Pinned Repositories
git-gpg
Use git on an untrusted server via a GPG-encrypted repository.
heinsen_attention
Reference implementation of "Softmax Attention with Constant Cost per Token" (Heinsen, 2024)
heinsen_position_embeddings
Reference implementation of "Encoding Position by Decaying and Updating Different Exponentiated States Differently," as proposed by Franz A. Heinsen in March 2024
heinsen_routing
Reference implementation of "An Algorithm for Routing Vectors in Sequences" (Heinsen, 2022) and "An Algorithm for Routing Capsules in All Domains" (Heinsen, 2019), for composing deep neural networks.
heinsen_routing_2022_paper
Instructions for replicating the results in "An Algorithm for Routing Vectors in Sequences" (Heinsen, 2022).
heinsen_sequence
Code implementing "Efficient Parallelization of a Ubiquitious Sequential Computation" (Heinsen, 2023)
heinsen_tree
Reference implementation of "Tree Methods for Hierarchical Classification in Parallel" (Heinsen, 2022) in PyTorch.
torch_lsuv_init
Layer Sequential Unit Variance (LSUV) Initialization for PyTorch
torch_parallel_scan
A simple implementation of parallel scan over sequences of tensors for PyTorch (Franz A. Heinsen, 2024)
torch_train_test_loop
Composable training/testing of PyTorch models with minimal overhead
GlassRoom's Repositories
glassroom/heinsen_routing
Reference implementation of "An Algorithm for Routing Vectors in Sequences" (Heinsen, 2022) and "An Algorithm for Routing Capsules in All Domains" (Heinsen, 2019), for composing deep neural networks.
glassroom/heinsen_sequence
Code implementing "Efficient Parallelization of a Ubiquitious Sequential Computation" (Heinsen, 2023)
glassroom/git-gpg
Use git on an untrusted server via a GPG-encrypted repository.
glassroom/heinsen_attention
Reference implementation of "Softmax Attention with Constant Cost per Token" (Heinsen, 2024)
glassroom/heinsen_tree
Reference implementation of "Tree Methods for Hierarchical Classification in Parallel" (Heinsen, 2022) in PyTorch.
glassroom/heinsen_position_embeddings
Reference implementation of "Encoding Position by Decaying and Updating Different Exponentiated States Differently," as proposed by Franz A. Heinsen in March 2024
glassroom/torch_lsuv_init
Layer Sequential Unit Variance (LSUV) Initialization for PyTorch
glassroom/heinsen_routing_2022_paper
Instructions for replicating the results in "An Algorithm for Routing Vectors in Sequences" (Heinsen, 2022).
glassroom/torch_parallel_scan
A simple implementation of parallel scan over sequences of tensors for PyTorch (Franz A. Heinsen, 2024)
glassroom/torch_train_test_loop
Composable training/testing of PyTorch models with minimal overhead
glassroom/heinsen_routing_2019_paper
Instructions for replicating the results in "An Algorithm for Routing Capsules in All Domains" (Heinsen, 2019).