Skip to content

CNN weights example #730

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

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

Jump to bottom
Draft
clewis7 wants to merge 2 commits into
base: main
Choose a base branch
from
Draft

CNN weights example #730

clewis7 wants to merge 2 commits into from

Conversation

@clewis7
Copy link
Member

@clewis7 clewis7 commented Feb 17, 2025 •
edited
Loading

WIP!

Started playing around with a simple CNN in pytorch to visualize the weights as the model is training using the MNIST dataset

image

The weights are definitely being updated every epoch, but the plot will not update until the last training epoch is done...not sure why

Another thing of note, pytorch tensors cannot directly be used with fpl right now because

File [~/repos/fastplotlib/fastplotlib/graphics/_features/_image.py:101](http://localhost:8888/lab/tree/examples/fastplotlib/graphics/_features/_image.py#line=100), in TextureArray._fix_data(self, data)
     95     raise ValueError(
     96         "image data must be 2D with or without an RGB(A) dimension, i.e. "
     97         "it must be of shape [rows, cols], [rows, cols, 3] or [rows, cols, 4]"
     98     )
    100 # let's just cast to float32 always
--> 101 return data.astype(np.float32)

AttributeError: 'Tensor' object has no attribute 'astype'

@kushalkolar
Copy link
Member

kushalkolar commented Feb 17, 2025

Are you trying to update a graphic in a for loop? You can't do this right now, see the discussions post. For now we can use zmq or see rendercanvas and try to expose the ondemand rendering mode in fpl

Torch tensors can't be used directly but if you do np.asarray it will not make a copy

@clewis7
Copy link
Member Author

clewis7 commented Feb 17, 2025 •
edited
Loading

Torch tensors can't be used directly but if you do np.asarray it will not make a copy

Even when I do np.asarray I still get issues that the arrays are on different devices

TypeError: can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first.

Are you trying to update a graphic in a for loop? You can't do this right now, see the discussions post. For now we can use zmq or see rendercanvas and try to expose the ondemand rendering mode in fpl

I was trying to use for loop, I will try using zmq for now

@kushalkolar
Copy link
Member

kushalkolar commented Feb 17, 2025

Yea you need to send to cpu

@clewis7
Copy link
Member Author

clewis7 commented Feb 17, 2025 •
edited
Loading

Okay, still a bit messy and inelegant...but progress

Right now I have the weights updating after every training epoch

net_weights.mp4

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@kushalkolar kushalkolar Awaiting requested review from kushalkolar kushalkolar will be requested when the pull request is marked ready for review kushalkolar is a code owner

At least 1 approving review is required to merge this pull request.

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@clewis7 @kushalkolar