optimizer.add_hook(chainer.optimizer.WeightDecay(1e-4))
scheduler = AdamScheduler(optimizer, 0.5, args.lr_step_list, 0)
train_iterator = iterators.MultithreadIterator(train_dataset,
args.batch_size,
n_threads=args.nb_jobs)
train_eval = Evaluator("train", args)
valid_iterator = iterators.MultithreadIterator(valid_dataset,
args.batch_size,
False,
False,
n_threads=args.nb_jobs)
valid_eval = Evaluator("valid", args)
logger.info("Training...")
train_eval.reset()
st = time.time()
# Training loop
for iter_cnt, batch in enumerate(train_iterator):
if iter_cnt == args.nb_iters:
break
chainer.config.train = True
chainer.config.enable_backprop = True
batch_array = [
convert.concat_examples([x[idx] for x in batch], args.gpu)
for idx in data_idxs
]
model.cleargrads()
loss, pred_y, _ = model(tuple(map(Variable, batch_array)))
loss.backward()
input("here")
train_iterator = iterators.MultithreadIterator(train_dataset,
args.batch_size,
n_threads=args.nb_jobs)
train_eval = Evaluator("train", args)
valid_iterator = iterators.MultithreadIterator(valid_dataset,
args.batch_size,
False,
False,
n_threads=args.nb_jobs)
valid_eval = Evaluator("valid", args)
logger.info("Training...")
train_eval.reset()
st = time.time()
total_loss = 0
# Training loop
for iter_cnt, batch in enumerate(train_iterator):
if iter_cnt == args.nb_iters:
break
batch_array = [
convert.concat_examples([x[idx] for x in batch], args.gpu)
for idx in data_idxs
]
batch_array = torch.FloatTensor([x.tolist() for x in batch_array])
batch_array = Variable(batch_array)
optimizer.zero_grad() # zero out gradients
loss, pred_y = net(batch_array)
