1
|
{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 0, "epochs": 15, "batch_size": 128, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 16, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.2, "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 1, "device": "cuda", "log_every": 100, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "nuis_ctrl_v1_rho0p2_raw_s1", "n_in": 784}, "provenance": {"git_commit": "c8a372cae2612839bdbdafcf9c114f52cb4a6de3", "git_dirty": false}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.327335834503174, "cos_r_negg": [0.017816580832004547, -0.012290592305362225, -0.008806847967207432], "cos_innovation_negg": [0.021385151892900467, -0.013261372223496437, -0.007389807142317295], "cos_apical_negg": [0.017816580832004547, -0.012290592305362225, -0.008806847967207432], "cos_Ac_negg": [0.021384619176387787, -0.013261749409139156, -0.007390002720057964], "r_norm": [5.40655517578125, 5.13895320892334, 5.029324531555176], "ldr": {"dL_sdil": -0.023440837860107422, "dL_bp": -0.19552373886108398, "ratio": 0.1198874264406416}}, {"step": 100, "epoch": 0, "train_loss": 0.4356670379638672, "cos_r_negg": [0.1689734160900116, 0.26602211594581604, 0.615007758140564], "cos_innovation_negg": [0.39415937662124634, 0.405183345079422, 0.8060305714607239], "cos_apical_negg": [0.1689734160900116, 0.26602211594581604, 0.615007758140564], "cos_Ac_negg": [0.39402318000793457, 0.40512290596961975, 0.8059067726135254], "r_norm": [3.8986620903015137, 3.890565872192383, 3.8833794593811035]}, {"step": 200, "epoch": 0, "train_loss": 0.38534340262413025, "cos_r_negg": [0.17627941071987152, 0.27030643820762634, 0.6998161673545837], "cos_innovation_negg": [0.4324456453323364, 0.4655967354774475, 0.8484436273574829], "cos_apical_negg": [0.17627941071987152, 0.27030643820762634, 0.6998161673545837], "cos_Ac_negg": [0.4322904348373413, 0.4656267762184143, 0.8483703136444092], "r_norm": [3.7278993129730225, 3.7348222732543945, 3.731517791748047]}, {"step": 300, "epoch": 0, "train_loss": 0.43883538246154785, "cos_r_negg": [0.1197032704949379, 0.2570406198501587, 0.5781093835830688], "cos_innovation_negg": [0.4359956383705139, 0.4337473511695862, 0.7447086572647095], "cos_apical_negg": [0.1197032704949379, 0.2570406198501587, 0.5781093835830688], "cos_Ac_negg": [0.4358632564544678, 0.433771014213562, 0.7446643114089966], "r_norm": [3.6789705753326416, 3.7146718502044678, 3.700066566467285]}, {"step": 400, "epoch": 0, "train_loss": 0.27429088950157166, "cos_r_negg": [0.12952469289302826, 0.2163768857717514, 0.5665967464447021], "cos_innovation_negg": [0.4441429376602173, 0.5018871426582336, 0.7688440680503845], "cos_apical_negg": [0.12952469289302826, 0.2163768857717514, 0.5665967464447021], "cos_Ac_negg": [0.4440825581550598, 0.5019638538360596, 0.7687709927558899], "r_norm": [3.6598286628723145, 3.6770894527435303, 3.6506810188293457]}, {"epoch_end": 0, "step": 469, "test_acc": 0.918, "test_loss": 0.32542384796142576}, {"step": 500, "epoch": 1, "train_loss": 0.39269447326660156, "cos_r_negg": [0.1062721312046051, 0.20505928993225098, 0.5644583106040955], "cos_innovation_negg": [0.42568862438201904, 0.4226480722427368, 0.7683092355728149], "cos_apical_negg": [0.1062721312046051, 0.20505928993225098, 0.5644583106040955], "cos_Ac_negg": [0.4255436062812805, 0.4224078059196472, 0.7680299282073975], "r_norm": [3.608336925506592, 3.623990535736084, 3.614894151687622], "ldr": {"dL_sdil": -0.013470113277435303, "dL_bp": -0.013411611318588257, "ratio": 1.0043620380472824}}, {"step": 600, "epoch": 1, "train_loss": 0.2612594962120056, "cos_r_negg": [0.10076740384101868, 0.20813807845115662, 0.5794553756713867], "cos_innovation_negg": [0.44401049613952637, 0.4399864077568054, 0.7417570352554321], "cos_apical_negg": [0.10076740384101868, 0.20813807845115662, 0.5794553756713867], "cos_Ac_negg": [0.4437589943408966, 0.4399217963218689, 0.7415154576301575], "r_norm": [3.6123552322387695, 3.6415977478027344, 3.628936290740967]}, {"step": 700, "epoch": 1, "train_loss": 0.2748524844646454, "cos_r_negg": [0.09547808766365051, 0.193529412150383, 0.6413819789886475], "cos_innovation_negg": [0.446560263633728, 0.577634334564209, 0.74764084815979], "cos_apical_negg": [0.09547808766365051, 0.193529412150383, 0.6413819789886475], "cos_Ac_negg": [0.4464041590690613, 0.5775347352027893, 0.7478909492492676], "r_norm": [3.6176598072052, 3.622282028198242, 3.6285762786865234]}, {"step": 800, "epoch": 1, "train_loss": 