bitsandbytes_ops_cpu

Python implementation of bisandbytes ops for CPU

Run tests

python test_ops.py

Run benchmarks

Eager vs. inductor (torch.compile)

# Latency
numactl -m <node id> -C <core list> python benchmarks.py --all
# Profiling
numactl -m <node id> -C <core list> python benchmarks.py --all --profile