microsoft / mutransformers
some common Huggingface transformers in maximal update parametrization (µP)
☆76Updated 2 years ago
Related projects: ⓘ
- A MAD laboratory to improve AI architecture designs 🧪☆84Updated 4 months ago
- Understand and test language model architectures on synthetic tasks.☆156Updated 4 months ago
- ☆66Updated 3 months ago
- A fast implementation of T5/UL2 in PyTorch using Flash Attention☆60Updated this week
- Inference code for LLaMA models in JAX☆108Updated 3 months ago
- ☆27Updated this week
- ☆69Updated 4 months ago
- Language models scale reliably with over-training and on downstream tasks☆91Updated 5 months ago
- Experiments for efforts to train a new and improved t5☆76Updated 5 months ago
- The source code of our work "Prepacking: A Simple Method for Fast Prefilling and Increased Throughput in Large Language Models"☆57Updated 5 months ago
- ☆64Updated 2 years ago
- ☆73Updated 5 months ago
- Code for the paper "The Impact of Positional Encoding on Length Generalization in Transformers", NeurIPS 2023☆123Updated 4 months ago
- ☆160Updated last year
- Code repository for the c-BTM paper☆105Updated 11 months ago
- ☆48Updated 3 months ago
- Tree Attention: Topology-aware Decoding for Long-Context Attention on GPU clusters☆94Updated 2 weeks ago
- Code for exploring Based models from "Simple linear attention language models balance the recall-throughput tradeoff"☆206Updated last month
- ☆129Updated last year
- Exploring finetuning public checkpoints on filter 8K sequences on Pile☆115Updated last year
- Randomized Positional Encodings Boost Length Generalization of Transformers☆78Updated 6 months ago
- Transformer with Mu-Parameterization, implemented in Jax/Flax. Supports FSDP on TPU pods.☆29Updated 3 weeks ago
- Large scale 4D parallelism pre-training for 🤗 transformers in Mixture of Experts *(still work in progress)*☆77Updated 9 months ago
- ☆53Updated 8 months ago
- Experiment of using Tangent to autodiff triton☆66Updated 7 months ago
- Cold Compress is a hackable, lightweight, and open-source toolkit for creating and benchmarking cache compression methods built on top of…☆73Updated last month
- Minimal (400 LOC) implementation Maximum (multi-node, FSDP) GPT training☆110Updated 5 months ago
- ☆68Updated 2 months ago
- Implementation of GateLoop Transformer in Pytorch and Jax☆86Updated 3 months ago
- Machine Learning eXperiment Utilities☆42Updated 3 months ago