0.17241355776786804, "cos_r_negg": [0.10088097304105759, 0.27291175723075867, 0.5625531077384949], "cos_innovation_negg": [0.431984007358551, 0.5703335404396057, 0.6875219345092773], "cos_apical_negg": [0.10088097304105759, 0.27291175723075867, 0.5625531077384949], "cos_Ac_negg": [0.4318956136703491, 0.5702531337738037, 0.687503457069397], "r_norm": [3.567176103591919, 3.582148313522339, 3.5719332695007324]}, {"step": 900, "epoch": 1, "train_loss": 0.31387853622436523, "cos_r_negg": [0.09953892230987549, 0.2921331226825714, 0.5925216674804688], "cos_innovation_negg": [0.44864189624786377, 0.6091399192810059, 0.822791337966919], "cos_apical_negg": [0.09953892230987549, 0.2921331226825714, 0.5925216674804688], "cos_Ac_negg": [0.4485782980918884, 0.6092501282691956, 0.8225663304328918], "r_norm": [3.5736279487609863, 3.5842716693878174, 3.576432228088379]}, {"epoch_end": 1, "step": 938, "test_acc": 0.9375, "test_loss": 0.25177192306518553}, {"step": 1000, "epoch": 2, "train_loss": 0.17049872875213623, "cos_r_negg": [0.10577884316444397, 0.26538562774658203, 0.5601075291633606], "cos_innovation_negg": [0.4354759454727173, 0.5381630063056946, 0.8366440534591675], "cos_apical_negg": [0.10577884316444397, 0.26538562774658203, 0.5601075291633606], "cos_Ac_negg": [0.43545007705688477, 0.5382001399993896, 0.8364940285682678], "r_norm": [3.5971617698669434, 3.6115221977233887, 3.5773520469665527], "ldr": {"dL_sdil": -0.02288752794265747, "dL_bp": -0.023692429065704346, "ratio": 0.9660270746906234}}, {"step": 1100, "epoch": 2, "train_loss": 0.3073243796825409, "cos_r_negg": [0.06935171037912369, 0.23983794450759888, 0.49295639991760254], "cos_innovation_negg": [0.4274466633796692, 0.5495652556419373, 0.724226713180542], "cos_apical_negg": [0.06935171037912369, 0.23983794450759888, 0.49295639991760254], "cos_Ac_negg": [0.42741042375564575, 0.5495082139968872, 0.7241969108581543], "r_norm": [3.55491304397583, 3.565094470977783, 3.5558106899261475]}, {"step": 1200, "epoch": 2, "train_loss": 0.3340785503387451, "cos_r_negg": [0.08844020962715149, 0.26136013865470886, 0.6662324666976929], "cos_innovation_negg": [0.4919504225254059, 0.564724326133728, 0.7682862877845764], "cos_apical_negg": [0.08844020962715149, 0.26136013865470886, 0.6662324666976929], "cos_Ac_negg": [0.49199527502059937, 0.5644968748092651, 0.7676172852516174], "r_norm": [3.5524349212646484, 3.563265562057495, 3.5444555282592773]}, {"step": 1300, "epoch": 2, "train_loss": 0.2755480408668518, "cos_r_negg": [0.0826110914349556, 0.2796480059623718, 0.7072727084159851], "cos_innovation_negg": [0.49278944730758667, 0.6161142587661743, 0.791374146938324], "cos_apical_negg": [0.0826110914349556, 0.2796480059623718, 0.7072727084159851], "cos_Ac_negg": [0.49275916814804077, 0.6161353588104248, 0.791243314743042], "r_norm": [3.5689969062805176, 3.5792593955993652, 3.56032133102417]}, {"step": 1400, "epoch": 2, "train_loss": 0.2892515957355499, "cos_r_negg": [0.04617100954055786, 0.19800525903701782, 0.5841293334960938], "cos_innovation_negg": [0.4735809564590454, 0.542998194694519, 0.8083404302597046], "cos_apical_negg": [0.04617100954055786, 0.19800525903701782, 0.5841293334960938], "cos_Ac_negg": [0.4735938310623169, 0.5430171489715576, 0.8083703517913818], "r_norm": [3.529519557952881, 3.53776478767395, 3.5263075828552246]}, {"epoch_end": 2, "step": 1407, "test_acc": 0.9436, "test_loss": 0.22161899795532228}, {"step": 1500, "epoch": 3, "train_loss": 0.18442478775978088, "cos_r_negg": [0.08789674937725067, 0.29980576038360596, 0.7146511077880859], "cos_innovation_negg": [0.4723778963088989, 0.605851411819458, 0.8120461702346802], "cos_apical_negg": [0.08789674937725067, 0.29980576038360596, 0.7146511077880859], "cos_Ac_negg": [0.47239476442337036, 0.6057537198066711, 0.8118860125541687], "r_norm": [3.5510218143463135, 3.5556280612945557, 3.5461888313293457], "ldr": {"dL_sdil": -0.00818738341331482, "dL_bp": -0.009622335433959961, "ratio": 0.8508727916945414}}, {"step": 1600, "epoch": 3, "train_loss": 0.21677272021770477, "cos_r_negg": [0.08460087329149246, 0.2268841415643692, 0.656872570514679], "cos_innovation_negg": [0.43371522426605225, 0.512417733669281, 0.8025482296943665], "cos_apical_negg": [0.08460087329149246, 0.2268841415643692, 0.656872570514679], "cos_Ac_negg": [0.43371880054473877, 0.512782633304596, 0.8022817969322205], "r_norm": [3.5290679931640625, 3.539044141769409, 3.5225727558135986]}, {"step": 1700, "epoch": 3, "train_loss": 0.11149293184280396, "cos_r_negg": [0.08747819066047668, 