Closed utkarsh530 closed 6 months ago
Fitzhugh Nagamo Benchmarks https://docs.sciml.ai/SciMLBenchmarksOutput/stable/ParameterEstimation/FitzHughNagumoParameterEstimation/:
BenchmarkTools.Trial: 28 samples with 1 evaluation.
Range (min … max): 170.824 ms … 211.453 ms ┊ GC (min … max): 0.00% … 11.76%
Time (median): 179.649 ms ┊ GC (median): 0.00%
Time (mean ± σ): 180.377 ms ± 9.862 ms ┊ GC (mean ± σ): 1.84% ± 2.55%
█ ▁▁
█▄▆▆▁▁▁▁▁▁▁▄▁▁▄▁▁▁▁▁▁▁██▄▄▁▁▁▁▁▁▄▄▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▄ ▁
171 ms Histogram: frequency by time 211 ms <
Memory estimate: 2.06 MiB, allocs estimate: 42709.
julia> @show gsol
gsol = PSOGPU.SPSOGBest{SVector{4, Float32}, Float32}(Float32[1.4175674, 0.934993, 0.049795125, 0.5001013], 3.5554986f-5)
PSOGPU.SPSOGBest{SVector{4, Float32}, Float32}(Float32[1.4175674, 0.934993, 0.049795125, 0.5001013], 3.5554986f-5)
Lotka Volterra https://docs.sciml.ai/SciMLSensitivity/stable/tutorials/parameter_estimation_ode/ :
julia> @benchmark PSOGPU.parameter_estim_ode!($prob, $(deepcopy(solver_cache)),
$lb,
$ub, Val(adaptive); saveat = t, dt = 0.1, backend = CUDABackend())
BenchmarkTools.Trial: 57 samples with 1 evaluation.
Range (min … max): 78.997 ms … 117.178 ms ┊ GC (min … max): 0.00% … 21.38%
Time (median): 80.868 ms ┊ GC (median): 0.00%
Time (mean ± σ): 87.706 ms ± 10.700 ms ┊ GC (mean ± σ): 3.43% ± 4.84%
█▃▃
███▆▄▃▁▁▁▃▁▁▃▁▄▁▃▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▆▇▃▃▁▃▃▃▁▃▁▃▁▃▁▁▁▁▁▁▃▁▁▁▁▁▁▃ ▁
79 ms Histogram: frequency by time 114 ms <
Memory estimate: 2.13 MiB, allocs estimate: 43106.
julia> @time gsol = PSOGPU.parameter_estim_ode!(prob, solver_cache,
lb,
ub, Val(adaptive); saveat = t, dt = 0.1, maxiters = 100)
0.079444 seconds (43.10 k allocations: 2.124 MiB)
PSOGPU.SPSOGBest{SVector{4, Float64}, Float64}([1.8749830934479235, 1.874982914866944, 2.1886256202246708, 2.18862664910988], 4.5790030071118923e-11)
Why not make these SciMLBenchmarks runs?
Yeah, I plan to do it once I tag the release and populate some docs.
Checklist
Additional context
Add any other context about the problem here.