Smt #45
Open
Smt #45
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
* adds fastMNIST dataset and parameters * fixes conv lca params to match between pytorch & tf1x - no longer getting NAN * fixes relative imports for utils/loaders.py - need to propagate to other files * fixes minor bugs in notebooks and run_utils * improves some error messaging
removes incomplete aversarial_analysis script all pytorch tests pass
adds convolutional MLP model with max pooling reorders all expected datashapes to have channels first moves typical log outputs to base model adds optimizer to checkpoint writing adds ability to load checkpoints from a log file adds ability to boot from checkpoint at the start of training (untested) adds ability to ignore gradients or include them when training ensemble models (untested) datasets now include a 'num_pixels' parameter in their output
adds utility to laod parameters from a log file adds preprocessing capability to FastMNIST minor typo fixes
…o smt Conflicts: modules/mlp_module.py
logging now includes computer environment and model architecture details lca now outputs fraction nonzero for channel location as well as convolutional spatial map fixes a checkpoint loading bug for ensemble models fixes a bug with ensemble modules that caused submodules of the same type to clobber removes flatten_feature_map function from utils.data_processing in favor of one-line option reorganizes mlp and lca forward function calls for cleaner integration into ensembles renames lca num_latent params to layer_channels to match mlp specification adds pooling params to the test suite changed some logging apis to be more intuitive / general
standardization can now use the dataset mean & std dataset_utils now outputs original dataset mean & std
removes unnecessary thresholding operators from modules/activations cleans up inhibitory connectivity function & adds convolutional version switched to torch.mm instead of torch.matmul when matrices are 2D layer_channels behaves like mlp, and now must include input channels minor comment addition/removal in pooling_module returned cifar preprocessing to be samplewise standardization fixed bug in tests with new dataset outputs updates ensemble lca test with comments & fixes ensemble state dict loading bug
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
No description provided.