0.23266366124153137, 0.689258873462677], "cos_innovation_negg": [0.41761064529418945, 0.5147852301597595, 0.7864898443222046], "cos_apical_negg": [0.08747819066047668, 0.23266366124153137, 0.689258873462677], "cos_Ac_negg": [0.4176199436187744, 0.5148299932479858, 0.7862241268157959], "r_norm": [3.5327134132385254, 3.551211357116699, 3.532529830932617]}, {"step": 1800, "epoch": 3, "train_loss": 0.11369071900844574, "cos_r_negg": [0.08348599076271057, 0.24422600865364075, 0.6891183257102966], "cos_innovation_negg": [0.4450009763240814, 0.4909214675426483, 0.7027865648269653], "cos_apical_negg": [0.08348599076271057, 0.24422600865364075, 0.6891183257102966], "cos_Ac_negg": [0.4448184072971344, 0.4904190003871918, 0.7016391754150391], "r_norm": [3.567490577697754, 3.5742855072021484, 3.5611674785614014]}, {"epoch_end": 3, "step": 1876, "test_acc": 0.9467, "test_loss": 0.23838634185791016}, {"step": 1900, "epoch": 4, "train_loss": 0.11205631494522095, "cos_r_negg": [0.07415357977151871, 0.28101688623428345, 0.680112898349762], "cos_innovation_negg": [0.4501284956932068, 0.5675541162490845, 0.7829769849777222], "cos_apical_negg": [0.07415357977151871, 0.28101688623428345, 0.680112898349762], "cos_Ac_negg": [0.4500575661659241, 0.567550003528595, 0.7826475501060486], "r_norm": [3.5320348739624023, 3.548214912414551, 3.5277538299560547]}, {"step": 2000, "epoch": 4, "train_loss": 0.15781386196613312, "cos_r_negg": [0.047394976019859314, 0.24725601077079773, 0.7115209698677063], "cos_innovation_negg": [0.4403102993965149, 0.5349092483520508, 0.8005425930023193], "cos_apical_negg": [0.047394976019859314, 0.24725601077079773, 0.7115209698677063], "cos_Ac_negg": [0.44032394886016846, 0.5343272686004639, 0.800477147102356], "r_norm": [3.5300722122192383, 3.5361928939819336, 3.532763957977295], "ldr": {"dL_sdil": -0.01455448567867279, "dL_bp": -0.016517803072929382, "ratio": 0.8811393146178003}}, {"step": 2100, "epoch": 4, "train_loss": 0.1095762625336647, "cos_r_negg": [0.08439536392688751, 0.24140521883964539, 0.534023642539978], "cos_innovation_negg": [0.4613109529018402, 0.5413509607315063, 0.6667870283126831], "cos_apical_negg": [0.08439536392688751, 0.24140521883964539, 0.534023642539978], "cos_Ac_negg": [0.4611716866493225, 0.5415787696838379, 0.6719840168952942], "r_norm": [3.531738758087158, 3.5340096950531006, 3.5337023735046387]}, {"step": 2200, "epoch": 4, "train_loss": 0.15755702555179596, "cos_r_negg": [0.08772367238998413, 0.19960877299308777, 0.538733184337616], "cos_innovation_negg": [0.4514702260494232, 0.5397301912307739, 0.6183557510375977], "cos_apical_negg": [0.08772367238998413, 0.19960877299308777, 0.538733184337616], "cos_Ac_negg": [0.45148801803588867, 0.539559543132782, 0.6182036399841309], "r_norm": [3.506218433380127, 3.5101022720336914, 3.4992876052856445]}, {"step": 2300, "epoch": 4, "train_loss": 0.3984989821910858, "cos_r_negg": [0.06273116171360016, 0.19438037276268005, 0.5086185336112976], "cos_innovation_negg": [0.37528741359710693, 0.472161203622818, 0.7189158201217651], "cos_apical_negg": [0.06273116171360016, 0.19438037276268005, 0.5086185336112976], "cos_Ac_negg": [0.37531930208206177, 0.47167930006980896, 0.7185438275337219], "r_norm": [3.5103940963745117, 3.52298641204834, 3.505107879638672]}, {"epoch_end": 4, "step": 2345, "test_acc": 0.95, "test_loss": 0.20280362167358398}, {"step": 2400, "epoch": 5, "train_loss": 0.21440386772155762, "cos_r_negg": [0.07072244584560394, 0.23109771311283112, 0.7181301116943359], "cos_innovation_negg": [0.43955373764038086, 0.5710488557815552, 0.7419193983078003], "cos_apical_negg": [0.07072244584560394, 0.23109771311283112, 0.7181301116943359], "cos_Ac_negg": [0.4395294785499573, 0.5707806944847107, 0.7417935729026794], "r_norm": [3.5005040168762207, 3.5065855979919434, 3.4987382888793945]}, {"step": 2500, "epoch": 5, "train_loss": 0.12291234731674194, "cos_r_negg": [0.08408364653587341, 0.21224187314510345, 0.671499490737915], "cos_innovation_negg": [0.46145349740982056, 0.572177529335022, 0.8321336507797241], "cos_apical_negg": [0.08408364653587341, 0.21224187314510345, 0.671499490737915], "cos_Ac_negg": [0.46149587631225586, 0.5722059011459351, 0.8317503929138184], "r_norm": [3.502736806869507, 3.506289005279541, 3.49263858795166], "ldr": {"dL_sdil": -0.01088804006576538, "dL_bp": -0.012451842427253723, "ratio": 0.8744119699053049}}, {"step": 2600, "epoch": 5, "train_loss": 0.160813570022583, "cos_r_negg": [0.06952044367790222, 