Training Custom Mixtral Model with DeepSpeed

less than 1 minute read

Placeholder for the blog logging how I trained a custom Mixtral model with DeepSpeed.

Twitter Facebook LinkedIn

Comments

Efficient Gather-and-scatter Feed-forward Network Kernel with Triton

13 minute read

In our recent work Learn to be efficient: Build structured sparsity in large language models, we propose a novel method to build structured sparsity in large...

Use Nsight System to Profile a Model Training with DeepSpeed on Multi-Node Cluster

11 minute read

This post is to log how I managed to profile a model training running on multiple nodes in a cluster with DeepSpeed and Nsight System. Click here to jump to ...

Custom Gather-scatter Operator by CUTLASS

19 minute read

This blog is to log my experience of building efficient custom operator based on CUTLASS. Jump to the final implementation of gather and scatter matrix multi...

Compact Inference with CUDA graph and StaticCache

14 minute read

This post is to log a minimum prototype of LLM inference with CUDA graph to eliminate bubbles between kernel launches. Click here to jump to the final implem...

Xueshen Liu

Comments

You May Also Enjoy

Efficient Gather-and-scatter Feed-forward Network Kernel with Triton

Use Nsight System to Profile a Model Training with DeepSpeed on Multi-Node Cluster

Custom Gather-scatter Operator by CUTLASS

Compact Inference with CUDA graph and StaticCache