Laz4rz / mupLinks
Minimal (truly) muP implementation, consistent with TP4 and TP5 papers notation
☆14Updated 3 months ago
Alternatives and similar repositories for mup
Users that are interested in mup are comparing it to the libraries listed below
Sorting:
- Code for the paper "Function-Space Learning Rates"☆23Updated 2 months ago
- Minimal (400 LOC) implementation Maximum (multi-node, FSDP) GPT training☆131Updated last year
- The simplest, fastest repository for training/finetuning medium-sized GPTs.☆156Updated 2 months ago
- ☆87Updated last year
- WIP☆94Updated last year
- A basic pure pytorch implementation of flash attention☆16Updated 10 months ago
- supporting pytorch FSDP for optimizers☆84Updated 8 months ago
- Tiny re-implementation of MDM in style of LLaDA and nano-gpt speedrun☆56Updated 5 months ago
- Universal Neurons in GPT2 Language Models☆30Updated last year
- Code for NeurIPS 2024 Spotlight: "Scaling Laws and Compute-Optimal Training Beyond Fixed Training Durations"☆82Updated 10 months ago
- ☆85Updated last year
- The official github repo for "Diffusion Language Models are Super Data Learners".☆107Updated 3 weeks ago
- ☆25Updated 3 months ago
- Unofficial Implementation of Selective Attention Transformer☆17Updated 10 months ago
- Official PyTorch implementation and models for paper "Diffusion Beats Autoregressive in Data-Constrained Settings". We find diffusion mod…☆86Updated this week
- The official repository for SkyLadder: Better and Faster Pretraining via Context Window Scheduling☆33Updated last week
- ☆34Updated 7 months ago
- $100K or 100 Days: Trade-offs when Pre-Training with Academic Resources☆143Updated 3 months ago
- Language models scale reliably with over-training and on downstream tasks☆98Updated last year
- Code to reproduce "Transformers Can Do Arithmetic with the Right Embeddings", McLeish et al (NeurIPS 2024)☆192Updated last year
- ☆11Updated 6 months ago
- 📄Small Batch Size Training for Language Models☆57Updated this week
- ☆69Updated last year
- Simple implementation of muP, based on Spectral Condition for Feature Learning. The implementation is SGD only, dont use it for Adam☆85Updated last year
- ☆53Updated last year
- [NeurIPS 2024] Low rank memory efficient optimizer without SVD☆30Updated 2 months ago
- [ICML 2025] Roll the dice & look before you leap: Going beyond the creative limits of next-token prediction☆65Updated 3 months ago
- Code for reproducing our paper "Not All Language Model Features Are Linear"☆77Updated 9 months ago
- The evaluation framework for training-free sparse attention in LLMs☆91Updated 2 months ago
- Implementations of attention with the softpick function, naive and FlashAttention-2☆83Updated 4 months ago