1
|
{"args": {"method": "ep", "dataset": "mnist", "depth": 1, "width": 500, "epochs": 25, "batch_size": 20, "train_examples": 50000, "max_batches": 0, "eta": null, "momentum": 0.9, "act": "tanh", "residual": 0, "feedback_scale": 1.0, "pepita_f_scale": 0.05, "pepita_decay_epochs": null, "keep_prob": 0.9, "ff_threshold": 2.0, "ep_beta": null, "ep_dt": 0.5, "ep_free_steps": null, "ep_nudge_steps": null, "ep_persistent": 1, "canonical_preprocess": 1, "probe_bs": 512, "seed": 0, "device": "cuda", "outdir": "results", "tag": "ep_original_v2_mnist_ep_w500_d1_s0"}, "resolved_protocol": {"beta": 0.5, "free_steps": 20, "nudge_steps": 4, "canonical_depth": true, "learning_rates": [0.1, 0.05], "persistent_particles": true, "train_examples": 50000}, "method_source": {"paper": "https://www.frontiersin.org/journals/computational-neuroscience/articles/10.3389/fncom.2017.00024/full", "code": "https://github.com/bscellier/Towards-a-Biologically-Plausible-Backprop", "protocol": "two-phase energy-gradient dynamics; persistent free particles; random beta sign"}, "provenance": {"git_commit": "5b9de3f36c91ea8ad57fc4dae5aa0c04f08a7a26", "git_dirty": false}, "steps": [{"epoch_end": 0, "test_acc": 0.955, "test_loss": 0.07738499794006348, "train_loss": 0.01003960333764553}, {"epoch_end": 1, "test_acc": 0.9648, "test_loss": 0.06048058605194092, "train_loss": 0.0037819012068212032}, {"epoch_end": 2, "test_acc": 0.9658, "test_loss": 0.056952181243896484, "train_loss": 0.0026113467756658792}, {"epoch_end": 3, "test_acc": 0.9668, "test_loss": 0.057208501434326174, "train_loss": 0.0011136147659271955}, {"epoch_end": 4, "test_acc": 0.971, "test_loss": 0.050622566032409666, "train_loss": 0.00023063279513735324}, {"epoch_end": 5, "test_acc": 0.9687, "test_loss": 0.052987321662902834, "train_loss": 0.00016184779815375805}, {"epoch_end": 6, "test_acc": 0.9693, "test_loss": 0.05012830390930176, "train_loss": 0.000521468638908118}, {"epoch_end": 7, "test_acc": 0.9723, "test_loss": 0.05139166889190674, "train_loss": 6.770728759875055e-06}, {"epoch_end": 8, "test_acc": 0.9657, "test_loss": 0.06045001754760742, "train_loss": 0.00023175214300863445}, {"epoch_end": 9, "test_acc": 0.9677, "test_loss": 0.0564575044631958, "train_loss": 5.406913624028675e-05}, {"epoch_end": 10, "test_acc": 0.9681, "test_loss": 0.056905843353271486, "train_loss": 1.3698017653496208e-07}, {"epoch_end": 11, "test_acc": 0.9686, "test_loss": 0.05428710899353027, "train_loss": 0.0}, {"epoch_end": 12, "test_acc": 0.971, "test_loss": 0.04870379829406738, "train_loss": 0.0}, {"epoch_end": 13, "test_acc": 0.9742, "test_loss": 0.04683217372894287, "train_loss": 0.0}, {"epoch_end": 14, "test_acc": 0.97, "test_loss": 0.04981306781768799, "train_loss": 0.0}, {"epoch_end": 15, "test_acc": 0.973, "test_loss": 0.049349253273010256, "train_loss": 1.141531811299501e-05}, {"epoch_end": 16, "test_acc": 0.9707, "test_loss": 0.04930688552856445, "train_loss": 0.0}, {"epoch_end": 17, "test_acc": 0.9687, "test_loss": 0.05342885799407959, "train_loss": 0.0}, {"epoch_end": 18, "test_acc": 0.9725, "test_loss": 0.049415488052368166, "train_loss": 0.0}, {"epoch_end": 19, "test_acc": 0.9706, "test_loss": 0.050550705146789554, "train_loss": 0.0}, {"epoch_end": 20, "test_acc": 0.9713, "test_loss": 0.0508961033821106, "train_loss": 0.0}, {"epoch_end": 21, "test_acc": 0.9715, "test_loss": 0.04933317308425903, "train_loss": 0.0}, {"epoch_end": 22, "test_acc": 0.9706, "test_loss": 0.051107603645324705, "train_loss": 0.0}, {"epoch_end": 23, "test_acc": 0.9719, "test_loss": 0.050396513938903806, "train_loss": 0.0}, {"epoch_end": 24, "test_acc": 0.9723, "test_loss": 0.05007862377166748, "train_loss": 0.0}], "final": {"test_acc": 0.9723, "test_loss": 0.05007862377166748, "wall_s": 895.3023679256439}}
|
