Add pytorch MNIST Workflow tutorial with FedProx #1158
Conversation
teoparvanov
changed the title
|
@ynonflumintel I think this is not the correct file in this commit, I dont see workflow api getting used in this file. |
teoparvanov
requested review from
teoparvanov,
kta-intel and
MasterSkepticista
openfl-tutorials/Federated_FedProx_PyTorch_MNIST_Workflow_Tutorial.ipynb
Outdated
Show resolved
Hide resolved
ynonflumintel
force-pushed
the
yf/add-pytorch-workflow-example
branch
from
3624c38
to
c73dd09
Compare
I pushed the wrong version of the file before, please take a look now |
There was a problem hiding this comment.
Looks great, thanks @ynonflumintel ! I just have a couple of small suggestions:
openfl-tutorials/Federated_FedProx_PyTorch_MNIST_Workflow_Tutorial.ipynb
Outdated
Show resolved
Hide resolved
openfl-tutorials/Federated_FedProx_PyTorch_MNIST_Workflow_Tutorial.ipynb
Outdated
Show resolved
Hide resolved
openfl-tutorials/Federated_FedProx_PyTorch_MNIST_Workflow_Tutorial.ipynb
Outdated
Show resolved
Hide resolved
openfl-tutorials/Federated_FedProx_PyTorch_MNIST_Workflow_Tutorial.ipynb
Outdated
Show resolved
Hide resolved
There was a problem hiding this comment.
General comment on the entire notebook: Include output of all the cells. New users should have visibility into the cells and evaluate for themselves if the notebook fulfils their goal of visit. Not showing output, can put-off newcomers because they end goal is not "visible".
openfl-tutorials/Federated_FedProx_PyTorch_MNIST_Workflow_Tutorial.ipynb
Outdated
Show resolved
Hide resolved
openfl-tutorials/Federated_FedProx_PyTorch_MNIST_Workflow_Tutorial.ipynb
Outdated
Show resolved
Hide resolved
| " loss.backward()\n", | ||
| " self.optimizer.step()\n", | ||
| "\n", | ||
| " if (len(data) * batch_idx) / len(self.train_loader.dataset) >= log_threshold:\n", |
There was a problem hiding this comment.
Instead of this complicated logic of calculating "quarter" of samples used, it is OK to print every 10 or 100 steps. We need to be cautious of all the cognitive overhead we put on the user.
There was a problem hiding this comment.
I ran experiments where there were a couple of thousand steps (small chunks), it's too spammy IMO
ynonflumintel
force-pushed
the
yf/add-pytorch-workflow-example
branch
from
c73dd09
to
edd7e00
Compare
Thought it would be too verbose and increase the file size significantly, but I added them as per your request |
ynonflumintel
force-pushed
the
yf/add-pytorch-workflow-example
branch
3 times, most recently
from
b8fbfbc
to
30ecf9e
Compare
Signed-off-by: Ynon Flum <[email protected]>
ynonflumintel
force-pushed
the
yf/add-pytorch-workflow-example
branch
from
30ecf9e
to
448240c
Compare
This branch introduces a tutorial for setting up a neural network training federation using pytorch, the MNIST dataset using the openfl workflow API.