Spaces:
Sleeping
Sleeping
| import re | |
| import os | |
| import tensorflow as tf | |
| import keras | |
| import keras_nlp | |
| MAX_ENCODER_SEQUENCE_LENGTH = 512 | |
| MAX_DECODER_SEQUENCE_LENGTH = 128 | |
| MODEL_PATH = r"TextSummarizationModel" | |
| WEIGHT_PATH = r"new_model.weights.h5" | |
| WEIGHT_PATH = os.path.join(MODEL_PATH, WEIGHT_PATH) | |
| def cleanText(text): | |
| text = str(text) | |
| text = re.sub(r'[^a-zA-Z0-9\s]', '', text) | |
| text = text.lower() | |
| return text | |
| preprocessor = keras_nlp.models.BartSeq2SeqLMPreprocessor.from_preset(MODEL_PATH, encoder_sequence_length=MAX_ENCODER_SEQUENCE_LENGTH,decoder_sequence_length=MAX_DECODER_SEQUENCE_LENGTH,) | |
| model = keras_nlp.models.BartSeq2SeqLM.from_preset(MODEL_PATH, preprocessor=preprocessor) | |
| model.load_weights(WEIGHT_PATH) | |
| def generateText(input_text, model=model, max_length=200): | |
| input_text = cleanText(input_text) | |
| output = model.generate(input_text, max_length=max_length) | |
| return output |