0.2068316638469696, 0.6928015947341919], "cos_innovation_negg": [0.46311214566230774, 0.5328207015991211, 0.7307355403900146], "cos_apical_negg": [0.06952044367790222, 0.2068316638469696, 0.6928015947341919], "cos_Ac_negg": [0.4631936550140381, 0.5325440168380737, 0.7305200099945068], "r_norm": [3.5053138732910156, 3.5166616439819336, 3.495210647583008]}, {"step": 2700, "epoch": 5, "train_loss": 0.1364130973815918, "cos_r_negg": [0.09724610298871994, 0.2638774514198303, 0.7222675085067749], "cos_innovation_negg": [0.44005417823791504, 0.5589592456817627, 0.819959282875061], "cos_apical_negg": [0.09724610298871994, 0.2638774514198303, 0.7222675085067749], "cos_Ac_negg": [0.4400399625301361, 0.5588173866271973, 0.8197481632232666], "r_norm": [3.5365207195281982, 3.5479297637939453, 3.520289421081543]}, {"step": 2800, "epoch": 5, "train_loss": 0.25221869349479675, "cos_r_negg": [0.06721948087215424, 0.23946689069271088, 0.6013596057891846], "cos_innovation_negg": [0.4509427547454834, 0.5987998843193054, 0.7860029935836792], "cos_apical_negg": [0.06721948087215424, 0.23946689069271088, 0.6013596057891846], "cos_Ac_negg": [0.4507969617843628, 0.5983481407165527, 0.7851682305335999], "r_norm": [3.478328227996826, 3.4902567863464355, 3.4767823219299316]}, {"epoch_end": 5, "step": 2814, "test_acc": 0.9475, "test_loss": 0.20962441024780273}, {"step": 2900, "epoch": 6, "train_loss": 0.07104501128196716, "cos_r_negg": [0.050086066126823425, 0.18649601936340332, 0.5987727642059326], "cos_innovation_negg": [0.43430572748184204, 0.5571109652519226, 0.7648565173149109], "cos_apical_negg": [0.050086066126823425, 0.18649601936340332, 0.5987727642059326], "cos_Ac_negg": [0.4344123601913452, 0.5572800040245056, 0.764534056186676], "r_norm": [3.4993033409118652, 3.507411003112793, 3.4867401123046875]}, {"step": 3000, "epoch": 6, "train_loss": 0.3882247805595398, "cos_r_negg": [0.035422470420598984, 0.2375493347644806, 0.7203845977783203], "cos_innovation_negg": [0.43740540742874146, 0.5923582315444946, 0.831477701663971], "cos_apical_negg": [0.035422470420598984, 0.2375493347644806, 0.7203845977783203], "cos_Ac_negg": [0.4374907910823822, 0.592270016670227, 0.8316681385040283], "r_norm": [3.508951187133789, 3.515498399734497, 3.506570816040039], "ldr": {"dL_sdil": -0.005574941635131836, "dL_bp": -0.009410247206687927, "ratio": 0.592433069257701}}, {"step": 3100, "epoch": 6, "train_loss": 0.2562331557273865, "cos_r_negg": [0.06972925364971161, 0.21012566983699799, 0.5793290138244629], "cos_innovation_negg": [0.4135417342185974, 0.536263644695282, 0.7021746635437012], "cos_apical_negg": [0.06972925364971161, 0.21012566983699799, 0.5793290138244629], "cos_Ac_negg": [0.41355645656585693, 0.53633713722229, 0.7020848393440247], "r_norm": [3.497947931289673, 3.5051608085632324, 3.4805431365966797]}, {"step": 3200, "epoch": 6, "train_loss": 0.24760746955871582, "cos_r_negg": [0.04600472003221512, 0.2612273097038269, 0.6718626618385315], "cos_innovation_negg": [0.44852331280708313, 0.5309064984321594, 0.7615406513214111], "cos_apical_negg": [0.04600472003221512, 0.2612273097038269, 0.6718626618385315], "cos_Ac_negg": [0.44859862327575684, 0.5307490229606628, 0.7611421346664429], "r_norm": [3.4771642684936523, 3.487788677215576, 3.4756035804748535]}, {"epoch_end": 6, "step": 3283, "test_acc": 0.9516, "test_loss": 0.20080115280151367}, {"step": 3300, "epoch": 7, "train_loss": 0.0777486264705658, "cos_r_negg": [0.058215558528900146, 0.2528151571750641, 0.5984994769096375], "cos_innovation_negg": [0.4828711152076721, 0.6769227981567383, 0.798736035823822], "cos_apical_negg": [0.058215558528900146, 0.2528151571750641, 0.5984994769096375], "cos_Ac_negg": [0.4828805923461914, 0.6768467426300049, 0.7987381219863892], "r_norm": [3.482384204864502, 3.4872336387634277, 3.475127935409546]}, {"step": 3400, "epoch": 7, "train_loss": 0.2704688012599945, "cos_r_negg": [0.10562082380056381, 0.24704620242118835, 0.5503202676773071], "cos_innovation_negg": [0.42678457498550415, 0.5736896991729736, 0.6482416391372681], "cos_apical_negg": [0.10562082380056381, 0.24704620242118835, 0.5503202676773071], "cos_Ac_negg": [0.42634135484695435, 0.5731626152992249, 0.6474989652633667], "r_norm": [3.504730224609375, 3.507523536682129, 3.487412452697754]}, {"step": 3500, "epoch": 7, "train_loss": 0.1841968595981598, "cos_r_negg": [0.03964517265558243, 0.2190065085887909, 0.2805502414703369], "cos_innovation_negg": [0.3974052369594574, 0.5362122058868408, 