Restructuring 2 - Add Utility functions and Run all Tests #12
Conversation
kartikdutt18
force-pushed
the
Restructuring-2-a-UtilityFunctions
branch
4 times, most recently
from
7a3666b to
07d732a
Compare
|
This works on windows too now, Cleaning commits and outputs. |
kartikdutt18
force-pushed
the
Restructuring-2-a-UtilityFunctions
branch
from
07d732a to
3d74ee1
Compare
Sure, we can do that, is there anything else you like to add here, this looks pretty good to me. |
I am done with this PR from my side aside from placing mnist_train/test in mnist.tar.gz. If this makes sense I'll ask on the IRC once I incorporate all suggestions. |
|
Makes sense to me. |
|
Looks good to me as well, please let us know if this is ready to go :) |
Trained a lenet1 model Add All Weights
Completed everything Fix build, Style fixes next Use force local for windows while unzipping tar files Use force local for windows while unzipping tar files
kartikdutt18
force-pushed
the
Restructuring-2-a-UtilityFunctions
branch
from
3d74ee1 to
9d90a21
Compare
Added Utility Function Style Fix Fix Typo causing build error in windows Fix Typo causing build error in windows Fix const issue for windows Extract in data folder Reposition force local This should work Print Path in windows for debugging Print Path in windows for debugging Print Path in windows for debugging Stip components of tar Stip components of tar strip componenets Relative Path Relative Path Relative Path Add List Dir utility function Print Print again Print again Squash this Update utils.hpp Squash this Clean Up Added more tests Style Fix Style Fix
kartikdutt18
force-pushed
the
Restructuring-2-a-UtilityFunctions
branch
from
585905f to
6ae432e
Compare
|
This is ready from my side. Just tested it everything works fine. I think the builds will result in the same output. |
|
Sorry pushed in the wrong branch, reverting now. Fixed now. |
kartikdutt18
force-pushed
the
Restructuring-2-a-UtilityFunctions
branch
from
e51aaa9 to
6ae432e
Compare
tests/dataloader_tests.cpp
Outdated
|
|
||
|
|
There was a problem hiding this comment.
Can you use 1 newline and add a comment about what we're checking in this block?
And, perhaps, I missed something... can you let me know why the data were distributed like below?
Train : 8,400
Valid: 33,600
Test: 28,000
There was a problem hiding this comment.
This is the same split as mlpack::data::split. Here 0.8 means 80% testing and 20 % training data. mlpack takes in test ratio. I though it would be best not to invert. Let me know if I should do that or keep it same as mlpack. Here is the link for data file.
There was a problem hiding this comment.
Ok, then, it would be better to change the parameter name ratio to testRatio for clarification in DataLoader class :)
There was a problem hiding this comment.
Sure, I'll make the change.
There was a problem hiding this comment.
Also, kindly update the parameter description as well in here
There was a problem hiding this comment.
Ah, ok. I thought it's from the same file :)
There was a problem hiding this comment.
Should I use validRatio instead I think that makes even more sense.
There was a problem hiding this comment.
Nice. that makes sense to me as well.
There was a problem hiding this comment.
Right, making the changes.
kartikdutt18
force-pushed
the
Restructuring-2-a-UtilityFunctions
branch
from
66cd0d7 to
ed7a643
Compare
|
Great. Thanks a lot for the reviews. |
![mlpack-bot[bot]](https://avatars.githubusercontent.com/in/23934?s=60&v=4){kind=small}
|
Thanks! @kartikdutt18 👍 |
|
Thanks @KimSangYeon-DGU, @zoq for the reviews and all the help. |
This PR is transferred from #8, Since FFN weights some time cause invalid memory access which seems related to mlpack/mlpack#2314. So I am adding utility function to unzip files here and I'll continue working on image dataloader and then we can get back to the LeNet model.
I hope this makes sense.
This PR provides the following :
(Probably we should get mnist_test/train in mnist.tar.gz since mnist test and train are fixed).
Looking forward to your suggestions.
Regards.
CC: @KimSangYeon-DGU