cloneofsimo / min-fsdpLinks
☆87Updated last year
Alternatives and similar repositories for min-fsdp
Users that are interested in min-fsdp are comparing it to the libraries listed below
Sorting:
- Minimal (400 LOC) implementation Maximum (multi-node, FSDP) GPT training☆130Updated last year
- supporting pytorch FSDP for optimizers☆84Updated 8 months ago
- The simplest, fastest repository for training/finetuning medium-sized GPTs.☆152Updated last month
- Experiment of using Tangent to autodiff triton☆80Updated last year
- WIP☆94Updated last year
- ☆53Updated last year
- ☆118Updated last year
- A library for unit scaling in PyTorch☆129Updated last month
- ☆34Updated 11 months ago
- Simple implementation of muP, based on Spectral Condition for Feature Learning. The implementation is SGD only, dont use it for Adam☆85Updated last year
- Understand and test language model architectures on synthetic tasks.☆221Updated last month
- An implementation of the Llama architecture, to instruct and delight☆21Updated 2 months ago
- nanoGPT-like codebase for LLM training☆102Updated 3 months ago
- A set of Python scripts that makes your experience on TPU better☆54Updated last year
- A fusion of a linear layer and a cross entropy loss, written for pytorch in triton.☆70Updated last year
- A MAD laboratory to improve AI architecture designs 🧪☆124Updated 8 months ago
- Large scale 4D parallelism pre-training for 🤗 transformers in Mixture of Experts *(still work in progress)*☆87Updated last year
- 📄Small Batch Size Training for Language Models☆42Updated 2 weeks ago
- JAX implementation of the Llama 2 model☆219Updated last year
- Demo of the unit_scaling library, showing how a model can be easily adapted to train in FP8.☆46Updated last year
- Language models scale reliably with over-training and on downstream tasks☆98Updated last year
- Minimal but scalable implementation of large language models in JAX☆35Updated last month
- Automatically take good care of your preemptible TPUs☆36Updated 2 years ago
- Simple and efficient pytorch-native transformer training and inference (batched)☆78Updated last year
- Triton Implementation of HyperAttention Algorithm☆48Updated last year
- Code for NeurIPS 2024 Spotlight: "Scaling Laws and Compute-Optimal Training Beyond Fixed Training Durations"☆81Updated 9 months ago
- A place to store reusable transformer components of my own creation or found on the interwebs☆59Updated last week
- CUDA and Triton implementations of Flash Attention with SoftmaxN.☆73Updated last year
- ☆20Updated 2 years ago
- Easily run PyTorch on multiple GPUs & machines☆46Updated last month