0.7107136845588684], "cos_apical_negg": [0.03964517265558243, 0.2190065085887909, 0.2805502414703369], "cos_Ac_negg": [0.3967805504798889, 0.5360937118530273, 0.7105822563171387], "r_norm": [3.4916341304779053, 3.5076818466186523, 3.4862499237060547], "ldr": {"dL_sdil": -0.010433733463287354, "dL_bp": -0.013596653938293457, "ratio": 0.7673750843876308}}, {"step": 3600, "epoch": 7, "train_loss": 0.08116768300533295, "cos_r_negg": [0.0007512769661843777, 0.1967654526233673, 0.5335896015167236], "cos_innovation_negg": [0.4380810260772705, 0.5319277048110962, 0.7935611605644226], "cos_apical_negg": [0.0007512769661843777, 0.1967654526233673, 0.5335896015167236], "cos_Ac_negg": [0.43796706199645996, 0.5299386978149414, 0.7933510541915894], "r_norm": [3.474682331085205, 3.4827327728271484, 3.467479705810547]}, {"step": 3700, "epoch": 7, "train_loss": 0.060786016285419464, "cos_r_negg": [0.033579256385564804, 0.18584689497947693, 0.5792797803878784], "cos_innovation_negg": [0.3946492671966553, 0.5032944679260254, 0.8709812164306641], "cos_apical_negg": [0.033579256385564804, 0.18584689497947693, 0.5792797803878784], "cos_Ac_negg": [0.3946901559829712, 0.5028430223464966, 0.8706419467926025], "r_norm": [3.454129695892334, 3.4617252349853516, 3.4539918899536133]}, {"epoch_end": 7, "step": 3752, "test_acc": 0.9535, "test_loss": 0.20100207901000977}, {"step": 3800, "epoch": 8, "train_loss": 0.14766934514045715, "cos_r_negg": [0.0455578938126564, 0.199472576379776, 0.5037887692451477], "cos_innovation_negg": [0.4330657422542572, 0.5209136605262756, 0.7766220569610596], "cos_apical_negg": [0.0455578938126564, 0.199472576379776, 0.5037887692451477], "cos_Ac_negg": [0.4327586889266968, 0.5200568437576294, 0.7765603065490723], "r_norm": [3.463716983795166, 3.475459098815918, 3.4624266624450684]}, {"step": 3900, "epoch": 8, "train_loss": 0.26989418268203735, "cos_r_negg": [0.04300380498170853, 0.19460409879684448, 0.6525254249572754], "cos_innovation_negg": [0.4327391982078552, 0.5639145374298096, 0.8385655879974365], "cos_apical_negg": [0.04300380498170853, 0.19460409879684448, 0.6525254249572754], "cos_Ac_negg": [0.4327658712863922, 0.5635520219802856, 0.8384553790092468], "r_norm": [3.4783992767333984, 3.483643054962158, 3.473872423171997]}, {"step": 4000, "epoch": 8, "train_loss": 0.1700727790594101, "cos_r_negg": [0.0691879391670227, 0.27972501516342163, 0.6195012927055359], "cos_innovation_negg": [0.4578794538974762, 0.6413708925247192, 0.7179861664772034], "cos_apical_negg": [0.0691879391670227, 0.27972501516342163, 0.6195012927055359], "cos_Ac_negg": [0.4578973352909088, 0.6412854194641113, 0.7177086472511292], "r_norm": [3.4794163703918457, 3.484488010406494, 3.467482089996338], "ldr": {"dL_sdil": -0.0059860944747924805, "dL_bp": -0.008245319128036499, "ratio": 0.7259991253022586}}, {"step": 4100, "epoch": 8, "train_loss": 0.07273297011852264, "cos_r_negg": [0.06468099355697632, 0.19027966260910034, 0.5791721343994141], "cos_innovation_negg": [0.4342232942581177, 0.5594127774238586, 0.808403491973877], "cos_apical_negg": [0.06468099355697632, 0.19027966260910034, 0.5791721343994141], "cos_Ac_negg": [0.4342646896839142, 0.5592778325080872, 0.8083246350288391], "r_norm": [3.4819517135620117, 3.480591297149658, 3.471940279006958]}, {"step": 4200, "epoch": 8, "train_loss": 0.3326941728591919, "cos_r_negg": [0.0850774273276329, 0.19953115284442902, 0.6112887859344482], "cos_innovation_negg": [0.41510963439941406, 0.48489782214164734, 0.7873479127883911], "cos_apical_negg": [0.0850774273276329, 0.19953115284442902, 0.6112887859344482], "cos_Ac_negg": [0.4148026406764984, 0.48095858097076416, 0.7850573062896729], "r_norm": [3.4779555797576904, 3.482177257537842, 3.463304042816162]}, {"epoch_end": 8, "step": 4221, "test_acc": 0.9522, "test_loss": 0.19817049865722655}, {"step": 4300, "epoch": 9, "train_loss": 0.0950564593076706, "cos_r_negg": [0.07509563863277435, 0.24522316455841064, 0.6603970527648926], "cos_innovation_negg": [0.46850359439849854, 0.5816950798034668, 0.7866798639297485], "cos_apical_negg": [0.07509563863277435, 0.24522316455841064, 0.6603970527648926], "cos_Ac_negg": [0.46845144033432007, 0.5801467895507812, 0.786354660987854], "r_norm": [3.4757020473480225, 3.477034568786621, 3.471956253051758]}, {"step": 4400, "epoch": 9, "train_loss": 0.03483837842941284, "cos_r_negg": [0.0756162479519844, 0.26046431064605713, 0.5944513082504272], "cos_innovation_negg": [0.492334246635437, 0.684106707572937, 