automl / unlocking_state_trackingLinks
Expanding linear RNN state-transition matrix eigenvalues to include negatives improves state-tracking tasks and language modeling without added training or inference costs.
☆15Updated 2 months ago
Alternatives and similar repositories for unlocking_state_tracking
Users that are interested in unlocking_state_tracking are comparing it to the libraries listed below
Sorting:
- Experiments on the impact of depth in transformers and SSMs.☆30Updated 6 months ago
- Efficient PScan implementation in PyTorch☆16Updated last year
- Code for "Theoretical Foundations of Deep Selective State-Space Models" (NeurIPS 2024)☆12Updated 4 months ago
- Engineering the state of RNN language models (Mamba, RWKV, etc.)☆32Updated last year
- ☆46Updated last year
- Official repository of paper "RNNs Are Not Transformers (Yet): The Key Bottleneck on In-context Retrieval"☆27Updated last year
- Parallel Associative Scan for Language Models☆18Updated last year
- ☆31Updated last year
- Here we will test various linear attention designs.☆58Updated last year
- Code and data for paper "(How) do Language Models Track State?"☆14Updated 2 months ago
- ☆54Updated 10 months ago
- ☆32Updated 8 months ago
- Xmixers: A collection of SOTA efficient token/channel mixers☆11Updated 6 months ago
- ☆28Updated 6 months ago
- Code for the paper: https://arxiv.org/pdf/2309.06979.pdf☆19Updated 10 months ago
- Reference implementation of "Softmax Attention with Constant Cost per Token" (Heinsen, 2024)☆24Updated 11 months ago
- ☆53Updated 8 months ago
- ☆17Updated last month
- ☆23Updated 8 months ago
- ☆20Updated last year
- Official repository for the paper "Exploring the Promise and Limits of Real-Time Recurrent Learning" (ICLR 2024)☆10Updated last year
- ☆11Updated last year
- Parallelizing non-linear sequential models over the sequence length☆51Updated 4 months ago
- Official Code Repository for the paper "Key-value memory in the brain"☆26Updated 3 months ago
- Combining SOAP and MUON☆16Updated 3 months ago
- ☆52Updated last year
- Blog post☆17Updated last year
- RWKV-X is a Linear Complexity Hybrid Language Model based on the RWKV architecture, integrating Sparse Attention to improve the model's l…☆37Updated last month
- Official code for the paper "Attention as a Hypernetwork"☆36Updated 11 months ago
- Using FlexAttention to compute attention with different masking patterns☆43Updated 8 months ago