feat: topic categorizer on spaces
Browse files- app.py +31 -0
- requirement.txt +0 -0
app.py
ADDED
|
@@ -0,0 +1,31 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import gradio as gr
|
| 2 |
+
|
| 3 |
+
from transformers import AutoTokenizer, AutoModelForSequenceClassification
|
| 4 |
+
|
| 5 |
+
tokenizer = AutoTokenizer.from_pretrained("WebOrganizer/TopicClassifier-NoURL")
|
| 6 |
+
model = AutoModelForSequenceClassification.from_pretrained(
|
| 7 |
+
"WebOrganizer/TopicClassifier-NoURL",
|
| 8 |
+
trust_remote_code=True,
|
| 9 |
+
use_memory_efficient_attention=False)
|
| 10 |
+
|
| 11 |
+
def predict(text):
|
| 12 |
+
inputs = tokenizer([text], return_tensors="pt")
|
| 13 |
+
outputs = model(**inputs)
|
| 14 |
+
probs = outputs.logits.softmax(dim=-1)
|
| 15 |
+
pred_index = probs.argmax(dim=-1).item()
|
| 16 |
+
confidence_score = probs[0, pred_index]
|
| 17 |
+
id2label = model.config.id2label
|
| 18 |
+
pred_label = id2label[pred_index]
|
| 19 |
+
|
| 20 |
+
return {'topic': pred_label, 'confidence': confidence_score}
|
| 21 |
+
|
| 22 |
+
title = "URL content Topic Categorizer"
|
| 23 |
+
|
| 24 |
+
topic = gr.Interface(
|
| 25 |
+
fn=predict,
|
| 26 |
+
inputs='text',
|
| 27 |
+
outputs='label',
|
| 28 |
+
title=title,
|
| 29 |
+
)
|
| 30 |
+
|
| 31 |
+
topic.launch()
|
requirement.txt
ADDED
|
File without changes
|