torch.cuda.manual_seed(const.SEED)
torch.backends.cudnn.deterministic = True
model = Transformer(
source_vocab_size=SOURCE_VOCAB_SIZE,
target_vocab_size=TARGET_VOCAB_SIZE,
source_padding_index=SRC_PAD_IDX,
target_padding_index=TRG_PAD_IDX,
embedding_size=const.EMBEDDING_SIZE,
number_of_layers=const.NUMBER_OF_LAYERS,
number_of_heads=const.NUMBER_OF_HEADS,
forward_expansion=const.FORWARD_EXPANSION,
device=device,
).to(device)
model.apply(model_utils.initialize_weights)
optimizer = torch.optim.Adam(model.parameters(), lr=const.LEARNING_RATE)
cross_entropy = nn.CrossEntropyLoss(ignore_index=TRG_PAD_IDX)
print(f'The model has {model_utils.count_parameters(model):,} trainable parameters')
trainer = Trainer(
const=const,
optimizer=optimizer,
criterion=cross_entropy,
device=device,
)
trainer.train(
model=model,
train_iterator=train_iterator,
model = Transformer(src_pad_idx=src_pad_idx,
trg_pad_idx=trg_pad_idx,
trg_sos_idx=trg_sos_idx,
d_model=d_model,
enc_voc_size=enc_voc_size,
dec_voc_size=dec_voc_size,
max_len=max_len,
ffn_hidden=ffn_hidden,
n_head=n_head,
n_layers=n_layers,
drop_prob=drop_prob,
device=device).to(device)
model.apply(initialize_weights)
optimizer = Adam(params=model.parameters(),
lr=init_lr,
weight_decay=weight_decay,
eps=adam_eps)
criterion = nn.CrossEntropyLoss(ignore_index=src_pad_idx)
def train(model, iterator, optimizer, criterion, clip):
model.train()
epoch_loss = 0
for i, batch in enumerate(iterator):
src = batch.src
trg = batch.trg