0.7453880906105042], "cos_apical_negg": [0.0756162479519844, 0.26046431064605713, 0.5944513082504272], "cos_Ac_negg": [0.4923914670944214, 0.6840108633041382, 0.7452437877655029], "r_norm": [3.470336675643921, 3.4723639488220215, 3.462059497833252]}, {"step": 4500, "epoch": 9, "train_loss": 0.1985168159008026, "cos_r_negg": [0.05870479717850685, 0.21263810992240906, 0.4580349922180176], "cos_innovation_negg": [0.4854428470134735, 0.5557766556739807, 0.626973569393158], "cos_apical_negg": [0.05870479717850685, 0.21263810992240906, 0.4580349922180176], "cos_Ac_negg": [0.4855104386806488, 0.5557011365890503, 0.6268703937530518], "r_norm": [3.484640121459961, 3.4871811866760254, 3.4654431343078613], "ldr": {"dL_sdil": -0.007090747356414795, "dL_bp": -0.011536553502082825, "ratio": 0.6146330752190957}}, {"step": 4600, "epoch": 9, "train_loss": 0.0680193081498146, "cos_r_negg": [0.04870690777897835, 0.253722608089447, 0.510011613368988], "cos_innovation_negg": [0.4142921566963196, 0.5051026344299316, 0.743492066860199], "cos_apical_negg": [0.04870690777897835, 0.253722608089447, 0.510011613368988], "cos_Ac_negg": [0.41395288705825806, 0.5022687911987305, 0.7433861494064331], "r_norm": [3.467735767364502, 3.4734983444213867, 3.462430953979492]}, {"epoch_end": 9, "step": 4690, "test_acc": 0.9538, "test_loss": 0.19944113998413085}, {"step": 4700, "epoch": 10, "train_loss": 0.09776610136032104, "cos_r_negg": [0.05886831879615784, 0.22570858895778656, 0.5670293569564819], "cos_innovation_negg": [0.4429883658885956, 0.5386934876441956, 0.6798692941665649], "cos_apical_negg": [0.05886831879615784, 0.22570858895778656, 0.5670293569564819], "cos_Ac_negg": [0.44310957193374634, 0.5391794443130493, 0.6795532703399658], "r_norm": [3.4625110626220703, 3.470669984817505, 3.455833673477173]}, {"step": 4800, "epoch": 10, "train_loss": 0.08414127677679062, "cos_r_negg": [0.050551027059555054, 0.10287302732467651, 0.3448510766029358], "cos_innovation_negg": [0.39102932810783386, 0.46998846530914307, 0.6855603456497192], "cos_apical_negg": [0.050551027059555054, 0.10287302732467651, 0.3448510766029358], "cos_Ac_negg": [0.3910481333732605, 0.46947142481803894, 0.6893148422241211], "r_norm": [3.4664206504821777, 3.472391366958618, 3.464334011077881]}, {"step": 4900, "epoch": 10, "train_loss": 0.11371088773012161, "cos_r_negg": [0.04447276517748833, 0.1574743390083313, 0.5095987319946289], "cos_innovation_negg": [0.4299008548259735, 0.5574938058853149, 0.7616347074508667], "cos_apical_negg": [0.04447276517748833, 0.1574743390083313, 0.5095987319946289], "cos_Ac_negg": [0.42991816997528076, 0.5574855804443359, 0.761502742767334], "r_norm": [3.468412160873413, 3.4738636016845703, 3.460758924484253]}, {"step": 5000, "epoch": 10, "train_loss": 0.05606111139059067, "cos_r_negg": [0.037474289536476135, 0.18544331192970276, 0.6179916262626648], "cos_innovation_negg": [0.44923120737075806, 0.6171839237213135, 0.7551383972167969], "cos_apical_negg": [0.037474289536476135, 0.18544331192970276, 0.6179916262626648], "cos_Ac_negg": [0.44958367943763733, 0.6168597340583801, 0.7548431754112244], "r_norm": [3.4625275135040283, 3.4685163497924805, 3.4564919471740723], "ldr": {"dL_sdil": -0.004449918866157532, "dL_bp": -0.0074905455112457275, "ratio": 0.5940714010050091}}, {"step": 5100, "epoch": 10, "train_loss": 0.14297005534172058, "cos_r_negg": [0.06512479484081268, 0.23240914940834045, 0.6401159167289734], "cos_innovation_negg": [0.447102427482605, 0.6113333106040955, 0.7934198379516602], "cos_apical_negg": [0.06512479484081268, 0.23240914940834045, 0.6401159167289734], "cos_Ac_negg": [0.44726502895355225, 0.6106896996498108, 0.7930235862731934], "r_norm": [3.4628477096557617, 3.4697208404541016, 3.449944496154785]}, {"epoch_end": 10, "step": 5159, "test_acc": 0.9543, "test_loss": 0.20183923797607423}, {"step": 5200, "epoch": 11, "train_loss": 0.09912631660699844, "cos_r_negg": [0.07001079618930817, 0.2584608793258667, 0.5197109580039978], "cos_innovation_negg": [0.43941575288772583, 0.5701873898506165, 0.7516946792602539], "cos_apical_negg": [0.07001079618930817, 0.2584608793258667, 0.5197109580039978], "cos_Ac_negg": [0.439582496881485, 0.5694912075996399, 0.751593828201294], "r_norm": [3.462938070297241, 3.4726412296295166, 3.448319911956787]}, {"step": 5300, "epoch": 11, "train_loss": 0.11607149988412857, "cos_r_negg": [0.05213943123817444, 0.21287769079208374, 0.4091430604457855], "cos_innovation_negg": [0.39561790227890015, 0.5195497274398804, 