bloc97 / DeMoLinks
DeMo: Decoupled Momentum Optimization
☆190Updated 8 months ago
Alternatives and similar repositories for DeMo
Users that are interested in DeMo are comparing it to the libraries listed below
Sorting:
- an open source reproduction of NVIDIA's nGPT (Normalized Transformer with Representation Learning on the Hypersphere)☆103Updated 4 months ago
- supporting pytorch FSDP for optimizers☆84Updated 7 months ago
- Repo for "LoLCATs: On Low-Rank Linearizing of Large Language Models"☆244Updated 6 months ago
- ☆100Updated last week
- research impl of Native Sparse Attention (2502.11089)☆58Updated 5 months ago
- PyTorch implementation of models from the Zamba2 series.☆184Updated 6 months ago
- Normalized Transformer (nGPT)☆185Updated 8 months ago
- The simplest, fastest repository for training/finetuning medium-sized GPTs.☆149Updated last month
- Explorations into the proposal from the paper "Grokfast, Accelerated Grokking by Amplifying Slow Gradients"☆101Updated 7 months ago
- Collection of autoregressive model implementation☆86Updated 3 months ago
- Tree Attention: Topology-aware Decoding for Long-Context Attention on GPU clusters☆127Updated 8 months ago
- Focused on fast experimentation and simplicity☆76Updated 7 months ago
- RWKV-7: Surpassing GPT☆94Updated 8 months ago
- NanoGPT-speedrunning for the poor T4 enjoyers☆68Updated 3 months ago
- Tiny re-implementation of MDM in style of LLaDA and nano-gpt speedrun☆55Updated 4 months ago
- Q-GaLore: Quantized GaLore with INT4 Projection and Layer-Adaptive Low-Rank Gradients.☆198Updated last year
- smolLM with Entropix sampler on pytorch☆150Updated 9 months ago
- EvaByte: Efficient Byte-level Language Models at Scale☆103Updated 3 months ago
- ☆134Updated 11 months ago
- WIP☆93Updated 11 months ago
- Pytorch implementation of the PEER block from the paper, Mixture of A Million Experts, by Xu Owen He at Deepmind☆127Updated 11 months ago
- Attention Kernels for Symmetric Power Transformers☆102Updated last week
- Getting crystal-like representations with harmonic loss☆193Updated 4 months ago
- PCCL (Prime Collective Communications Library) implements fault tolerant collective communications over IP☆98Updated 2 weeks ago
- ☆81Updated last year
- Implementations of attention with the softpick function, naive and FlashAttention-2☆81Updated 3 months ago
- Quick implementation of nGPT, learning entirely on the hypersphere, from NvidiaAI☆288Updated 2 months ago
- Token Omission Via Attention☆128Updated 9 months ago
- A MAD laboratory to improve AI architecture designs 🧪☆123Updated 7 months ago
- Minimal (400 LOC) implementation Maximum (multi-node, FSDP) GPT training☆130Updated last year