I have a fully functional LSTM neural network in Keras. I found ignite as the closest alternative to Keras for PyTorch(which I have to use instead of Tensorflow due to some requirements), and need to implement exactly the same model in Ignite. I have my training data as a CSV file, and am attaching the code that needs to be converted below.
import keras import pandas as pd import numpy as np import csv from unidecode import unidecode from keras.preprocessing.text import Tokenizer def _removeNonAscii(s): return "".join(i for i in s if ord(i)<128) x_train =  y_train =  with open('/Users/anubhav/Desktop/dataset_/final.csv') as f: reader = csv.reader(f) for row in reader: x_train.append(_removeNonAscii(str(row))) y_train.append(str(row)) tk = Tokenizer() tk.fit_on_texts(x_train) xtrain = tk.texts_to_sequences(x_train) flattened =  for b in xtrain: for t in b: flattened.append(t) flattened npFlat = np.asarray(flattened) ytrain =  for i in range(0, (len(y_train)-1)): try: ytrain.append(int(y_train[i])) except ValueError: print("Error", i) vocabulary_size = len(tk.word_index)+1 print(vocabulary_size) model = keras.Sequential() model.add(keras.layers.Embedding(input_dim=vocabulary_size+1, output_dim=50)) model.add(keras.layers.LSTM(units=50,return_sequences=True)) model.add(keras.layers.LSTM(units=10)) model.add(keras.layers.Dropout(0.5)) model.add(keras.layers.Dense(8)) model.add(keras.layers.Dense(1, activation="sigmoid")) model.compile(optimizer='adam', loss='mean_squared_error') yTRain = np.asarray(ytrain) trainingReshape = np.zeros(401,) target = np.append(yTRain, trainingReshape) model.fit(npFlat, target, epochs=29, batch_size=7, callbacks=[keras.callbacks.EarlyStopping(patience=5)])
How would I convert this Keras model to Pytorch Ignite? This is really urgent, and help would be much appreciated.