@@ -25,6 +25,8 @@ PR or email us. We'd very much like to hear from you!
[library](https://www.mosaicml.com/blog/gpt-3-quality-for-500k). Composer is a
library for efficient neural network training.
- EleutherAI's [GPT-NeoX](https://github.com/EleutherAI/gpt-neox/pull/725). This is a research library for training large language transformer models at scale based on NVIDIA's Megatron-LM and Microsoft's DeepSpeed.
## MLPerf benchmarks
[MLPerf](https://mlcommons.org/en/) is a competitive machine learning performance benchmark. FlashAttention