0.8210018277168274], "cos_apical_negg": [0.05213943123817444, 0.21287769079208374, 0.4091430604457855], "cos_Ac_negg": [0.39566001296043396, 0.5188983678817749, 0.8210616111755371], "r_norm": [3.4646692276000977, 3.4711191654205322, 3.455239772796631]}, {"step": 5400, "epoch": 11, "train_loss": 0.26368260383605957, "cos_r_negg": [0.06291709095239639, 0.24170327186584473, 0.532926619052887], "cos_innovation_negg": [0.3816204369068146, 0.5461244583129883, 0.8120216131210327], "cos_apical_negg": [0.06291709095239639, 0.24170327186584473, 0.532926619052887], "cos_Ac_negg": [0.38147851824760437, 0.5410433411598206, 0.8119370937347412], "r_norm": [3.466569185256958, 3.4699583053588867, 3.457618474960327]}, {"step": 5500, "epoch": 11, "train_loss": 0.22858896851539612, "cos_r_negg": [0.06074442341923714, 0.18535567820072174, 0.45614051818847656], "cos_innovation_negg": [0.41328179836273193, 0.5018080472946167, 0.7735607624053955], "cos_apical_negg": [0.06074442341923714, 0.18535567820072174, 0.45614051818847656], "cos_Ac_negg": [0.4130654036998749, 0.5000206828117371, 0.7733755111694336], "r_norm": [3.454829216003418, 3.459933280944824, 3.4497036933898926], "ldr": {"dL_sdil": -0.005135253071784973, "dL_bp": -0.010742738842964172, "ratio": 0.478020842436121}}, {"step": 5600, "epoch": 11, "train_loss": 0.09115181863307953, "cos_r_negg": [0.04053889214992523, 0.2067248821258545, 0.6530458927154541], "cos_innovation_negg": [0.3989574611186981, 0.5432908535003662, 0.767601728439331], "cos_apical_negg": [0.04053889214992523, 0.2067248821258545, 0.6530458927154541], "cos_Ac_negg": [0.39899998903274536, 0.5434530973434448, 0.7665030360221863], "r_norm": [3.4796388149261475, 3.4847800731658936, 3.4624600410461426]}, {"epoch_end": 11, "step": 5628, "test_acc": 0.9549, "test_loss": 0.2047551902770996}, {"step": 5700, "epoch": 12, "train_loss": 0.08562565594911575, "cos_r_negg": [0.04938215762376785, 0.22137320041656494, 0.48624658584594727], "cos_innovation_negg": [0.43018898367881775, 0.5790407657623291, 0.7335515022277832], "cos_apical_negg": [0.04938215762376785, 0.22137320041656494, 0.48624658584594727], "cos_Ac_negg": [0.43125706911087036, 0.5718858242034912, 0.7333112359046936], "r_norm": [3.468977212905884, 3.4694833755493164, 3.4570069313049316]}, {"step": 5800, "epoch": 12, "train_loss": 0.01059839129447937, "cos_r_negg": [0.05219132453203201, 0.2245350182056427, 0.6449170112609863], "cos_innovation_negg": [0.44984960556030273, 0.5593054294586182, 0.7876733541488647], "cos_apical_negg": [0.05219132453203201, 0.2245350182056427, 0.6449170112609863], "cos_Ac_negg": [0.4500811696052551, 0.556477427482605, 0.7876458168029785], "r_norm": [3.4648661613464355, 3.4633777141571045, 3.452197313308716]}, {"step": 5900, "epoch": 12, "train_loss": 0.08707671612501144, "cos_r_negg": [0.06551908701658249, 0.24000033736228943, 0.576103150844574], "cos_innovation_negg": [0.4030078947544098, 0.6000269055366516, 0.8022713661193848], "cos_apical_negg": [0.06551908701658249, 0.24000033736228943, 0.576103150844574], "cos_Ac_negg": [0.40303221344947815, 0.5962719321250916, 0.8021576404571533], "r_norm": [3.4652485847473145, 3.4664688110351562, 3.453677177429199]}, {"step": 6000, "epoch": 12, "train_loss": 0.016570743173360825, "cos_r_negg": [0.04202159494161606, 0.19206121563911438, 0.4011518955230713], "cos_innovation_negg": [0.37041717767715454, 0.5084023475646973, 0.7606583833694458], "cos_apical_negg": [0.04202159494161606, 0.19206121563911438, 0.4011518955230713], "cos_Ac_negg": [0.37013882398605347, 0.5067340135574341, 0.761455774307251], "r_norm": [3.466170310974121, 3.4691123962402344, 3.4524691104888916], "ldr": {"dL_sdil": -0.006000012159347534, "dL_bp": -0.010488852858543396, "ratio": 0.5720370225672863}}, {"epoch_end": 12, "step": 6097, "test_acc": 0.9547, "test_loss": 0.1983235137939453}, {"step": 6100, "epoch": 13, "train_loss": 0.05860435217618942, "cos_r_negg": [0.057026926428079605, 0.19623446464538574, 0.4584203362464905], "cos_innovation_negg": [0.395458847284317, 0.5979858636856079, 0.7957639694213867], "cos_apical_negg": [0.057026926428079605, 0.19623446464538574, 0.4584203362464905], "cos_Ac_negg": [0.3953757584095001, 0.5976648926734924, 0.7956451773643494], "r_norm": [3.461276054382324, 3.4666571617126465, 3.4516501426696777]}, {"step": 6200, "epoch": 13, "train_loss": 0.11902886629104614, "cos_r_negg": [0.08081215620040894, 0.1969260573387146, 0.5059528350830078], "cos_innovation_negg": [0.41033506393432617, 0.534386157989502, 