automl / unlocking_state_tracking
Expanding linear RNN state-transition matrix eigenvalues to include negatives improves state-tracking tasks and language modeling without added training or inference costs.
☆14Updated last week
Alternatives and similar repositories for unlocking_state_tracking:
Users that are interested in unlocking_state_tracking are comparing it to the libraries listed below
- Experiments on the impact of depth in transformers and SSMs.☆23Updated 4 months ago
- Parallel Associative Scan for Language Models☆18Updated last year
- ☆33Updated last year
- ☆31Updated 11 months ago
- Efficient PScan implementation in PyTorch☆16Updated last year
- ☆47Updated last year
- ☆24Updated 6 months ago
- ☆30Updated 5 months ago
- Combining SOAP and MUON☆13Updated last month
- ☆14Updated 3 weeks ago
- The accompanying code for "Simplifying and Understanding State Space Models with Diagonal Linear RNNs" (Ankit Gupta, Harsh Mehta, Jonatha…☆20Updated 2 years ago
- Blog post☆17Updated last year
- Here we will test various linear attention designs.☆60Updated 11 months ago
- ☆30Updated 4 months ago
- Official Code Repository for the paper "Key-value memory in the brain"☆24Updated last month
- ☆52Updated 8 months ago
- ☆18Updated 9 months ago
- Code for the paper: https://arxiv.org/pdf/2309.06979.pdf☆19Updated 7 months ago
- ☆52Updated 5 months ago
- ☆38Updated last year
- Source-to-Source Debuggable Derivatives in Pure Python☆15Updated last year
- Reference implementation of "Softmax Attention with Constant Cost per Token" (Heinsen, 2024)☆24Updated 9 months ago
- ☆11Updated last year
- Engineering the state of RNN language models (Mamba, RWKV, etc.)☆32Updated 10 months ago
- Parallelizing non-linear sequential models over the sequence length☆51Updated 2 months ago
- ☆30Updated last year
- Xmixers: A collection of SOTA efficient token/channel mixers☆11Updated 4 months ago
- Official repository of paper "RNNs Are Not Transformers (Yet): The Key Bottleneck on In-context Retrieval"☆26Updated 11 months ago
- ☆14Updated 3 months ago
- Official repository for the paper "Exploring the Promise and Limits of Real-Time Recurrent Learning" (ICLR 2024)☆11Updated last year