Large, exploding loss in Pytorch transformer model - machine-learning

Question

I am trying to solve a sequence to sequence problem with a transformer model. The data is derived from a set of crossword puzzles.
The positional encoding and transformer classes are as follows:
class PositionalEncoding(nn.Module):
def __init__(self, d_model: int, dropout: float = 0.1, max_len: int = 3000):
super().__init__()
self.dropout = nn.Dropout(p=dropout)
position = torch.arange(max_len).unsqueeze(1)
div_term = torch.exp(torch.arange(0, d_model, 2) * (-math.log(10000.0) / d_model))
pe = torch.zeros(1, max_len, d_model)
pe[0, :, 0::2] = torch.sin(position * div_term)
pe[0, :, 1::2] = torch.cos(position * div_term)
self.register_buffer('pe', pe)
def debug(self, x):
return x.shape, x.size()
def forward(self, x: Tensor) -> Tensor:
x = x + self.pe[:, :x.size(1), :]
return self.dropout(x)
class Transformer(nn.Module):
def __init__(
self,
num_tokens,
dim_model,
num_heads,
num_encoder_layers,
num_decoder_layers,
batch_first,
dropout_p,
):
super().__init__()
self.model_type = "Transformer"
self.dim_model = dim_model
self.positional_encoder = PositionalEncoding(
d_model=dim_model, dropout=dropout_p, max_len=3000
)
self.embedding = nn.Embedding.from_pretrained(vec_weights, freeze=False)#nn.Embedding(num_tokens, dim_model)
self.transformer = nn.Transformer(
d_model=dim_model,
nhead=num_heads,
num_encoder_layers=num_encoder_layers,
num_decoder_layers=num_decoder_layers,
dropout=dropout_p,
batch_first = batch_first
)
self.out = nn.Linear(dim_model, num_tokens)
def forward(self, src, tgt, tgt_mask=None, src_pad_mask=None, tgt_pad_mask=None):
src = self.embedding(src)*math.sqrt(self.dim_model)
tgt = self.embedding(tgt)*math.sqrt(self.dim_model)
src = self.positional_encoder(src)
tgt = self.positional_encoder(tgt)
transformer_out = self.transformer(src, tgt, tgt_mask=tgt_mask, src_key_padding_mask=src_pad_mask, tgt_key_padding_mask=tgt_pad_mask)
out = self.out(transformer_out)
return out
def get_tgt_mask(self, size) -> torch.tensor:
mask = torch.tril(torch.ones(size, size) == 1)
mask = mask.float()
mask = mask.masked_fill(mask == 0, float('-inf'))
mask = mask.masked_fill(mask == 1, float(0.0))
return mask
def create_pad_mask(self, matrix: torch.tensor, pad_token: int) -> torch.tensor:
return (matrix == pad_token)
The input tensors are a source tensor of size N by S, where N is the batch size and S is the source sequence length, and a target tensor of size N by T, where T is the target sequence length. S is about 10 and T is about 5, while the total number of items is about 160,000-200,000, divided into batch sizes of 512. They are torch.IntTensors, with elements in the range from 0 to V, where V is the vocabulary length.
The first layer is an embedding layer that takes the input from N by S to N by S by E, where E is the embedding dimension (300), or to N by T by E in the case of the target. The second layer adds position encoding without changing the shape. Then both tensors are passed through the transformer layer, which outputs an N by T by E tensor. Finally, we pass this output through a linear layer, which produces an N by T by V output, where V is the size of the vocabulary used in the problem. Here V is about 56,697. The most frequent tokens (words) appear about 50-60 times in the target tensor.
The transformer class also contains the functions for implementing the masking matrices.
Then we create the model and run it (this process is wrapped in a function).
device = "cuda"
src_train, src_test = torch.utils.data.random_split(src_t, [int(0.9*len(src_t)), len(src_t)-int(0.9*len(src_t))])
src_train, src_test = src_train[:512], src_test[:512]
tgt_train, tgt_test = torch.utils.data.random_split(tgt_t, [int(0.9*len(tgt_t)), len(tgt_t)-int(0.9*len(tgt_t))])
tgt_train, tgt_test = tgt_train[:512], tgt_test[:512]
train_data, test_data = list(zip(src_train, tgt_train)), list(zip(src_test, tgt_test))
train, test = torch.utils.data.DataLoader(dataset=train_data), torch.utils.data.DataLoader(dataset=test_data)
model = Transformer(num_tokens=ntokens, dim_model=300, num_heads=2, num_encoder_layers=3, num_decoder_layers=3, batch_first = True, dropout_p=0.1).to(device)
loss_function = nn.CrossEntropyLoss()
optimizer = torch.optim.SGD(model.parameters(), lr=0.0000001)
n_epochs = 50
def train_model(model, optimizer, loss_function, n_epochs):
loss_value=0
for epoch in range(n_epochs):
print(f"Starting epoch {epoch}")
for batch, data in enumerate(train):
x, y = data
if batch%100 == 0:
print(f"Batch is {batch}")
batch += 1
optimizer.zero_grad()
x, y = torch.tensor(x).to(device), torch.tensor(y).to(device)
y_input, y_base = y[:, :-1], y[:, 1:]
y_input, y_base = y_input.to(device), y_base.to(device)
tgt_mask = model.get_tgt_mask(y_input.shape[1]).to(device)
pad_token = vocabulary_table[embeddings.key_to_index["/"]]
src_pad_mask = model.create_pad_mask(x, pad_token).to(device)
tgt_pad_mask = model.create_pad_mask(y_input, pad_token).to(device)
z = model(x, y_input, tgt_mask, src_pad_mask, tgt_pad_mask)
z = z.permute(0, 2, 1).to(device)
y_base = y_base.long().to(device)
loss = loss_function(z, y_base).to(device)
loss.backward()
nn.utils.clip_grad_norm_(model.parameters(), max_norm=2.0, norm_type=2)
optimizer.step()
loss_value += float(loss)
if batch%100 == 0:
print(f"For epoch {epoch}, batch {batch} the cross-entropy loss is {loss_value}")
#Free GPU memory.
del z
del x
del y
del y_input
del y_base
del loss
torch.cuda.empty_cache()
return model.parameters(), loss_value
Basically, we split the data into test and training sets and use an SGD optimizer and cross-entropy loss. We create a masking matrix for the padding for both the target and source tensors, and a masking matrix for unseen elements for the target tensor. We then do the usual gradient update steps. Right now, there is no validation loop, because I cannot even get the training loss to decrease.
The loss is very high, reaching more than 1000 after 100 batches. More concerningly, the loss also increases rapidly during training, rather than decreasing. In the code that I included, I tried clipping the gradients, lowering the learning rate, and using a much smaller sample to debug the code.
What could be causing this behavior?

Answer 1

You are only adding things to your loss, so naturally it can only increase.
loss_value += float(loss)
You're supposed to set it to zero after every epoch. Now you set it to zero once, in the beginning of the training process. There is a training loop template here if you're interested (https://pytorch.org/tutorials/beginner/blitz/cifar10_tutorial.html). This explains the increasing loss. To further troubleshoot (if needed) I'd throw in a validation loop.