DeepSeed
Deep learning optimization library making distributed training easy
...With just a single GPU, ZeRO-Offload of DeepSpeed can train models with over 10B parameters, 10x bigger than the state of arts, democratizing multi-billion-parameter model training such that many deep learning scientists can explore bigger and better models. Sparse attention of DeepSpeed powers an order-of-magnitude longer input sequence and obtains up to 6x faster execution comparing with dense transformers.