File size: 4,603 Bytes
3013601
212fd6e
3013601
aeda638
3013601
aeda638
 
 
 
3013601
 
aeda638
3013601
aeda638
 
3013601
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
aeda638
3013601
 
aeda638
 
3013601
78989d7
aeda638
 
78989d7
212fd6e
3013601
 
 
 
 
 
aeda638
 
3013601
 
 
 
aeda638
 
 
3013601
 
 
aeda638
 
 
 
 
212fd6e
 
aeda638
 
3013601
 
 
 
 
 
 
 
 
 
 
 
 
aeda638
 
3013601
 
aeda638
 
 
3013601
 
aeda638
 
 
 
3013601
 
 
 
 
 
 
 
 
 
aeda638
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3013601
212fd6e
3013601
 
 
 
 
aeda638
 
 
 
 
 
 
 
3013601
 
 
 
 
 
 
 
 
 
aeda638
 
 
 
 
 
 
 
 
3013601
 
 
 
 
aeda638
 
 
3013601
 
124c9e2
3013601
124c9e2
3013601
 
aeda638
 
3013601
 
aeda638
 
212fd6e
 
3013601
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203

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
@spaces.GPU(duration=60)
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()