Upload 4 files

app.py

@@ -0,0 +1,58 @@
+
+import torch
+import gradio as gr
+from model import SentimentAnalysisModel
+
+# Load the pre-trained sentiment analysis model
+model = SentimentAnalysisModel(bert_model_name="SamLowe/roberta-base-go_emotions", num_labels=7)
+model.load_state_dict(torch.load("best_model_0.75.pth", map_location=torch.device('cpu')))
+model.eval()
+
+# Mapping from predicted class to emoji
+emoji_to_emotion = {
+    0: 'joy 😆',
+    1: 'fear 😱',
+    2: 'anger 😡',
+    3: 'sadness 😭',
+    4: 'disgust 🤮',
+    5: 'shame 😳',
+    6: 'guilt 😞'
+}
+
+# Function to make predictions
+def predict_sentiment(text):
+    inputs = model.tokenizer(text, return_tensors="pt", truncation=True, padding=True)
+    input_ids = inputs["input_ids"]
+    attention_mask = inputs["attention_mask"]
+
+    with torch.no_grad():
+        outputs = model(input_ids=input_ids, attention_mask=attention_mask)
+
+    logits = outputs.logits
+    _, predicted_class = torch.max(logits, dim=1)
+    
+    # Map predicted class to emoji
+    result = emoji_to_emotion[predicted_class.item()]
+    
+    return result
+
+# Create title, description and article strings
+title = "Emoji-aware Sentiment Analysis using Roberta Model"
+description = "Explore the power of sentiment analysis with our Emotion Detector! Simply input a sentence or text, and let our model predict the underlying emotion. Discover the magic of AI in understanding human sentiments."
+article = "Sentiment Analysis, also known as opinion mining, is a branch of Natural Language Processing (NLP) that involves determining the emotional tone behind a piece of text. This powerful tool allows us to uncover the underlying feelings, attitudes, and opinions expressed in written communication."
+
+
+# Interface for Gradio
+iface = gr.Interface(
+    fn=predict_sentiment,
+    inputs="text",
+    outputs="text",
+    live=True,
+    theme="huggingface",
+    interpretation="default",
+    title=title,
+    description=description,
+    article=article)
+
+# Launch the Gradio interface
+iface.launch()

best_model_75.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:05496d2e0ba1ddb147181357c259ece6f7bcced221e259c48acdf503d1c1207a
+size 1496074346

model.py

@@ -0,0 +1,22 @@
+
+import torch
+from transformers import AutoModelForSequenceClassification, AutoTokenizer
+
+class SentimentAnalysisModel(torch.nn.Module):
+    def __init__(self, bert_model_name, num_labels=7, dropout_rate=0.4):
+        super(SentimentAnalysisModel, self).__init__()
+        self.model = AutoModelForSequenceClassification.from_pretrained(bert_model_name)
+        self.model.classifier.out_proj = torch.nn.Linear(self.model.config.hidden_size, num_labels)
+        self.model.classifier.dropout = torch.nn.Dropout(p=dropout_rate, inplace=False)
+        # ปรับ dropout ในทุก RobertaLayer
+        for layer in self.model.roberta.encoder.layer:
+            layer.attention.self.dropout = torch.nn.Dropout(p=dropout_rate)
+            layer.attention.output.dropout = torch.nn.Dropout(p=dropout_rate)
+            layer.intermediate.dropout = torch.nn.Dropout(p=dropout_rate)
+            layer.output.dropout = torch.nn.Dropout(p=dropout_rate)
+
+        self.tokenizer = AutoTokenizer.from_pretrained(bert_model_name)
+
+    def forward(self, input_ids, attention_mask, labels=None):
+        outputs = self.model(input_ids, attention_mask=attention_mask, labels=labels)
+        return outputs

requirements.txt

@@ -0,0 +1,3 @@
+torch==1.10.0
+transformers==4.11.3
+gradio==2.4.0