main.py

# coding: utf-8

# Sampled-Softmax for PyTorch
# Lei Mao
# University of Chicago

import argparse
import time
import math
import torch
import torch.nn as nn
from torch.autograd import Variable
import torch.optim as optim

import data
import model
import utils

parser = argparse.ArgumentParser(description='PTB RNN/LSTM Language Model: Main Function')
parser.add_argument('--data', type=str, default='./data/ptb',
                    help='location of the data corpus')
parser.add_argument('--model', type=str, default='LSTM',
                    help='type of recurrent net (RNN_TANH, RNN_RELU, LSTM, GRU)')
parser.add_argument('--emsize', type=int, default=100,
                    help='size of word embeddings')
parser.add_argument('--nhid', type=int, default=128,
                    help='number of hidden units per layer')
parser.add_argument('--nlayers', type=int, default=1,
                    help='number of layers')
parser.add_argument('--lr', type=float, default=20,
                    help='initial learning rate')
parser.add_argument('--clip', type=float, default=0.25,
                    help='gradient clipping')
parser.add_argument('--epochs', type=int, default=5,
                    help='upper epoch limit')
parser.add_argument('--batch_size', type=int, default=32, metavar='N',
                    help='batch size')
parser.add_argument('--bptt', type=int, default=35,
                    help='sequence length')
parser.add_argument('--dropout', type=float, default=0.2,
                    help='dropout applied to layers (0 = no dropout)')
parser.add_argument('--tied', action='store_true',
                    help='tie the word embedding and softmax weights')
parser.add_argument('--seed', type=int, default=1111,
                    help='random seed')
parser.add_argument('--log_interval', type=int, default=200, metavar='N',
                    help='report interval')
parser.add_argument('--save', type=str,  default='model.pt',
                    help='path to save the final model')
parser.add_argument('--softmax_nsampled', type=int,  default=100,
                    help='number of random sample generated for sampled softmax')
parser.add_argument('--lr_decay', type=float, default=1.0,
                    help='learning rate decay over each epoch, default 1.0')


args = parser.parse_args()


# Set the random seed manually for reproducibility.
torch.manual_seed(args.seed)

###############################################################################
# Load data
###############################################################################

corpus_raw = data.Corpus(args.data)


# word_freq_sorted: list idx is the rank of the word frequency, list content value is the word_idx
# word_freq_sorted = word_freq_ordered(corpus = corpus_raw)

# word_rank: list idx is the word_idx, list content value is the frequency rank
word_rank = utils.word_rank_dictionary(corpus = corpus_raw)

#######################################################
# We use the word_rank as the input to the model !
#######################################################

corpus = utils.Rand_Idxed_Corpus(corpus = corpus_raw, word_rank = word_rank)



# Starting from sequential data, batchify arranges the dataset into columns.
# For instance, with the alphabet as the sequence and batch size 4, we'd get
# ┌ a g m s ┐
# │ b h n t │
# │ c i o u │
# │ d j p v │
# │ e k q w │
# └ f l r x ┘.
# These columns are treated as independent by the model, which means that the
# dependence of e. g. 'g' on 'f' can not be learned, but allows more efficient
# batch processing.

def batchify(data, bsz):
    # Work out how cleanly we can divide the dataset into bsz parts.
    nbatch = data.size(0) // bsz
    # Trim off any extra elements that wouldn't cleanly fit (remainders).
    data = data.narrow(0, 0, nbatch * bsz)
    # Evenly divide the data across the bsz batches.
    data = data.view(bsz, -1).t().contiguous()
    return data

eval_batch_size = 10
train_data = batchify(corpus.train, args.batch_size)
val_data = batchify(corpus.valid, eval_batch_size)
test_data = batchify(corpus.test, eval_batch_size)

###############################################################################
# Build the model
###############################################################################

ntokens = len(corpus.dictionary)
model = model.RNNModel(args.model, ntokens, args.emsize, args.nhid, args.nlayers, args.dropout, args.tied)
criterion = nn.CrossEntropyLoss()
#decoder = utils.LinearDecoder(args.nhid, ntokens)
encoder = utils.Word2VecEncoder(ntokens, args.emsize, args.dropout)
sampled_softmax = utils.SampledSoftmax(ntokens = ntokens, nsampled = args.softmax_nsampled, nhid = args.nhid, tied_weight = None)


model.add_module("encoder", encoder)
model.add_module("decoder", sampled_softmax)

optimizer = optim.Adagrad(model.parameters(), args.lr, weight_decay=1e-5)

#optimizer = optim.Adam(model.parameters(), args.lr, betas=(0.9, 0.999), eps=1e-08, weight_decay=1e-4)


#optimizer = optim.SGD(model.parameters(), lr=args.lr, momentum=0, dampening=0, weight_decay=1e-4, nesterov=False)


###############################################################################
# Training code
###############################################################################

def repackage_hidden(h):
    """Wraps hidden states in new Variables, to detach them from their history."""
    if type(h) == Variable:
        return Variable(h.data)
    else:
        return tuple(repackage_hidden(v) for v in h)


# get_batch subdivides the source data into chunks of length args.bptt.
# If source is equal to the example output of the batchify function, with
# a bptt-limit of 2, we'd get the following two Variables for i = 0:
# ┌ a g m s ┐ ┌ b h n t ┐
# └ b h n t ┘ └ c i o u ┘
# Note that despite the name of the function, the subdivison of data is not
# done along the batch dimension (i.e. dimension 1), since that was handled
# by the batchify function. The chunks are along dimension 0, corresponding
# to the seq_len dimension in the LSTM.

