Skip to content
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: Add CFAIR Experiemts #62

Open
wants to merge 51 commits into
base: main
Choose a base branch
from
Open

DRAFT: Add CFAIR Experiemts #62

wants to merge 51 commits into from

Conversation

fariedabuzaid
Copy link
Collaborator

@fariedabuzaid fariedabuzaid commented Sep 29, 2024 •
edited
Loading

  • Define data classes
  • Define experiment
  • Test

fariedabuzaid and others added 30 commits September 29, 2024 19:21
@fariedabuzaid
Add CFAIR split
fdd7098
@fariedabuzaid
Define CFAIR experiment
7ed6c8b
@fariedabuzaid
Add L^{\inf}-radial MNIST experiment configuration
e575efd
@YalcinerMustafa
Add import that is required when running the MNIST training file as is.
34bd806
@YalcinerMustafa
update this keyword to match with the latest code-base.
169f9ae
@YalcinerMustafa
Fix configuration file and add gpu version to train on dgx.
4b4a9ba
@YalcinerMustafa
tryout a fix for loading tensors on the wrong device.
635de70
@YalcinerMustafa
Revert "tryout a fix for loading tensors on the wrong device."
ad655fd 
This reverts commit 635de70.
@YalcinerMustafa
dont use lu layers, there seems to be a bug there.
f0a487b
@YalcinerMustafa
update the config file with l_inf norm.
40ee80c
@YalcinerMustafa
update config file to include lu layers.
661b993
@YalcinerMustafa
revert the use of lu layers.
975bebb
@YalcinerMustafa
Add configuration entry for radial-exponential.
6a8b63d
@YalcinerMustafa
add/fix config files.
478f6e0
@YalcinerMustafa
Add comment.
e32ec29
@YalcinerMustafa
Initialize the dequantizer superclass to smooth the cifar dataset, as…
bef50ba 
… well. Also, add a filter method for training on only one class of cifar. adjust the config file for cpu. remain backward compatible to not break the mnist training. contains the following bug: the training process stops due to mismatch of tensor dimensions, also often, the report folders by ray explode in size probably due to a bug.
@YalcinerMustafa
Add configuration for training with lognormal distribution with vario…
f0dbd1c 
…us parameters.
@YalcinerMustafa
Add cofig file for ablation with lognormal.
1f70b8b
@YalcinerMustafa
update config file.
082a0cc
@YalcinerMustafa
Make the models significantly smaller.
6232a9b
@YalcinerMustafa
Change experiment name to not confuse folder names.
862a2e5
@YalcinerMustafa
Add config entry for learning slightly larger models on mnist.
3fb1a27
@YalcinerMustafa
add new experiment config.
d4f7133
@YalcinerMustafa
Add method for showing samples from the dataset. Rescaling to 14*14 s…
ed440d2 
…eems meaningful.
@YalcinerMustafa
allow for rescaling mnist to 14*14 (instead of only 10*10).
3f857a9
@YalcinerMustafa
Update config file for training on mnist 14*14.
3a8fdb2
@YalcinerMustafa
change experiment name and minor param change.
65c9d7b
@YalcinerMustafa
improve config file for training mnist on 14*14.
9457ff9
@YalcinerMustafa
update config file to try divers hyperparameters- training keeps term…
5a29004 
…inating prematurely.
@YalcinerMustafa
Fix config file.
75efe8d
YalcinerMustafa and others added 21 commits October 21, 2024 11:48
@YalcinerMustafa
Increase patience.
00c41cf
@YalcinerMustafa
Narrow down the parameters to the ones that worked best.
de211d0
@YalcinerMustafa
Add config file for training on all mnist digts rescaled to 14 times …
6adf28d 
…14 pixels.
@YalcinerMustafa
Slightly reduce the number of hyperopt samples to run on all digits.
6558f7c
@YalcinerMustafa
Add a fixed and simplified configuration entry for learning multiple …
0809203 
…digits.
@YalcinerMustafa
Fix config file to actually train differente digits (dont use the ove…
29f901d 
…rride feature).
@YalcinerMustafa
Add config entry to test multiple location params.
d29d9f0
@YalcinerMustafa
Add new training config for all digits that does not use the override…
419fda1 
… keyword.
@YalcinerMustafa
rename config file.
c02f54c
@YalcinerMustafa
add config entry for training with lu layers.
af228ac
@YalcinerMustafa
Search for best loc param under use of LU layers.
610af31
@YalcinerMustafa
Turn on lu layers.
8e99bfb
@YalcinerMustafa
Add config file for testing smaller networks with lu layer.
6460ddc
@YalcinerMustafa
Reduce numer of coupling nn layers to avoid model getting too big.
52b6e75
@YalcinerMustafa
Add training for larger models.
0dca0bc
@YalcinerMustafa
Increase patience and allow for even lower learning rates to avoid ea…
136eeac 
…rly termination in local minima.
@YalcinerMustafa
Change loc param and adjust lr.
ccb6ca1
update dependencies (ray version no longer available)
9d4829b
Minor adoptions for CFAIR
4853111
Correct dimension
a07bd78
update config
4d84c7f
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers
No reviews
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
2 participants
@fariedabuzaid @YalcinerMustafa