Estoy entrenando una auto-encoder
red con Adam
optimizador (con amsgrad=True
) y MSE loss
para la tarea de separación de fuente de audio de un solo canal. Cada vez que disminuyo la tasa de aprendizaje por un factor, la pérdida de la red salta abruptamente y luego disminuye hasta la próxima disminución en la tasa de aprendizaje.
Estoy usando Pytorch para la implementación y capacitación de redes.
Following are my experimental setups:
Setup-1: NO learning rate decay, and
Using the same Adam optimizer for all epochs
Setup-2: NO learning rate decay, and
Creating a new Adam optimizer with same initial values every epoch
Setup-3: 0.25 decay in learning rate every 25 epochs, and
Creating a new Adam optimizer every epoch
Setup-4: 0.25 decay in learning rate every 25 epochs, and
NOT creating a new Adam optimizer every time rather
using PyTorch's "multiStepLR" and "ExponentialLR" decay scheduler
every 25 epochs
Estoy obteniendo resultados muy sorprendentes para las configuraciones # 2, # 3, # 4 y no puedo razonar ninguna explicación al respecto. Los siguientes son mis resultados:
Setup-1 Results:
Here I'm NOT decaying the learning rate and
I'm using the same Adam optimizer. So my results are as expected.
My loss decreases with more epochs.
Below is the loss plot this setup.
Parcela-1:
optimizer = torch.optim.Adam(lr=m_lr,amsgrad=True, ...........)
for epoch in range(num_epochs):
running_loss = 0.0
for i in range(num_train):
train_input_tensor = ..........
train_label_tensor = ..........
optimizer.zero_grad()
pred_label_tensor = model(train_input_tensor)
loss = criterion(pred_label_tensor, train_label_tensor)
loss.backward()
optimizer.step()
running_loss += loss.item()
loss_history[m_lr].append(running_loss/num_train)
Setup-2 Results:
Here I'm NOT decaying the learning rate but every epoch I'm creating a new
Adam optimizer with the same initial parameters.
Here also results show similar behavior as Setup-1.
Because at every epoch a new Adam optimizer is created, so the calculated gradients
for each parameter should be lost, but it seems that this doesnot affect the
network learning. Can anyone please help on this?
Parcela-2:
for epoch in range(num_epochs):
optimizer = torch.optim.Adam(lr=m_lr,amsgrad=True, ...........)
running_loss = 0.0
for i in range(num_train):
train_input_tensor = ..........
train_label_tensor = ..........
optimizer.zero_grad()
pred_label_tensor = model(train_input_tensor)
loss = criterion(pred_label_tensor, train_label_tensor)
loss.backward()
optimizer.step()
running_loss += loss.item()
loss_history[m_lr].append(running_loss/num_train)
Setup-3 Results:
As can be seen from the results in below plot,
my loss jumps every time I decay the learning rate. This is a weird behavior.
If it was happening due to the fact that I'm creating a new Adam
optimizer every epoch then, it should have happened in Setup #1, #2 as well.
And if it is happening due to the creation of a new Adam optimizer with a new
learning rate (alpha) every 25 epochs, then the results of Setup #4 below also
denies such correlation.
Parcela-3:
decay_rate = 0.25
for epoch in range(num_epochs):
optimizer = torch.optim.Adam(lr=m_lr,amsgrad=True, ...........)
if epoch % 25 == 0 and epoch != 0:
lr *= decay_rate # decay the learning rate
running_loss = 0.0
for i in range(num_train):
train_input_tensor = ..........
train_label_tensor = ..........
optimizer.zero_grad()
pred_label_tensor = model(train_input_tensor)
loss = criterion(pred_label_tensor, train_label_tensor)
loss.backward()
optimizer.step()
running_loss += loss.item()
loss_history[m_lr].append(running_loss/num_train)
Setup-4 Results:
In this setup, I'm using Pytorch's learning-rate-decay scheduler (multiStepLR)
which decays the learning rate every 25 epochs by 0.25.
Here also, the loss jumps everytime the learning rate is decayed.
Como lo sugirió @Dennis en los comentarios a continuación, probé con ambas ReLU
y sin 1e-02 leakyReLU
linealidades. Pero, los resultados parecen comportarse de manera similar y la pérdida primero disminuye, luego aumenta y luego se satura a un valor más alto de lo que alcanzaría sin disminuir la tasa de aprendizaje.
La gráfica 4 muestra los resultados.
Parcela-4:
scheduler = torch.optim.lr_scheduler.MultiStepLR(optimizer=optimizer, milestones=[25,50,75], gamma=0.25)
scheduler = torch.optim.lr_scheduler.ExponentialLR(optimizer=optimizer, gamma=0.95)
scheduler = ......... # defined above
optimizer = torch.optim.Adam(lr=m_lr,amsgrad=True, ...........)
for epoch in range(num_epochs):
scheduler.step()
running_loss = 0.0
for i in range(num_train):
train_input_tensor = ..........
train_label_tensor = ..........
optimizer.zero_grad()
pred_label_tensor = model(train_input_tensor)
loss = criterion(pred_label_tensor, train_label_tensor)
loss.backward()
optimizer.step()
running_loss += loss.item()
loss_history[m_lr].append(running_loss/num_train)
EDICIONES:
- Como se sugiere en los comentarios y la respuesta a continuación, he realizado cambios en mi código y he entrenado el modelo. He agregado el código y las parcelas para lo mismo.
- Probé con varios
lr_scheduler
enPyTorch (multiStepLR, ExponentialLR)
y parcelas de la misma se enumeran enSetup-4
lo sugerido por @Dennis en los comentarios a continuación. - Intentando con leakyReLU como lo sugiere @Dennis en los comentarios.
Alguna ayuda. Gracias