| from transformers import BertTokenizer, TFBertModel |
| import keras |
|
|
| def predict(data: str): |
| tokenizer = BertTokenizer.from_pretrained('./src/assets/') |
| model = keras.models.load_model('./src/model/scam_class.h5',custom_objects={"TFBertModel": TFBertModel}) |
| encoded = tokenizer(text=data,add_special_tokens=True,max_length=50,padding='max_length', |
| truncation=True,return_tensors='tf',return_token_type_ids=False,verbose=True,return_attention_mask=True) |
| input_obj = {'input_ids': encoded['input_ids'], 'attention_mask': encoded['attention_mask']} |
| prediction = model.predict(input_obj) |
| return prediction |
|
|
| result = predict("Mainkan terus games mu sekarang dan beli koinnya pakai pulsamu, cek caranya di http://tsel.me/jajanonline") |
| print(result) |
