principled-dfa-training
principled-dfa-training copied to clipboard
Code for our paper on best practices to train neural networks with direct feedback alignment (DFA).
Hi @iacolippo I was trying to run the `best_practises.py` but it ran into an error at https://github.com/lightonai/principled-dfa-training/blob/5a386652b2f425a769d4092de65117e01a1a9413/best_practices.py#L274 The base_name `3_test_DFA_FBnorm_initSTD_FC_ND_BN_Tanh()_CIFAR-10_e50_bs128_from_scratch_s` is not matched with any of the files' name that...
The open-source community is much more reluctant to adopt new code if there is no license associated with it, since technically we have no right to use the code without...