Spaces:
Sleeping
Sleeping
Download app.py from Pavan97/emotion_analysis_latest: direct link, hf CLI and curl.
- Browser
- Download file 4.6 kB
-
https://huggingface.co/spaces/Pavan97/emotion_analysis_latest/resolve/main/app.py
- Command line
-
hf download hf://spaces/Pavan97/emotion_analysis_latest/app.py
-
curl -L -o app.py https://huggingface.co/spaces/Pavan97/emotion_analysis_latest/resolve/main/app.py
4.6 kB
| import gradio as gr | |
| import torch | |
| import spaces | |
| from transformers import ( | |
| AutoModelForSequenceClassification, | |
| AutoTokenizer | |
| ) | |
| # Hugging Face model repository | |
| MODEL_ID = "Pavan97/pavan_EA_us_latest" | |
| # Optimized validation threshold for V1 | |
| THRESHOLD = 0.25 | |
| # Emotion label mapping | |
| id2label = { | |
| 0: "admiration", | |
| 1: "amusement", | |
| 2: "anger", | |
| 3: "annoyance", | |
| 4: "approval", | |
| 5: "caring", | |
| 6: "confusion", | |
| 7: "curiosity", | |
| 8: "desire", | |
| 9: "disappointment", | |
| 10: "disapproval", | |
| 11: "disgust", | |
| 12: "embarrassment", | |
| 13: "excitement", | |
| 14: "fear", | |
| 15: "gratitude", | |
| 16: "grief", | |
| 17: "joy", | |
| 18: "love", | |
| 19: "nervousness", | |
| 20: "optimism", | |
| 21: "pride", | |
| 22: "realization", | |
| 23: "relief", | |
| 24: "remorse", | |
| 25: "sadness", | |
| 26: "surprise", | |
| 27: "neutral" | |
| } | |
| label2id = {v: k for k, v in id2label.items()} | |
| # Load model once at startup | |
| print("Loading emotion analysis model...") | |
| tokenizer = AutoTokenizer.from_pretrained( | |
| MODEL_ID | |
| ) | |
| model = AutoModelForSequenceClassification.from_pretrained( | |
| MODEL_ID | |
| ) | |
| if model.config.num_labels != 28: | |
| raise ValueError( | |
| f"Expected 28 emotion labels, " | |
| f"but model has {model.config.num_labels}." | |
| ) | |
| # Keep model on CPU between ZeroGPU requests | |
| model.to("cpu") | |
| model.eval() | |
| print("Model loaded successfully!") | |
| # ZeroGPU prediction function | |
| def predict(text): | |
| if not text or not text.strip(): | |
| return {}, "Please enter some text." | |
| device = torch.device( | |
| "cuda" if torch.cuda.is_available() else "cpu" | |
| ) | |
| try: | |
| model.to(device) | |
| # Tokenize input text | |
| tokenized_text = tokenizer( | |
| text, | |
| return_tensors="pt", | |
| truncation=True, | |
| max_length=256 | |
| ) | |
| tokenized_text = { | |
| key: value.to(device) | |
| for key, value in tokenized_text.items() | |
| } | |
| # Perform multi-label inference | |
| with torch.inference_mode(): | |
| outputs = model(**tokenized_text) | |
| probabilities = torch.sigmoid( | |
| outputs.logits[0] | |
| ).cpu().tolist() | |
| # Convert predictions to dictionary | |
| predicted_labels = { | |
| id2label[idx]: float(score) | |
| for idx, score in enumerate(probabilities) | |
| } | |
| # Sort by highest probability | |
| predicted_labels = dict( | |
| sorted( | |
| predicted_labels.items(), | |
| key=lambda x: x[1], | |
| reverse=True | |
| ) | |
| ) | |
| # Select emotions above threshold | |
| detected_emotions = [ | |
| emotion | |
| for emotion, score in predicted_labels.items() | |
| if score >= THRESHOLD | |
| ] | |
| if detected_emotions: | |
| summary = "Detected emotions: " + ", ".join( | |
| detected_emotions | |
| ) | |
| else: | |
| summary = ( | |
| "No emotions exceeded the " | |
| f"{THRESHOLD} threshold." | |
| ) | |
| return predicted_labels, summary | |
| except Exception as e: | |
| print("Prediction error:", str(e)) | |
| raise gr.Error( | |
| "Prediction failed. Check the Space logs." | |
| ) | |
| finally: | |
| # Release GPU memory after each request | |
| model.to("cpu") | |
| if device.type == "cuda": | |
| torch.cuda.empty_cache() | |
| # Gradio Interface | |
| iface = gr.Interface( | |
| fn=predict, | |
| inputs=gr.Textbox( | |
| lines=3, | |
| placeholder="Enter text to analyze emotions...", | |
| label="Input Text" | |
| ), | |
| outputs=[ | |
| gr.Label( | |
| num_top_classes=5, | |
| label="Predicted Emotions" | |
| ), | |
| gr.Textbox( | |
| label="Detected Emotions" | |
| ) | |
| ], | |
| title="Emotion Analysis 2026", | |
| description=( | |
| "Analyze text using a fine-tuned RoBERTa model " | |
| "trained to recognize 28 different emotions. " | |
| "Uses multi-label classification with an " | |
| "optimized prediction threshold of 0.25." | |
| ), | |
| examples=[ | |
| ["Thanks a lot for watching the video. I really appreciate it."], | |
| ["I am so angry that they delayed the flight again!"], | |
| ["Wow, that is absolutely fascinating! Tell me more."], | |
| ["I feel nervous about my interview tomorrow."], | |
| ["I am really happy with my achievements!"], | |
| ["I am excited about my new job but also nervous."] | |
| ], | |
| flagging_mode="never", | |
| cache_examples=False | |
| ) | |
| # Launch application | |
| if __name__ == "__main__": | |
| iface.launch() | |