mrq
/
bitsandbytes-rocm
1
0
Fork 1
Code Issues 1 Pull Requests Packages Projects Releases Wiki Activity
bitsandbytes-rocm/benchmarking/switchback
History
Tim Dettmers dd562c24f1 Refactored simulated fp8 modules into research.nn. 2023-04-12 11:24:44 +07:00
..
README.md Refactored simulated fp8 modules into research.nn. 2023-04-12 11:24:44 +07:00
info_a100_py2.jsonl Refactored simulated fp8 modules into research.nn. 2023-04-12 11:24:44 +07:00
make_plot_with_jsonl.py Refactored simulated fp8 modules into research.nn. 2023-04-12 11:24:44 +07:00
plot_with_info.pdf Refactored simulated fp8 modules into research.nn. 2023-04-12 11:24:44 +07:00
speed_benchmark.py Refactored simulated fp8 modules into research.nn. 2023-04-12 11:24:44 +07:00

README.md

Steps:

  1. Run python speed_benchmark/speed_benchmark.py which times operations and writes their time to speed_benchmark/info_a100_py2.jsonl (change the name of the jsonl to a different name for your profiling).
  2. Run python speed_benchmark/make_plot_with_jsonl.py, which produces the speed_benchmark/plot_with_info.pdf. Again make sure you change the jsonl which is being processed.