Spaces:
Build error
Build error
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,54 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import torch
|
2 |
+
from transformers import DistilBertTokenizer, DistilBertForSequenceClassification
|
3 |
+
|
4 |
+
def load_model(model_path, device):
|
5 |
+
model = DistilBertForSequenceClassification.from_pretrained(model_path)
|
6 |
+
model.to(device)
|
7 |
+
model.eval()
|
8 |
+
return model
|
9 |
+
|
10 |
+
def run_inference(model, tokenizer, label_decoder, device, user_input):
|
11 |
+
model.eval() # Set the model to evaluation mode
|
12 |
+
|
13 |
+
# user_input = input("Enter a text for prediction: ")
|
14 |
+
|
15 |
+
# Tokenize user input
|
16 |
+
input_ids = tokenizer.encode(user_input, return_tensors="pt").to(device)
|
17 |
+
|
18 |
+
with torch.no_grad():
|
19 |
+
outputs = model(input_ids)
|
20 |
+
predicted_label = torch.argmax(outputs.logits, dim=1).tolist()
|
21 |
+
|
22 |
+
# Extracting the text and predicted outcome
|
23 |
+
input_text = tokenizer.decode(input_ids[0], skip_special_tokens=True)
|
24 |
+
predicted_outcome = label_decoder[predicted_label[0]]
|
25 |
+
|
26 |
+
# Display the results
|
27 |
+
print(f"Text: {input_text}")
|
28 |
+
print(f"Predicted Outcome: {predicted_outcome}")
|
29 |
+
print()
|
30 |
+
return predicted_outcome # Add a new line for better readability
|
31 |
+
|
32 |
+
# Example usage
|
33 |
+
model_path = "/home/lwasinam/AI_Projects/hate_speech_detection/model6" # Replace with the actual path to your model
|
34 |
+
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
35 |
+
tokenizer = DistilBertTokenizer.from_pretrained("distilbert-base-uncased") # Replace with your desired tokenizer
|
36 |
+
|
37 |
+
# Load model
|
38 |
+
model = load_model(model_path, device)
|
39 |
+
|
40 |
+
|
41 |
+
label_decoder = {0: "Not Hate", 1: "Hate",}
|
42 |
+
|
43 |
+
|
44 |
+
# Assuming you have label_decoder defined
|
45 |
+
|
46 |
+
|
47 |
+
import streamlit as st
|
48 |
+
|
49 |
+
st.title("Hate Speech Detection")
|
50 |
+
|
51 |
+
user_input = st.text_input("Enter your text:")
|
52 |
+
if user_input:
|
53 |
+
result = run_inference(model, tokenizer, label_decoder, device, user_input)
|
54 |
+
st.write("Inference Result:", result)
|