We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
the loss stop reducing after pruning (only one step). Is it any wrong with my code?
imp = tp.importance.MagnitudeImportance(p=2) ignored_layers = [] iterative_steps = 100 example_inputs = torch.randn(1, 1, FACE_SHAPE[0], FACE_SHAPE[1]) for m in net.modules(): if isinstance(m, torch.nn.Linear) and m.out_features == 5: print(m) ignored_layers.append(m) # DO NOT prune the final classifier! pruner = tp.pruner.MagnitudePruner(model=net, example_inputs=example_inputs, pruning_ratio=0.01, importance=imp, iterative_steps=iterative_steps, ignored_layers=ignored_layers, ) # ... for epoch in range(max_epochs): self.train(model=net, optimizer=optimizer, data_loader=train_loader, device=device, epoch=epoch, pruner=pruner) #... valid_loss, f1 = self.valid(model=net, valid_loader=valid_loader, device=device, epoch=epoch) if f1 > 0.9 and steps > 0: pruner.step() steps -= 1
def train(self, model, optimizer, data_loader, device, epoch, pruner): """ :param pruner :param device: :param data_loader: :param optimizer: :param model: :type epoch: int """ criterion = CustomLoss() model.train() sum_loss = 0 for batch_idx, (X, y) in enumerate(data_loader, 0): X = X.to(device) y = y.to(device) output = model(X) optimizer.zero_grad() # print(output) loss = criterion(output, y) pruner.regularize(model, loss) loss.backward() optimizer.step() sum_loss += loss.item() self.global_step += 1 self.trainWriter.add_scalar("loss", loss.item(), self.global_step) self.trainWriter.add_scalar("train loss", sum_loss / len(data_loader), epoch + 1) lr = optimizer.param_groups[0]['lr'] self.trainWriter.add_scalar("lr", lr, epoch + 1)
The text was updated successfully, but these errors were encountered:
No branches or pull requests
the loss stop reducing after pruning (only one step).
Is it any wrong with my code?
The text was updated successfully, but these errors were encountered: