Examples: query, "exact match", wildcard*, wild?ard, wild*rd
Fuzzy search: cake~ (finds cakes, bake)
Term boost: "red velvet"^4, chocolate^2
Field grouping: tags:(+work -"fun-stuff")
Escaping: Escape characters +-&|!(){}[]^"~*?:\ with \, e.g. \+
Range search: properties.timestamp:[1587729413488 TO *] (inclusive), properties.title:{A TO Z}(excluding A and Z)
Combinations: chocolate AND vanilla, chocolate OR vanilla, (chocolate OR vanilla) NOT "vanilla pudding"
Field search: properties.title:"The Title" AND text
Answered
Hi, I Just Started Working With Clearml And I Have A Question About One Of My Plots. Fig 1 Is What It Looks Like In My Pycharm Output. Fig 2 Is What Is Logged In The Clearml Server. Why Does The Figure Change So Drastically? And How Can I Solve It? Sideno

Hi,
I just started working with ClearML and I have a question about one of my plots.
Fig 1 is what it looks like in my pycharm output.
Fig 2 is what is logged in the clearML server.
Why does the figure change so drastically? And how can I solve it?
Sidenote: there seems to be a bug in the plot viewer, as the axis are a bit chaotic..
Thanks in advance!

  
  
Posted 2 years ago
Votes Newest

Answers 13


Hi Natan,
Thanks for your reply.

Fig1 = the left one, with the subplots
Fig2 = the right one, with the 'child' labels (which I did not create)

This is the plot that is automatically registered in the details of the experiment.

  
  
Posted 2 years ago

TartSeagull57 , what framework are you on? What version of ClearML are you using?

  
  
Posted 2 years ago

what framework are you on? What version of ClearML are you using?

I'm using PyTorch and my clearml version is 1.3.2.

What are you referring yo specifically? The data plots seem to be identical.

The top subplot disappears entirely. Also the 'child' labels are added, it seems like it tries to combine the two plots?

Do you mean the x/y intersection?

The x labels are not at the x-axis but in the middle of the plot. Also the y-axis seems to be duplicated? There are different labels written over each other.

  
  
Posted 2 years ago

I circumvented the problem by saving the variables as scalars, those automatic plots work well.
Thanks for your help!

  
  
Posted 2 years ago

TartSeagull57 , you said the problem was with automatic reporting. Can you give an example of how you solved the issue for yourself?

  
  
Posted 2 years ago

CostlyOstrich36 The problem was that the plot I created myself (with the loss and accuracy) was not stored properly in the results/plot tab. I removed the plotting from my script and instead I save the loss and accuracy using Logger.current_logger().report_scalar(). With the loss and accuracy properly saved as scalars, they are plotted nicely in the results/scalar tab.

  
  
Posted 2 years ago

The problem was that the plot I created myself

How was the plot created? Can you give me a small snippet to try and play around with?

  
  
Posted 2 years ago

TartSeagull57 , can you please open a github issue for tracking? 🙂

  
  
Posted 2 years ago

Sure!

  
  
Posted 2 years ago

TartSeagull57 , I couldn't make the sample you gave me work 😞

Can you please provide a self contained example that would reproduce the issue?

  
  
Posted 2 years ago

Why does the figure change so drastically? And how can I solve it?

What are you referring yo specifically? The data plots seem to be identical.

Sidenote: there seems to be a bug in the plot viewer, as the axis are a bit chaotic..

Do you mean the x/y intersection?

  
  
Posted 2 years ago

Hi TartSeagull57 ,

Which one is fig 1 and which one is fig 2?
How are you logging them?

  
  
Posted 2 years ago

fig, axs = plt.subplots(2, 1) axs[1].plot(range(epoch), accuracy_train, range(epoch), accuracy_valid) axs[1].legend(['Training accuracy', 'Validation accuracy']) axs[0].plot(range(epoch), loss_train, range(epoch), loss_valid) axs[0].legend(['Training loss', 'Validation loss']) plt.show()This is the code I used on the output of my model (epoch, accuracy_train, accuracy_valid, loss_train, loss_valid)

  
  
Posted 2 years ago
716 Views
13 Answers
2 years ago
one year ago
Tags