1
|
{"args": {"mode": "dfa", "dataset": "cifar10", "depth": 20, "width": 64, "act": "tanh", "residual": 1, "epochs": 5, "batch_size": 128, "train_examples": 0, "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 16, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 2, "device": "cuda", "log_every": 500, "max_steps": 0, "probe_bs": 256, "outdir": "results", "tag": "scale_v3_cifar10_dfa_w64_d20_s2", "n_in": 3072, "n_out": 10}, "provenance": {"git_commit": "d4aef1a571c21e6e4e4125c90aa58da7f14e62e3", "git_dirty": false}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.6389660835266113, "cos_r_negg": [0.07448157668113708, 0.0014600008726119995, -0.013762697577476501, 0.05444254353642464, 0.022922558709979057, -0.012131884694099426, -0.05589545890688896, 0.03273990377783775, 0.0222703218460083, 0.0850955918431282, 0.06275729835033417, -0.02497122623026371, -0.08311794698238373, 0.029074417427182198, 0.007704804185777903, 0.045739881694316864, 0.02155613899230957, -0.03865803778171539, 0.000761175062507391, -0.017601067200303078], "cos_innovation_negg": [0.07448157668113708, 0.0014600008726119995, -0.013762697577476501, 0.05444254353642464, 0.022922558709979057, -0.012131884694099426, -0.05589545890688896, 0.03273990377783775, 0.0222703218460083, 0.0850955918431282, 0.06275729835033417, -0.02497122623026371, -0.08311794698238373, 0.029074417427182198, 0.007704804185777903, 0.045739881694316864, 0.02155613899230957, -0.03865803778171539, 0.000761175062507391, -0.017601067200303078], "cos_apical_negg": [0.07448157668113708, 0.0014600008726119995, -0.013762697577476501, 0.05444254353642464, 0.022922558709979057, -0.012131884694099426, -0.05589545890688896, 0.03273990377783775, 0.0222703218460083, 0.0850955918431282, 0.06275729835033417, -0.02497122623026371, -0.08311794698238373, 0.029074417427182198, 0.007704804185777903, 0.045739881694316864, 0.02155613899230957, -0.03865803778171539, 0.000761175062507391, -0.017601067200303078], "cos_Ac_negg": [0.07448157668113708, 0.0014600008726119995, -0.013762697577476501, 0.05444254353642464, 0.022922558709979057, -0.012131884694099426, -0.05589545890688896, 0.03273990377783775, 0.0222703218460083, 0.0850955918431282, 0.06275729835033417, -0.02497122623026371, -0.08311794698238373, 0.029074417427182198, 0.007704804185777903, 0.045739881694316864, 0.02155613899230957, -0.03865803778171539, 0.000761175062507391, -0.017601067200303078], "r_norm": [2.4913158416748047, 2.4768152236938477, 2.57504940032959, 2.43980073928833, 2.4843506813049316, 2.402621030807495, 2.4488186836242676, 2.4657020568847656, 2.4578514099121094, 2.4779200553894043, 2.4961323738098145, 2.6098742485046387, 2.3924381732940674, 2.673389196395874, 2.4331655502319336, 2.397357940673828, 2.687264919281006, 2.5355918407440186, 2.4266104698181152, 2.3182177543640137]}, {"epoch_end": 0, "step": 391, "test_acc": 0.3355, "test_loss": 1.881385595703125}, {"step": 500, "epoch": 1, "train_loss": 1.7851390838623047, "cos_r_negg": [0.4640771150588989, 0.0030161715112626553, 0.014575157314538956, -0.018322177231311798, 0.01954386942088604, -0.0065841348841786385, 0.02585095912218094, -0.09374186396598816, 0.09363935887813568, 0.010640671476721764, -0.03224765509366989, -0.0026568558532744646, 0.030340882018208504, 0.00013389438390731812, 0.0073851775377988815, -0.04146915674209595, -0.059956565499305725, -0.052099596709012985, -0.0031441031023859978, 0.03645823150873184], "cos_innovation_negg": [0.4640771150588989, 0.0030161715112626553, 0.014575157314538956, -0.018322177231311798, 0.01954386942088604, -0.0065841348841786385, 0.02585095912218094, -0.09374186396598816, 0.09363935887813568, 0.010640671476721764, -0.03224765509366989, -0.0026568558532744646, 0.030340882018208504, 0.00013389438390731812, 0.0073851775377988815, -0.04146915674209595, -0.059956565499305725, -0.052099596709012985, -0.0031441031023859978, 0.03645823150873184], "cos_apical_negg": [0.4640771150588989, 0.0030161715112626553, 0.014575157314538956, -0.018322177231311798, 