Surajnitp commited on
Commit
686a3d6
ยท
verified ยท
1 Parent(s): 9217053

Create app.py

Browse files
Files changed (1) hide show
  1. app.py +199 -0
app.py ADDED
@@ -0,0 +1,199 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """
2
+ Trackmodel - Gradio Web App
3
+ This app loads your fine-tuned model and provides a web interface
4
+ """
5
+
6
+ import gradio as gr
7
+ from transformers import pipeline
8
+ import subprocess
9
+ import os
10
+
11
+ # Global variable to store the classifier
12
+ classifier = None
13
+
14
+
15
+ def load_model():
16
+ """Load the fine-tuned model"""
17
+ global classifier
18
+
19
+ try:
20
+ # Try loading from local first
21
+ if os.path.exists("./my_finetuned_model"):
22
+ classifier = pipeline(
23
+ "sentiment-analysis",
24
+ model="./my_finetuned_model",
25
+ tokenizer="./my_finetuned_model",
26
+ )
27
+ return "โœ… Loaded fine-tuned model from local"
28
+
29
+ # Fall back to Hugging Face Hub
30
+ model_name = "Surajnitp/sentiment-trackmodel"
31
+ classifier = pipeline(
32
+ "sentiment-analysis",
33
+ model=model_name,
34
+ tokenizer=model_name,
35
+ )
36
+ return f"โœ… Loaded model from Hub: {model_name}"
37
+
38
+ except Exception as e:
39
+ # Fall back to default model
40
+ classifier = pipeline(
41
+ "sentiment-analysis",
42
+ model="distilbert-base-uncased-finetuned-sst-2-english",
43
+ )
44
+ return f"โš ๏ธ Using default model (fine-tuned model not found): {str(e)}"
45
+
46
+
47
+ def predict(text):
48
+ """Predict sentiment of input text"""
49
+ if not text.strip():
50
+ return "โš ๏ธ Please enter some text to analyze."
51
+
52
+ try:
53
+ result = classifier(text)
54
+ label = result[0]['label']
55
+ score = result[0]['score']
56
+
57
+ # Emoji for visual feedback
58
+ emoji = "๐Ÿ˜Š" if label == "POSITIVE" else "๐Ÿ˜ž"
59
+
60
+ return f"""
61
+ ### ๐Ÿ“Š Sentiment Analysis Result
62
+
63
+ {emoji} **Label:** {label}
64
+ ๐ŸŽฏ **Confidence:** {score:.2%}
65
+
66
+ ---
67
+ ๐Ÿ“ **Text analyzed:** "{text[:100]}..."
68
+ """
69
+
70
+ except Exception as e:
71
+ return f"โŒ Error: {str(e)}"
72
+
73
+
74
+ def run_training():
75
+ """Run the training script and return status"""
76
+ try:
77
+ # Run the training script
78
+ result = subprocess.run(
79
+ ["python", "train.py"],
80
+ capture_output=True,
81
+ text=True,
82
+ timeout=600, # 10 minute timeout
83
+ )
84
+
85
+ if result.returncode == 0:
86
+ # Reload the model after training
87
+ load_model()
88
+ return f"""
89
+ โœ… **Training completed successfully!**
90
+
91
+ ๐Ÿ“Š **Output:**
92
+ {result.stdout[:500]}...
93
+
94
+ ๐ŸŽฏ The model has been saved to `./my_finetuned_model`
95
+ """
96
+ else:
97
+ return f"""
98
+ โŒ **Training failed!**
99
+
100
+ ๐Ÿ“Š **Error:**
101
+ {result.stderr[:500]}...
102
+ """
103
+
104
+ except subprocess.TimeoutExpired:
105
+ return "โฐ Training timed out after 10 minutes."
106
+
107
+ except Exception as e:
108
+ return f"โŒ Unexpected error: {str(e)}"
109
+
110
+
111
+ # Load the model on startup
112
+ load_status = load_model()
113
+
114
+ # Create the Gradio interface
115
+ with gr.Blocks(
116
+ title="Trackmodel - Sentiment Analysis",
117
+ theme=gr.themes.Soft(),
118
+ css=".gradio-container {max-width: 800px; margin: auto;}"
119
+ ) as demo:
120
+
121
+ gr.Markdown("""
122
+ # ๐ŸŽฏ Trackmodel - Sentiment Analysis
123
+
124
+ This Space demonstrates downloading, fine-tuning, and deploying models on Hugging Face.
125
+
126
+ **Model:** DistilBERT fine-tuned on IMDb reviews for sentiment classification.
127
+ """)
128
+
129
+ # Status indicator
130
+ gr.Markdown(f"**Status:** {load_status}")
131
+
132
+ with gr.Tab("๐Ÿ’ฌ Predict"):
133
+ gr.Markdown("### Enter text to analyze sentiment")
134
+
135
+ with gr.Row():
136
+ with gr.Column(scale=4):
137
+ text_input = gr.Textbox(
138
+ label="Your text",
139
+ placeholder="Type something like: 'I love this movie!'",
140
+ lines=4,
141
+ )
142
+ with gr.Column(scale=1):
143
+ predict_btn = gr.Button("๐Ÿ”ฎ Predict", variant="primary", size="lg")
144
+ clear_btn = gr.Button("๐Ÿ—‘๏ธ Clear", variant="secondary")
145
+
146
+ output = gr.Markdown(
147
+ label="Result",
148
+ value="๐Ÿ“ Enter text and click 'Predict' to see results.",
149
+ )
150
+
151
+ with gr.Tab("โš™๏ธ Training"):
152
+ gr.Markdown("""
153
+ ### Train the Model on Custom Data
154
+
155
+ Click the button below to fine-tune the model on the IMDb dataset.
156
+
157
+ โš ๏ธ **Note:** This will take ~3-5 minutes on a CPU instance.
158
+ """)
159
+
160
+ with gr.Row():
161
+ train_btn = gr.Button("๐Ÿš€ Start Training", variant="primary", size="lg")
162
+
163
+ train_output = gr.Markdown("โ„น๏ธ Click the button to start training.")
164
+
165
+ with gr.Tab("โ„น๏ธ About"):
166
+ gr.Markdown("""
167
+ ## About Trackmodel
168
+
169
+ ### What this Space does:
170
+ 1. **Downloads** a pre-trained DistilBERT model
171
+ 2. **Fine-tunes** it on the IMDb movie review dataset
172
+ 3. **Deploys** the trained model as a web app
173
+
174
+ ### How to use:
175
+ - Go to the **Predict** tab to analyze text sentiment
176
+ - Go to the **Training** tab to fine-tune the model yourself
177
+
178
+ ### Links:
179
+ - [Hugging Face Models](https://huggingface.co/models)
180
+ - [Gradio Documentation](https://gradio.app/docs)
181
+ - [Hugging Face Course](https://huggingface.co/learn/nlp-course)
182
+
183
+ ### Author:
184
+ **Surajnitp** ๐Ÿš€
185
+ """)
186
+
187
+ # Connect event handlers
188
+ predict_btn.click(fn=predict, inputs=text_input, outputs=output)
189
+ clear_btn.click(fn=lambda: "", inputs=None, outputs=text_input)
190
+ clear_btn.click(
191
+ fn=lambda: "๐Ÿ“ Enter text and click 'Predict' to see results.",
192
+ inputs=None,
193
+ outputs=output
194
+ )
195
+ train_btn.click(fn=run_training, inputs=None, outputs=train_output)
196
+
197
+ # Launch the app
198
+ if __name__ == "__main__":
199
+ demo.launch()