Download results/ablations/logs/eval_size_data.log from matCercola18/DiffusionControl: direct link, hf CLI and curl.
- Browser
- Download file 5.07 kB
-
https://huggingface.co/matCercola18/DiffusionControl/resolve/main/results/ablations/logs/eval_size_data.log
- Command line
-
hf download hf://matCercola18/DiffusionControl/results/ablations/logs/eval_size_data.log
-
curl -L -o eval_size_data.log https://huggingface.co/matCercola18/DiffusionControl/resolve/main/results/ablations/logs/eval_size_data.log
5.07 kB
| device=cuda training seeds=[42, 43, 44] eval noise seed=0 | |
| [size] hidden=64 (~1/4 width) (base_tag=size_quarter, seeds=[42, 43, 44]) | |
| evaluating size_quarter seed=42 <- models/Best_pendulum_abl_size_quarter.pth (eval noise seed=0) | |
| /home/cercola/matte_server/DiffusionControl/nonlinear/data_generation/pendulum_lmi.py:149: UserWarning: Solution may be inaccurate. Try another solver, adjusting the solver settings, or solve with verbose=True for more information. | |
| prob.solve(solver=solver, verbose=False) | |
| evaluating size_quarter seed=43 <- models/Best_pendulum_abl_size_quarter_s43.pth (eval noise seed=0) | |
| evaluating size_quarter seed=44 <- models/Best_pendulum_abl_size_quarter_s44.pth (eval noise seed=0) | |
| [size] hidden=256 (fair reference, ablation trainer) (base_tag=size_ref, seeds=[42, 43, 44]) | |
| evaluating size_ref seed=42 <- models/Best_pendulum_abl_size_ref.pth (eval noise seed=0) | |
| evaluating size_ref seed=43 <- models/Best_pendulum_abl_size_ref_s43.pth (eval noise seed=0) | |
| evaluating size_ref seed=44 <- models/Best_pendulum_abl_size_ref_s44.pth (eval noise seed=0) | |
| [size] hidden=512 (~2x width) (base_tag=size_2x, seeds=[42, 43, 44]) | |
| evaluating size_2x seed=42 <- models/Best_pendulum_abl_size_2x.pth (eval noise seed=0) | |
| evaluating size_2x seed=43 <- models/Best_pendulum_abl_size_2x_s43.pth (eval noise seed=0) | |
| evaluating size_2x seed=44 <- models/Best_pendulum_abl_size_2x_s44.pth (eval noise seed=0) | |
| [size] hidden=256 (DEPLOYED checkpoint -- pre-existing train_pendulum.py run, not part of the 3-seed sweep) <- models/Best_pendulum_cond.pth | |
| === SIZE ABLATION (pendulum, N=40, K=8, shared cached oracle, eval noise seed=0, 3 training seeds=[42, 43, 44]) === | |
| - hidden=64 (~1/4 width) | |
| params=255,563 time_to_best=405.9s | |
| median regret vs oracle: +0.68% +/- 0.16pp over seeds [s42=+0.58%, s43=+0.87%, s44=+0.60%] | |
| coverage (mean over seeds): certified>=1/plant 0.992 stable>=1/plant 1.000 per-sample stable 0.999 | |
| - hidden=256 (fair reference, ablation trainer) | |
| params=3,823,883 time_to_best=421.4s | |
| median regret vs oracle: +0.92% +/- 0.22pp over seeds [s42=+1.06%, s43=+0.67%, s44=+1.03%] | |
| coverage (mean over seeds): certified>=1/plant 1.000 stable>=1/plant 1.000 per-sample stable 1.000 | |
| - hidden=512 (~2x width) | |
| params=15,118,859 time_to_best=441.1s | |
| median regret vs oracle: +0.86% +/- 0.38pp over seeds [s42=+0.98%, s43=+0.43%, s44=+1.16%] | |
| coverage (mean over seeds): certified>=1/plant 1.000 stable>=1/plant 1.000 per-sample stable 1.000 | |
| - hidden=256 (DEPLOYED checkpoint -- pre-existing train_pendulum.py run, not part of the 3-seed sweep) | |
| params=3,823,883 time_to_best=232.0s | |
| median regret vs oracle: +1.15% +/- 0.00pp over seeds [ssingle=+1.15%] | |
| coverage (mean over seeds): certified>=1/plant 1.000 stable>=1/plant 1.000 per-sample stable 1.000 | |
| [data] 25% data (650/2600 examples) (base_tag=data_p25, seeds=[42, 43, 44]) | |
| evaluating data_p25 seed=42 <- models/Best_pendulum_abl_data_p25.pth (eval noise seed=0) | |
| evaluating data_p25 seed=43 <- models/Best_pendulum_abl_data_p25_s43.pth (eval noise seed=0) | |
| evaluating data_p25 seed=44 <- models/Best_pendulum_abl_data_p25_s44.pth (eval noise seed=0) | |
| [data] 50% data (1300/2600 examples) (base_tag=data_p50, seeds=[42, 43, 44]) | |
| evaluating data_p50 seed=42 <- models/Best_pendulum_abl_data_p50.pth (eval noise seed=0) | |
| evaluating data_p50 seed=43 <- models/Best_pendulum_abl_data_p50_s43.pth (eval noise seed=0) | |
| evaluating data_p50 seed=44 <- models/Best_pendulum_abl_data_p50_s44.pth (eval noise seed=0) | |
| [data] 100% data (= size_ref, same 3 seeds, reused) (base_tag=size_ref, seeds=[42, 43, 44]) | |
| [data] hidden=256 (DEPLOYED checkpoint -- pre-existing train_pendulum.py run, not part of the 3-seed sweep) <- models/Best_pendulum_cond.pth | |
| === DATA ABLATION (pendulum, N=40, K=8, shared cached oracle, eval noise seed=0, 3 training seeds=[42, 43, 44]) === | |
| - 25% data (650/2600 examples) | |
| params=3,823,883 time_to_best=261.8s | |
| median regret vs oracle: +0.29% +/- 0.27pp over seeds [s42=+0.61%, s43=+0.16%, s44=+0.12%] | |
| coverage (mean over seeds): certified>=1/plant 0.992 stable>=1/plant 1.000 per-sample stable 1.000 | |
| - 50% data (1300/2600 examples) | |
| params=3,823,883 time_to_best=355.6s | |
| median regret vs oracle: +1.00% +/- 0.23pp over seeds [s42=+1.22%, s43=+0.76%, s44=+1.02%] | |
| coverage (mean over seeds): certified>=1/plant 1.000 stable>=1/plant 1.000 per-sample stable 1.000 | |
| - 100% data (= size_ref, same 3 seeds, reused) | |
| params=3,823,883 time_to_best=421.4s | |
| median regret vs oracle: +0.92% +/- 0.22pp over seeds [s42=+1.06%, s43=+0.67%, s44=+1.03%] | |
| coverage (mean over seeds): certified>=1/plant 1.000 stable>=1/plant 1.000 per-sample stable 1.000 | |
| - hidden=256 (DEPLOYED checkpoint -- pre-existing train_pendulum.py run, not part of the 3-seed sweep) | |
| params=3,823,883 time_to_best=232.0s | |
| median regret vs oracle: +1.15% +/- 0.00pp over seeds [ssingle=+1.15%] | |
| coverage (mean over seeds): certified>=1/plant 1.000 stable>=1/plant 1.000 per-sample stable 1.000 | |