1
|
{"args": {"method": "ep", "dataset": "mnist", "depth": 2, "width": 500, "epochs": 60, "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": 3, "device": "cuda", "outdir": "results", "tag": "ep_depth2_v1_mnist_ep_w500_d2_s3"}, "resolved_protocol": {"beta": 1.0, "free_steps": 150, "nudge_steps": 6, "canonical_depth": true, "learning_rates": [0.4, 0.1, 0.01], "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", "code_revision": "cb73f76d997924b6198355e22b50ed7e97cec684", "protocol": "two-phase energy-gradient dynamics; persistent free particles; random beta sign"}, "provenance": {"git_commit": "07a9d5afe08fee5824c325edfb3807a2f9aede8f", "git_dirty": false}, "steps": [{"epoch_end": 0, "test_acc": 0.3758, "test_loss": 0.9226919128417969, "train_loss": 0.09278693050146103}, {"epoch_end": 1, "test_acc": 0.7534, "test_loss": 0.29963711700439455, "train_loss": 0.02746838517487049}, {"epoch_end": 2, "test_acc": 0.7617, "test_loss": 0.27579493408203126, "train_loss": 0.02499999850988388}, {"epoch_end": 3, "test_acc": 0.8472, "test_loss": 0.1984871795654297, "train_loss": 0.014302253723144531}, {"epoch_end": 4, "test_acc": 0.8551, "test_loss": 0.18292726135253906, "train_loss": 0.017146730795502663}, {"epoch_end": 5, "test_acc": 0.8581, "test_loss": 0.1791511276245117, "train_loss": 0.005986477714031935}, {"epoch_end": 6, "test_acc": 0.8597, "test_loss": 0.17956334228515625, "train_loss": 0.0027445852756500244}, {"epoch_end": 7, "test_acc": 0.8627, "test_loss": 0.1697449577331543, "train_loss": 0.0003132715355604887}, {"epoch_end": 8, "test_acc": 0.8523, "test_loss": 0.18929722747802735, "train_loss": 0.00027113239048048854}, {"epoch_end": 9, "test_acc": 0.8606, "test_loss": 0.1723261215209961, "train_loss": 0.00026587629690766335}, {"epoch_end": 10, "test_acc": 0.8561, "test_loss": 0.18276385192871095, "train_loss": 7.586369611090049e-05}, {"epoch_end": 11, "test_acc": 0.8656, "test_loss": 0.16506804809570313, "train_loss": 0.0008843948598951101}, {"epoch_end": 12, "test_acc": 0.8669, "test_loss": 0.16278900756835937, "train_loss": 0.005353138316422701}, {"epoch_end": 13, "test_acc": 0.8608, "test_loss": 0.17640072021484374, "train_loss": 0.00022469356190413237}, {"epoch_end": 14, "test_acc": 0.8686, "test_loss": 0.16051104354858398, "train_loss": 1.511078153271228e-06}, {"epoch_end": 15, "test_acc": 0.8701, "test_loss": 0.15654135131835936, "train_loss": 0.0006371229537762702}, {"epoch_end": 16, "test_acc": 0.8713, "test_loss": 0.15731585083007812, "train_loss": 0.0005890914471819997}, {"epoch_end": 17, "test_acc": 0.8648, "test_loss": 0.16680169219970703, "train_loss": 0.002603990025818348}, {"epoch_end": 18, "test_acc": 0.8757, "test_loss": 0.1467596176147461, "train_loss": 0.0006131798145361245}, {"epoch_end": 19, "test_acc": 0.8727, "test_loss": 0.1536305648803711, "train_loss": 0.0016642293194308877}, {"epoch_end": 20, "test_acc": 0.8686, "test_loss": 0.1609891098022461, "train_loss": 0.0}, {"epoch_end": 21, "test_acc": 0.8696, "test_loss": 0.1562847068786621, "train_loss": 0.0}, {"epoch_end": 22, "test_acc": 0.8759, "test_loss": 0.14773306274414064, "train_loss": 0.0}, {"epoch_end": 23, "test_acc": 0.8773, "test_loss": 0.14494658279418945, "train_loss": 0.0}, {"epoch_end": 24, "test_acc": 0.8717, "test_loss": 0.1594344612121582, "train_loss": 0.00014053747872821987}, {"epoch_end": 25, "test_acc": 0.8762, "test_loss": 0.1459928024291992, "train_loss": 0.00014689643285237253}, {"epoch_end": 