Closed beckermr closed 2 months ago
Comparing fix-bench
(5283773) with main
(d03e151)
⚡ 1
improvements
❌ 5 (👁 5)
regressions
✅ 5
untouched benchmarks
Benchmark | main |
fix-bench |
Change | |
---|---|---|---|---|
👁 | test_benchmarks_interpolated_image[compile] |
77.4 ms | 1,854.4 ms | -95.83% |
👁 | test_benchmarks_lanczos_interp[kval-conserve_dc-compile] |
42.7 ms | 5,358.5 ms | -99.2% |
👁 | test_benchmarks_lanczos_interp[kval-no_conserve_dc-compile] |
4.1 ms | 1,560 ms | -99.74% |
👁 | test_benchmarks_lanczos_interp[xval-conserve_dc-compile] |
2.5 ms | 315.5 ms | -99.2% |
⚡ | test_benchmarks_lanczos_interp[xval-conserve_dc-run] |
2.5 ms | 2.1 ms | +17.97% |
👁 | test_benchmarks_lanczos_interp[xval-no_conserve_dc-compile] |
1.7 ms | 261.6 ms | -99.35% |
Again this PR has no functional changes and is for testing changes in #101. I am going to merge but I can respond to any CR later.
This PR fixes a bug in #102 where the jax jit caches were not cleared properly.