0.01954386942088604, -0.0065841348841786385, 0.02585095912218094, -0.09374186396598816, 0.09363935887813568, 0.010640671476721764, -0.03224765509366989, -0.0026568558532744646, 0.030340882018208504, 0.00013389438390731812, 0.0073851775377988815, -0.04146915674209595, -0.059956565499305725, -0.052099596709012985, -0.0031441031023859978, 0.03645823150873184], "cos_Ac_negg": [0.4640771150588989, 0.0030161715112626553, 0.014575157314538956, -0.018322177231311798, 0.01954386942088604, -0.0065841348841786385, 0.02585095912218094, -0.09374186396598816, 0.09363935887813568, 0.010640671476721764, -0.03224765509366989, -0.0026568558532744646, 0.030340882018208504, 0.00013389438390731812, 0.0073851775377988815, -0.04146915674209595, -0.059956565499305725, -0.052099596709012985, -0.0031441031023859978, 0.03645823150873184], "r_norm": [2.2125487327575684, 2.19014310836792, 2.284943103790283, 2.101881980895996, 2.2105798721313477, 2.1341071128845215, 2.1729016304016113, 2.1816134452819824, 2.1557774543762207, 2.2032582759857178, 2.2203023433685303, 2.2814154624938965, 2.091397762298584, 2.331033706665039, 2.1622118949890137, 2.1681461334228516, 2.3640477657318115, 2.218207359313965, 2.142991065979004, 2.04206919670105]}, {"epoch_end": 1, "step": 782, "test_acc": 0.3411, "test_loss": 1.874661279296875}, {"step": 1000, "epoch": 2, "train_loss": 1.85595703125, "cos_r_negg": [0.558573842048645, 0.036978598684072495, 0.009148790501058102, 0.013155395165085793, 0.013183376751840115, 0.012356553226709366, -0.01104075089097023, -0.05993099510669708, 0.09335718303918839, 0.013943319208920002, 0.004368153400719166, 0.08683671057224274, 0.001207069493830204, -0.03124265745282173, 0.07321980595588684, 0.02006414160132408, -0.020863570272922516, -0.02153320610523224, -0.04226310923695564, 0.03823466598987579], "cos_innovation_negg": [0.558573842048645, 0.036978598684072495, 0.009148790501058102, 0.013155395165085793, 0.013183376751840115, 0.012356553226709366, -0.01104075089097023, -0.05993099510669708, 0.09335718303918839, 0.013943319208920002, 0.004368153400719166, 0.08683671057224274, 0.001207069493830204, -0.03124265745282173, 0.07321980595588684, 0.02006414160132408, -0.020863570272922516, -0.02153320610523224, -0.04226310923695564, 0.03823466598987579], "cos_apical_negg": [0.558573842048645, 0.036978598684072495, 0.009148790501058102, 0.013155395165085793, 0.013183376751840115, 0.012356553226709366, -0.01104075089097023, -0.05993099510669708, 0.09335718303918839, 0.013943319208920002, 0.004368153400719166, 0.08683671057224274, 0.001207069493830204, -0.03124265745282173, 0.07321980595588684, 0.02006414160132408, -0.020863570272922516, -0.02153320610523224, -0.04226310923695564, 0.03823466598987579], "cos_Ac_negg": [0.558573842048645, 0.036978598684072495, 0.009148790501058102, 0.013155395165085793, 0.013183376751840115, 0.012356553226709366, -0.01104075089097023, -0.05993099510669708, 0.09335718303918839, 0.013943319208920002, 0.004368153400719166, 0.08683671057224274, 0.001207069493830204, -0.03124265745282173, 0.07321980595588684, 0.02006414160132408, -0.020863570272922516, -0.02153320610523224, -0.04226310923695564, 0.03823466598987579], "r_norm": [2.198134660720825, 2.172987937927246, 2.275698661804199, 2.0858616828918457, 2.206936836242676, 2.114949941635132, 2.1676676273345947, 2.167811393737793, 2.1483511924743652, 2.195636749267578, 2.196794033050537, 2.2675693035125732, 2.0893359184265137, 2.3223891258239746, 2.1435770988464355, 2.1613082885742188, 2.3439152240753174, 2.209167003631592, 2.1303019523620605, 2.0290513038635254]}, {"epoch_end": 2, "step": 1173, "test_acc": 0.3491, "test_loss": 1.86499990234375}, {"step": 1500, "epoch": 3, "train_loss": 1.7456567287445068, "cos_r_negg": [0.5448051691055298, 0.024284038692712784, 0.024641944095492363, -0.01412207167595625, 0.08011756837368011, 0.01602393016219139, 0.021633237600326538, 0.013920004479587078, 0.10752232372760773, 0.01743071898818016, -0.01061627920717001, 