26, "test_acc": 0.8768, "test_loss": 0.14496204757690428, "train_loss": 0.004999999888241291}, {"epoch_end": 27, "test_acc": 0.8711, "test_loss": 0.15727540740966797, "train_loss": 0.0}, {"epoch_end": 28, "test_acc": 0.8769, "test_loss": 0.1434103614807129, "train_loss": 0.0}, {"epoch_end": 29, "test_acc": 0.8782, "test_loss": 0.1433744140625, "train_loss": 0.00025980689679272473}, {"epoch_end": 30, "test_acc": 0.8773, "test_loss": 0.14672983474731446, "train_loss": 0.0}, {"epoch_end": 31, "test_acc": 0.8784, "test_loss": 0.14388548126220704, "train_loss": 0.0}, {"epoch_end": 32, "test_acc": 0.8789, "test_loss": 0.14196547012329103, "train_loss": 0.00023333965509664267}, {"epoch_end": 33, "test_acc": 0.8719, "test_loss": 0.15591726684570312, "train_loss": 0.00011699247261276469}, {"epoch_end": 34, "test_acc": 0.8686, "test_loss": 0.1618768783569336, "train_loss": 0.009616614319384098}, {"epoch_end": 35, "test_acc": 0.8742, "test_loss": 0.15000681533813476, "train_loss": 0.0}, {"epoch_end": 36, "test_acc": 0.8763, "test_loss": 0.14675813598632811, "train_loss": 3.7980221350153442e-06}, {"epoch_end": 37, "test_acc": 0.8791, "test_loss": 0.14377740783691406, "train_loss": 0.0}, {"epoch_end": 38, "test_acc": 0.8764, "test_loss": 0.14671160430908203, "train_loss": 3.909141923941206e-06}, {"epoch_end": 39, "test_acc": 0.8745, "test_loss": 0.15199198684692383, "train_loss": 0.0005784915410913527}, {"epoch_end": 40, "test_acc": 0.8797, "test_loss": 0.13992799911499024, "train_loss": 0.0}, {"epoch_end": 41, "test_acc": 0.8789, "test_loss": 0.14377679138183594, "train_loss": 0.0}, {"epoch_end": 42, "test_acc": 0.878, "test_loss": 0.14777060470581055, "train_loss": 0.0005716081359423697}, {"epoch_end": 43, "test_acc": 0.877, "test_loss": 0.14930353469848634, "train_loss": 0.0018796103540807962}, {"epoch_end": 44, "test_acc": 0.8818, "test_loss": 0.13803529663085937, "train_loss": 0.0}, {"epoch_end": 45, "test_acc": 0.8769, "test_loss": 0.1464111068725586, "train_loss": 3.340703187859617e-05}, {"epoch_end": 46, "test_acc": 0.8786, "test_loss": 0.14494974441528322, "train_loss": 0.0}, {"epoch_end": 47, "test_acc": 0.8709, "test_loss": 0.1602149101257324, "train_loss": 0.0}, {"epoch_end": 48, "test_acc": 0.8787, "test_loss": 0.14363153839111328, "train_loss": 0.0}, {"epoch_end": 49, "test_acc": 0.8766, "test_loss": 0.1477624755859375, "train_loss": 2.7098310511064483e-06}, {"epoch_end": 50, "test_acc": 0.8808, "test_loss": 0.14025485076904298, "train_loss": 0.0}, {"epoch_end": 51, "test_acc": 0.8761, "test_loss": 0.15056638259887695, "train_loss": 0.0}, {"epoch_end": 52, "test_acc": 0.8751, "test_loss": 0.1535306442260742, "train_loss": 0.0}, {"epoch_end": 53, "test_acc": 0.8785, "test_loss": 0.14508495178222655, "train_loss": 0.0}, {"epoch_end": 54, "test_acc": 0.8737, "test_loss": 0.153459912109375, "train_loss": 0.0}, {"epoch_end": 55, "test_acc": 0.8765, "test_loss": 0.1486385353088379, "train_loss": 9.688658610684797e-05}, {"epoch_end": 56, "test_acc": 0.8763, "test_loss": 0.15067028503417967, "train_loss": 0.0}, {"epoch_end": 57, "test_acc": 0.8808, "test_loss": 0.1403446090698242, "train_loss": 0.0}, {"epoch_end": 58, "test_acc": 0.8822, "test_loss": 0.13764392013549806, "train_loss": 0.0}, {"epoch_end": 59, "test_acc": 0.8777, "test_loss": 0.14470885620117188, "train_loss": 0.0}], "final": {"test_acc": 0.8777, "test_loss": 0.14470885620117188, "wall_s": 17532.79643392563}}
|
