app.py

import gradio as gr
from transformers import AutoModelForSequenceClassification, AutoTokenizer
import torch
from sentence_transformers import SentenceTransformer, models
param_max_length=256

# Define a function that takes a text input and returns the result
def analyze_text(input):
    # Your processing or model inference code here
    result = predict_similarity(input)
    return result

param_model_name="CAMeL-Lab/bert-base-arabic-camelbert-msa-sixteenth"

tokenizer = AutoTokenizer.from_pretrained(param_model_name)

class BertForSTS(torch.nn.Module):

    def __init__(self):
        super(BertForSTS, self).__init__()
        #self.bert = models.Transformer('bert-base-uncased', max_seq_length=128)
        #self.bert = AutoModelForSequenceClassification.from_pretrained("CAMeL-Lab/bert-base-arabic-camelbert-msa-sixteenth")
        self.bert = models.Transformer(param_model_name, max_seq_length=param_max_length)


        dimension= self.bert.get_word_embedding_dimension()
        #print(dimension)
        self.pooling_layer = models.Pooling(dimension)
        self.dropout = torch.nn.Dropout(0.1)

        # relu activation function
        self.relu =  torch.nn.ReLU()

        # dense layer 1
        self.fc1 = torch.nn.Linear(dimension,512)

        # dense layer 2 (Output layer)
        self.fc2 = torch.nn.Linear(512,512)
        #self.pooling_layer = models.Pooling(self.bert.config.hidden_size)
        self.sts_bert = SentenceTransformer(modules=[self.bert,self.pooling_layer, self.fc1])
        #self.sts_bert = SentenceTransformer(modules=[self.bert,self.pooling_layer, self.fc1, self.relu, self.dropout,self.fc2])
    def forward(self, input_data):
        #print(input_data)
        x=self.bert(input_data)
        x=self.pooling_layer(x)
        x=self.fc1(x['sentence_embedding'])
        x = self.relu(x)
        x = self.dropout(x)
        #x = self.fc2(x)

        return x
model_load_path = "IBounhas/riadh/bert-sts-15.pt"
model = BertForSTS()
model.load_state_dict(torch.load(model_load_path, map_location=torch.device('cpu')))
model.to(device)

def predict_similarity(sentence_pair):
  test_input = tokenizer(sentence_pair, padding='max_length', max_length = param_max_length, truncation=True, return_tensors="pt").to(device)
  test_input['input_ids'] = test_input['input_ids']
  test_input['attention_mask'] = test_input['attention_mask']
  del test_input['token_type_ids']
  output = model(test_input)
  sim = torch.nn.functional.cosine_similarity(output[0], output[1], dim=0).item()*2-1

  return sim

# Create a Gradio interface with a text input zone
iface = gr.Interface(
    fn=analyze_text,          # The function to be called with user input
    inputs=[gr.Textbox(), gr.Textbox()],
    outputs="text"            # Display the result as text
)

# Launch the Gradio interface
iface.launch()