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.

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

Pointwise features generation #2

Open
wants to merge 18 commits into
base: main
Choose a base branch
from

Conversation

Sriharsha-hatwar
Copy link
Collaborator

This request contains point-wise feature code generation as a script.

Sriharsha-hatwar and others added 6 commits March 11, 2023 21:10
dhdhagar#39)

* Add --debug and --track_errors to log cvxpylayer errors
* Fix save_to_wandb call for hyperparameters
* Fix tensor serializable error
* Fix for --eval_only_split flow
* Add `sdp_scale` hyperparameter to scale the weight matrix to the SDP layer by the maximum element
* Add `gradient_accumulation` hyperparameter
* Change run defaults and sweep configs
* Add sweep prefix option to run_sweep
* Increase sweep agent memory
* Clamp cvxpy output to [0,1]
* Address meshgrid warning
* Add `weighted_loss` to e2e sweep config
* Modify run_sweep.sh to take in seed start and end values
* Add `use_sdp` hyperparam to control whether to use the SDP during training and inference or directly use the MLP output with HAC-cut
* Log errors before crashing; make error tracking the default behavior
* Exception handling improvements
* Make tqdm verbose even in silent mode
* Save best dev model before testing
* Add *-nosdp sweep configurations
* Add `e2e_loss` hyperparam to control whether to use Frobenius or BCE loss
* Change default subsampling to 80 (train) and 100 (dev)
* Add --local to run with wandb disabled, change default weighted_loss to false to stay consistent with icml23 submission
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants