"] = "-1"  # Disable GPU and enforce CPU execution
+
 import gradio as gr
 from transformers import (
     DistilBertTokenizerFast,
@@ -9,8 +12,11 @@ from huggingface_hub import hf_hub_download
 import torch
 import pickle
 import numpy as np
+from tensorflow.keras.models import load_model
+from tensorflow.keras.preprocessing.sequence import pad_sequences
+import re
 
-# Load models and tokenizers
+# Load pre-trained models and tokenizers
 models = {
     "DistilBERT": {
         "tokenizer": DistilBertTokenizerFast.from_pretrained("nhull/distilbert-sentiment-model"),
@@ -50,7 +56,30 @@ for model_data in models.values():
     if "model" in model_data:
         model_data["model"].to(device)
 
-# Functions for prediction
+# Load GRU model and tokenizer
+gru_repo_id = "arjahojnik/GRU-sentiment-model"
+gru_model_path = hf_hub_download(repo_id=gru_repo_id, filename="best_GRU_tuning_model.h5")
+gru_model = load_model(gru_model_path)
+gru_tokenizer_path = hf_hub_download(repo_id=gru_repo_id, filename="my_tokenizer.pkl")
+with open(gru_tokenizer_path, "rb") as f:
+    gru_tokenizer = pickle.load(f)
+
+# Preprocessing function for GRU
+def preprocess_text(text):
+    text = text.lower()
+    text = re.sub(r"[^a-zA-Z\s]", "", text).strip()
+    return text
+
+# GRU prediction function
+def predict_with_gru(text):
+    cleaned = preprocess_text(text)
+    seq = gru_tokenizer.texts_to_sequences([cleaned])
+    padded_seq = pad_sequences(seq, maxlen=200)  # Ensure maxlen matches the GRU training
+    probs = gru_model.predict(padded_seq)
+    predicted_class = np.argmax(probs, axis=1)[0]
+    return int(predicted_class + 1)
+
+# Functions for other model predictions
 def predict_with_distilbert(text):
     tokenizer = models["DistilBERT"]["tokenizer"]
     model = models["DistilBERT"]["model"]
@@ -99,6 +128,7 @@ def predict_with_roberta_ordek899(text):
 # Unified function for sentiment analysis and statistics
 def analyze_sentiment_and_statistics(text):
     results = {
+        "GRU Model": predict_with_gru(text),
         "DistilBERT": predict_with_distilbert(text),
         "Logistic Regression": predict_with_logistic_regression(text),
         "BERT Multilingual (NLP Town)": predict_with_bert_multilingual(text),
@@ -169,6 +199,7 @@ with gr.Blocks(css=".gradio-container { max-width: 900px; margin: auto; padding:
 
     with gr.Row():
         with gr.Column():
+            gru_output = gr.Textbox(label="Predicted Sentiment (GRU Model)", interactive=False)
             distilbert_output = gr.Textbox(label="Predicted Sentiment (DistilBERT)", interactive=False)
             log_reg_output = gr.Textbox(label="Predicted Sentiment (Logistic Regression)", interactive=False)
             bert_output = gr.Textbox(label="Predicted Sentiment (BERT Multilingual)", interactive=False)
@@ -183,6 +214,7 @@ with gr.Blocks(css=".gradio-container { max-width: 900px; margin: auto; padding:
         results, statistics = analyze_sentiment_and_statistics(text_input)
         if "Message" in statistics:
             return (
+                f"{results['GRU Model']}",
                 f"{results['DistilBERT']}",
                 f"{results['Logistic Regression']}",
                 f"{results['BERT Multilingual (NLP Town)']}",
@@ -192,6 +224,7 @@ with gr.Blocks(css=".gradio-container { max-width: 900px; margin: auto; padding:
             )
         else:
             return (
+                f"{results['GRU Model']}",
                 f"{results['DistilBERT']}",
                 f"{results['Logistic Regression']}",
                 f"{results['BERT Multilingual (NLP Town)']}",
@@ -204,6 +237,7 @@ with gr.Blocks(css=".gradio-container { max-width: 900px; margin: auto; padding:
         process_input_and_analyze,
         inputs=[text_input],
         outputs=[
+            gru_output,
             distilbert_output, 
             log_reg_output, 
             bert_output,