Liger Kernel Integration

Section under construction. Feel free to contribute!

Liger Kernel is a collection of Triton kernels designed specifically for LLM training. It can effectively increase multi-GPU training throughput by 20% and reduce memory usage by 60%. That way, we can 4x our context length, as described in the benchmark below. They have implemented Hugging Face compatible RMSNorm, RoPE, SwiGLU, CrossEntropy, FusedLinearCrossEntropy, with more to come. The kernel works out of the box with FlashAttention, PyTorch FSDP, and Microsoft DeepSpeed.

With this memory reduction, you can potentially turn off cpu_offloading or gradient checkpointing to further boost the performance.

Speed Up	Memory Reduction

1. To use Liger-Kernel in SFTTrainer, first install it by:

   Copied

   pip install liger-kernel

2. Once installed, set use_liger_kernel in SFTConfig. No other changes are needed!

training_args = SFTConfig(
    use_liger_kernel=True,
    ...
)

To learn more about Liger-Kernel, visit their official repository.