0.05441971495747566, 0.004082793835550547, 0.06802317500114441, 0.03371230512857437, 0.015438380651175976, 0.055324114859104156, 0.0017619188874959946, 0.07272281497716904, 0.11185211688280106], "cos_innovation_negg": [0.5448051691055298, 0.024284038692712784, 0.024641944095492363, -0.01412207167595625, 0.08011756837368011, 0.01602393016219139, 0.021633237600326538, 0.013920004479587078, 0.10752232372760773, 0.01743071898818016, -0.01061627920717001, 0.05441971495747566, 0.004082793835550547, 0.06802317500114441, 0.03371230512857437, 0.015438380651175976, 0.055324114859104156, 0.0017619188874959946, 0.07272281497716904, 0.11185211688280106], "cos_apical_negg": [0.5448051691055298, 0.024284038692712784, 0.024641944095492363, -0.01412207167595625, 0.08011756837368011, 0.01602393016219139, 0.021633237600326538, 0.013920004479587078, 0.10752232372760773, 0.01743071898818016, -0.01061627920717001, 0.05441971495747566, 0.004082793835550547, 0.06802317500114441, 0.03371230512857437, 0.015438380651175976, 0.055324114859104156, 0.0017619188874959946, 0.07272281497716904, 0.11185211688280106], "cos_Ac_negg": [0.5448051691055298, 0.024284038692712784, 0.024641944095492363, -0.01412207167595625, 0.08011756837368011, 0.01602393016219139, 0.021633237600326538, 0.013920004479587078, 0.10752232372760773, 0.01743071898818016, -0.01061627920717001, 0.05441971495747566, 0.004082793835550547, 0.06802317500114441, 0.03371230512857437, 0.015438380651175976, 0.055324114859104156, 0.0017619188874959946, 0.07272281497716904, 0.11185211688280106], "r_norm": [2.1963696479797363, 2.174484968185425, 2.2724430561065674, 2.0813329219818115, 2.2107338905334473, 2.132361888885498, 2.1762070655822754, 2.1791486740112305, 2.143925666809082, 2.192976474761963, 2.2072956562042236, 2.257190704345703, 2.086545944213867, 2.3229029178619385, 2.1514406204223633, 2.1592295169830322, 2.3501033782958984, 2.2090721130371094, 2.1251113414764404, 2.039553642272949]}, {"epoch_end": 3, "step": 1564, "test_acc": 0.3531, "test_loss": 1.8557219116210937}, {"epoch_end": 4, "step": 1955, "test_acc": 0.3641, "test_loss": 1.8239850830078126}], "final": {"test_acc": 0.3641, "test_loss": 1.8239850830078126, "wall_s": 23.91855549812317, "cos_r_negg": [0.552725076675415, 0.07765524089336395, 0.04594659060239792, 0.07279789447784424, 0.09308507293462753, -0.01618456467986107, 0.05110277980566025, 0.023583944886922836, 0.06028150022029877, 0.025180421769618988, 0.03129574656486511, 0.07333260774612427, 0.06705927848815918, 0.04275612160563469, 0.07446826994419098, 0.04325999319553375, 0.12488226592540741, 0.04010283201932907, 0.011573133058845997, 0.08657374978065491], "cos_innovation_negg": [0.552725076675415, 0.07765524089336395, 0.04594659060239792, 0.07279789447784424, 0.09308507293462753, -0.01618456467986107, 0.05110277980566025, 0.023583944886922836, 0.06028150022029877, 0.025180421769618988, 0.03129574656486511, 0.07333260774612427, 0.06705927848815918, 0.04275612160563469, 0.07446826994419098, 0.04325999319553375, 0.12488226592540741, 0.04010283201932907, 0.011573133058845997, 0.08657374978065491], "cos_apical_negg": [0.552725076675415, 0.07765524089336395, 0.04594659060239792, 0.07279789447784424, 0.09308507293462753, -0.01618456467986107, 0.05110277980566025, 0.023583944886922836, 0.06028150022029877, 0.025180421769618988, 0.03129574656486511, 0.07333260774612427, 0.06705927848815918, 0.04275612160563469, 0.07446826994419098, 0.04325999319553375, 0.12488226592540741, 0.04010283201932907, 0.011573133058845997, 0.08657374978065491], "cos_Ac_negg": [0.552725076675415, 0.07765524089336395, 0.04594659060239792, 0.07279789447784424, 0.09308507293462753, -0.01618456467986107, 0.05110277980566025, 0.023583944886922836, 0.06028150022029877, 0.025180421769618988, 0.03129574656486511, 0.07333260774612427, 0.06705927848815918, 0.04275612160563469, 0.07446826994419098, 0.04325999319553375, 0.12488226592540741, 0.04010283201932907, 0.011573133058845997, 0.08657374978065491]}}
|
