Show HN: Fast NF4 dequantization Triton kernel (1.41x faster than bitsandbytes)

Griffith-71 pts0 comments

I wanted to see if I could optimize the dequantization bottleneck during 4-bit LLM inference. By writing a custom kernel in Triton to optimize memory access patterns, I managed to get up to a 1.41x speedup over the standard bitsandbytes implementation. Check out the source code and benchmarks, feedback is highly appreciated!

dequantization triton kernel bitsandbytes quot optimize

Related Articles