import torch
import torch.nn as nn
from transformers import BertTokenizer, BertModel, BertForSequenceClassification
import gradio as gr

# Your model repo
model_name = "keethu/bert-emotion-classifier"

# Load tokenizer
tokenizer = BertTokenizer.from_pretrained(model_name)

# Load base BERT model
base_bert = BertModel.from_pretrained(model_name)

# Define your classifier architecture (same as training)
class BERTClassifier(nn.Module):
    def __init__(self, bert_model, num_labels=5, dropout=0.3):
        super(BERTClassifier, self).__init__()
        self.bert = bert_model
        self.dropout = nn.Dropout(dropout)
        self.classifier = nn.Linear(self.bert.config.hidden_size, num_labels)
    
    def forward(self, input_ids, attention_mask):
        outputs = self.bert(input_ids=input_ids, attention_mask=attention_mask)
        pooled_output = outputs.pooler_output
        pooled_output = self.dropout(pooled_output)
        logits = self.classifier(pooled_output)
        return logits

# Create model instance
model = BERTClassifier(base_bert, num_labels=5, dropout=0.3)

# Load the trained weights - USE from_pretrained properly
from huggingface_hub import hf_hub_download
import os

# Download the model file
model_path = hf_hub_download(repo_id=model_name, filename="pytorch_model.bin")

# Load state dict
state_dict = torch.load(model_path, map_location='cpu')
model.load_state_dict(state_dict)
model.eval()

emotion_labels = ['anger', 'fear', 'joy', 'sadness', 'surprise']

def predict_emotions(text):
    # Tokenize input
    inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True, max_length=128)
    
    # Get predictions
    with torch.no_grad():
        outputs = model(inputs['input_ids'], inputs['attention_mask'])
        probs = torch.sigmoid(outputs)[0].numpy()
    
    # Create results dictionary
    results = {emotion_labels[i]: float(probs[i]) for i in range(len(emotion_labels))}
    return results

# Create Gradio interface
iface = gr.Interface(
    fn=predict_emotions,
    inputs=gr.Textbox(lines=3, placeholder="Enter text here to detect emotions..."),
    outputs=gr.Label(num_top_classes=5),
    title="Emotion Classification",
    description="Multi-label emotion detection using fine-tuned BERT. Enter any text to detect anger, fear, joy, sadness, and surprise.",
    examples=[
        ["I am so happy and excited about this!"],
        ["This is terrible and makes me angry."],
        ["I can't believe this happened, it's shocking!"]
    ]
)

iface.launch()