xrsrke / pipegoose
Large scale 4D parallelism pre-training for ๐ค transformers in Mixture of Experts *(still work in progress)*
โ81Updated last year
Alternatives and similar repositories for pipegoose:
Users that are interested in pipegoose are comparing it to the libraries listed below
- some common Huggingface transformers in maximal update parametrization (ยตP)โ80Updated 3 years ago
- โ75Updated 8 months ago
- โ20Updated last year
- โ43Updated last year
- Minimal (400 LOC) implementation Maximum (multi-node, FSDP) GPT trainingโ122Updated 10 months ago
- Language models scale reliably with over-training and on downstream tasksโ96Updated 11 months ago
- A set of Python scripts that makes your experience on TPU betterโ50Updated 8 months ago
- โ58Updated 3 years ago
- The simplest, fastest repository for training/finetuning medium-sized GPTs.โ100Updated 3 months ago
- Automatically take good care of your preemptible TPUsโ36Updated last year
- Experiment of using Tangent to autodiff tritonโ76Updated last year
- Collection of autoregressive model implementationโ83Updated last month
- supporting pytorch FSDP for optimizersโ79Updated 3 months ago
- Exploring finetuning public checkpoints on filter 8K sequences on Pileโ115Updated last year
- Experiments with generating opensource language model assistantsโ97Updated last year
- Simple and efficient pytorch-native transformer training and inference (batched)โ69Updated 11 months ago
- A MAD laboratory to improve AI architecture designs ๐งชโ107Updated 2 months ago
- Inference code for LLaMA models in JAXโ116Updated 9 months ago
- Code for the paper "The Impact of Positional Encoding on Length Generalization in Transformers", NeurIPS 2023โ130Updated 10 months ago
- Understand and test language model architectures on synthetic tasks.โ183Updated last week
- A toolkit for scaling law research โโ48Updated last month
- Prune transformer layersโ68Updated 9 months ago
- Comprehensive analysis of difference in performance of QLora, Lora, and Full Finetunes.โ82Updated last year
- See https://github.com/cuda-mode/triton-index/ instead!โ11Updated 10 months ago
- Fast, Modern, Memory Efficient, and Low Precision PyTorch Optimizersโ86Updated 7 months ago
- โ49Updated last year
- JAX implementation of the Llama 2 modelโ216Updated last year
- โ53Updated last year
- Pytorch/XLA SPMD Test code in Google TPUโ23Updated 11 months ago
- HomebrewNLP in JAX flavour for maintable TPU-Trainingโ48Updated last year