Plotting loss curve from training loop

AI Discussions
1

hello, I am looking for a way to plot the loss curve in my training loop pytorch. Any ideas on how do I do it? I cant find the right one for me

for epoch in range(num_epochs):
train_loss = 0.0
valid_loss = 0.0

model.train() # prep model for training
for it, training_data in enumerate(training_loader):
    optimizer.zero_grad()
    training_label = training_data[0]
    output = model(training_input)
    loss = criterion(output, training_label)
    loss.backward()

    optimizer.step()
    # update running training loss
    if it % 25 == 0:
        print(f"loss at step {it}: {loss}")
    if it == 50000:
        break
1 Like
2

Have you seen this?

1 Like
3

I did. However, it is not a loss curve graph. I am looking for something like a matplotlib.

I am currently considering printing out the loss and graph it in excel

4

Once you get the loss, package like matplotlib can be used to plot the loss.
Please explain in detail on what the issue is.

5

I wasnt able to implement the matplotlib and I cant find anythin on the internet that works for mine

I think its because I didnt have grad=true?

6

Please click my name and message your file as an attachment.

7

Hey, I am very sorry. I just found out how to do it just now and felt really. Its only declaring loss array and .append function to store the loss.