def get_batch(source, i, evaluation=False):
    seq_len = min(args.bptt, len(source) - 1 - i)
    data = Variable(source[i:i+seq_len], volatile=evaluation)
    target = Variable(source[i+1:i+1+seq_len].view(-1))
    return data, target


def evaluate(data_source):
    # Turn on evaluation mode which disables dropout.
    encoder.eval()
    model.eval()
    #decoder.eval()
    total_loss = 0
    ntokens = len(corpus.dictionary)
    hidden = model.init_hidden(eval_batch_size)
    for i in range(0, data_source.size(0) - 1, args.bptt):
        data, targets = get_batch(data_source, i, evaluation=True)
        emb = encoder(data)
        output, hidden = model(emb, hidden)
        #logits = decoder(output)
        logits = sampled_softmax.full(output.view(-1, output.size(2)))
        #loss = criterion(logits.view(-1, logits.size(-1)), new_targets)

        total_loss += len(data) * criterion(logits.view(-1, logits.size(-1)), targets).data # +1 for the true label
        #total_loss += len(data) * criterion(logits.view(-1, args.softmax_nsampled+1), new_targets).data

        #logits_flat = logits.view(-1, ntokens)
        #total_loss += len(data) * criterion(logits_flat, targets).data
        hidden = repackage_hidden(hidden)
    return total_loss[0] / len(data_source)


def train():
    # Turn on training mode which enables dropout.
    model.train()
    total_loss = 0
    start_time = time.time()
    ntokens = len(corpus.dictionary)
    hidden = model.init_hidden(args.batch_size)
    for batch, i in enumerate(range(0, train_data.size(0) - 1, args.bptt)):
        data, targets = get_batch(train_data, i)

        optimizer.zero_grad()

        #encoder.zero_grad()
        #decoder.zero_grad()
        #sampled_softmax.zero_grad()

        #start_time_forward = time.time()
        emb = encoder(data)
        # Starting each batch, we detach the hidden state from how it was previously produced.
        # If we didn't, the model would try backpropagating all the way to start of the dataset.
        hidden = repackage_hidden(hidden)
        #model.zero_grad()
        output, hidden = model(emb, hidden)
        #logits = decoder(output)
        logits, new_targets = sampled_softmax(output.view(-1, output.size(2)), targets)
        loss = criterion(logits.view(-1, args.softmax_nsampled+1), new_targets)
        #loss = criterion(logits.view(-1, ntokens), targets)
        #print("forward time: %f ms" %(1000*(time.time()-start_time_forward)))
        #print(loss)
        #start_time_backward = time.time()
        loss.backward()

        # `clip_grad_norm` helps prevent the exploding gradient problem in RNNs / LSTMs.
        torch.nn.utils.clip_grad_norm(encoder.parameters(), args.clip)
        #for p in encoder.parameters():
        #    p.data.add_(-lr, p.grad.data)
        torch.nn.utils.clip_grad_norm(model.parameters(), args.clip)
        #for p in model.parameters():
        #    p.data.add_(-lr, p.grad.data)
        torch.nn.utils.clip_grad_norm(sampled_softmax.parameters(), args.clip)
        #for p in sampled_softmax.parameters():
        #    p.data.add_(-lr, p.grad.data)
        #torch.nn.utilss.clip_grad_norm(decoder.parameters(), args.clip)
        #for p in decoder.parameters():
        #    p.data.add_(-lr, p.grad.data)
        #print("backpropagation time: %f ms" %(1000*(time.time()-start_time_backward)))

        optimizer.step()

        total_loss += loss.data

        if batch % args.log_interval == 0 and batch > 0:
            cur_loss = total_loss[0] / args.log_interval
            elapsed = time.time() - start_time
            print('| epoch {:3d} | {:5d}/{:5d} batches | lr {:02.2f} | ms/batch {:5.2f} | '
                    'loss {:5.2f} | perplexity {:8.2f}'.format(
                epoch, batch, len(train_data) // args.bptt, lr,
                elapsed * 1000 / args.log_interval, cur_loss, math.exp(cur_loss)))
            total_loss = 0
            start_time = time.time()

# Loop over epochs.
lr = args.lr
best_val_loss = None

# At any point you can hit Ctrl + C to break out of training early.
try:
    for epoch in range(1, args.epochs+1):
        epoch_start_time = time.time()
        train()
        val_loss = evaluate(val_data)
        print('-' * 89)
        print('| end of epoch {:3d} | time: {:5.2f}s | valid loss {:5.2f} | '
                'valid perplexity {:8.2f}'.format(epoch, (time.time() - epoch_start_time),
                                           val_loss, math.exp(val_loss)))
        print('-' * 89)

        # Learning rate decay
        lr *= args.lr_decay

        # Save the model if the validation loss is the best we've seen so far.
        if not best_val_loss or val_loss < best_val_loss:
            with open(args.save, 'wb') as f:
                model_state = {'model': model, 'encoder': encoder, 'decoder': sampled_softmax}
                # torch.save(model, f)
                torch.save(model_state, f)
            best_val_loss = val_loss
        else:
            # Anneal the learning rate if no improvement has been seen in the validation dataset.
            # lr /= 4.0
            lr = lr
except KeyboardInterrupt:
    print('-' * 89)
    print('Exiting from training early')

# Load the best saved model.
with open(args.save, 'rb') as f:
    model_saved = torch.load(f)

model = model_saved['model']
encoder = model_saved['encoder']
sampled_softmax = model_saved['decoder']

# Run on test data.
test_loss = evaluate(test_data)
print('=' * 89)
print('| End of training | test loss {:5.2f} | test perplexity {:8.2f}'.format(
    test_loss, math.exp(test_loss)))
print('=' * 89)