0.8059005737304688], "cos_apical_negg": [0.08081215620040894, 0.1969260573387146, 0.5059528350830078], "cos_Ac_negg": [0.41017386317253113, 0.5336558818817139, 0.8060274124145508], "r_norm": [3.4563701152801514, 3.4618825912475586, 3.4465322494506836]}, {"step": 6300, "epoch": 13, "train_loss": 0.07047811150550842, "cos_r_negg": [0.029118582606315613, 0.16016680002212524, 0.527755081653595], "cos_innovation_negg": [0.39475560188293457, 0.5635925531387329, 0.8127559423446655], "cos_apical_negg": [0.029118582606315613, 0.16016680002212524, 0.527755081653595], "cos_Ac_negg": [0.39480024576187134, 0.563109815120697, 0.812713086605072], "r_norm": [3.4493629932403564, 3.452237129211426, 3.4401230812072754]}, {"step": 6400, "epoch": 13, "train_loss": 0.09686963260173798, "cos_r_negg": [0.036843959242105484, 0.19907917082309723, 0.6437855958938599], "cos_innovation_negg": [0.3797476887702942, 0.5322127938270569, 0.7541957497596741], "cos_apical_negg": [0.036843959242105484, 0.19907917082309723, 0.6437855958938599], "cos_Ac_negg": [0.3802404999732971, 0.5302999019622803, 0.7538274526596069], "r_norm": [3.4386394023895264, 3.443452835083008, 3.4289498329162598]}, {"step": 6500, "epoch": 13, "train_loss": 0.030908331274986267, "cos_r_negg": [0.025387559086084366, 0.20244021713733673, 0.5220205187797546], "cos_innovation_negg": [0.41954177618026733, 0.4961056113243103, 0.7337448000907898], "cos_apical_negg": [0.025387559086084366, 0.20244021713733673, 0.5220205187797546], "cos_Ac_negg": [0.41956377029418945, 0.495990514755249, 0.733537495136261], "r_norm": [3.462789297103882, 3.4721198081970215, 3.445420742034912], "ldr": {"dL_sdil": -0.01279139518737793, "dL_bp": -0.019137457013130188, "ratio": 0.6683957632720883}}, {"epoch_end": 13, "step": 6566, "test_acc": 0.9547, "test_loss": 0.2109088150024414}, {"step": 6600, "epoch": 14, "train_loss": 0.05749582499265671, "cos_r_negg": [0.031158369034528732, 0.14849227666854858, 0.46749401092529297], "cos_innovation_negg": [0.3765043020248413, 0.5662267208099365, 0.732268750667572], "cos_apical_negg": [0.031158369034528732, 0.14849227666854858, 0.46749401092529297], "cos_Ac_negg": [0.3768375515937805, 0.565014123916626, 0.7313430309295654], "r_norm": [3.443437337875366, 3.4483394622802734, 3.433666944503784]}, {"step": 6700, "epoch": 14, "train_loss": 0.07075092196464539, "cos_r_negg": [0.035666707903146744, 0.13649296760559082, 0.527536153793335], "cos_innovation_negg": [0.37695565819740295, 0.5032856464385986, 0.7208537459373474], "cos_apical_negg": [0.035666707903146744, 0.13649296760559082, 0.527536153793335], "cos_Ac_negg": [0.37247520685195923, 0.4940648078918457, 0.7214541435241699], "r_norm": [3.4513092041015625, 3.458841562271118, 3.4443211555480957]}, {"step": 6800, "epoch": 14, "train_loss": 0.16527383029460907, "cos_r_negg": [0.0403471365571022, 0.1276024580001831, 0.5496867895126343], "cos_innovation_negg": [0.37198305130004883, 0.4789280891418457, 0.7920528054237366], "cos_apical_negg": [0.0403471365571022, 0.1276024580001831, 0.5496867895126343], "cos_Ac_negg": [0.37165147066116333, 0.47502797842025757, 0.7918992042541504], "r_norm": [3.4475739002227783, 3.4508376121520996, 3.435202121734619]}, {"step": 6900, "epoch": 14, "train_loss": 0.05497385933995247, "cos_r_negg": [0.06007204204797745, 0.19484585523605347, 0.5344549417495728], "cos_innovation_negg": [0.42855221033096313, 0.552445650100708, 0.8151719570159912], "cos_apical_negg": [0.06007204204797745, 0.19484585523605347, 0.5344549417495728], "cos_Ac_negg": [0.42856401205062866, 0.5520523190498352, 0.8149268627166748], "r_norm": [3.450343370437622, 3.454378604888916, 3.4388420581817627]}, {"step": 7000, "epoch": 14, "train_loss": 0.1180526539683342, "cos_r_negg": [0.021235257387161255, 0.14549796283245087, 0.41770482063293457], "cos_innovation_negg": [0.40436187386512756, 0.49030813574790955, 0.746916651725769], "cos_apical_negg": [0.021235257387161255, 0.14549796283245087, 0.41770482063293457], "cos_Ac_negg": [0.4050508439540863, 0.4935333728790283, 0.7477433085441589], "r_norm": [3.452204942703247, 3.4590353965759277, 3.4412460327148438], "ldr": {"dL_sdil": -0.0052800774574279785, "dL_bp": -0.01100870966911316, "ratio": 0.4796272784123057}}, {"epoch_end": 14, "step": 7035, "test_acc": 0.9539, "test_loss": 0.2062372871398926}], "final": {"test_acc": 0.9539, "test_loss": 0.2062372871398926, "wall_s": 22.702091693878174, "cos_r_negg": [9.238254278898239e-06, 0.1693142205476761, 0.5549061298370361], "cos_innovation_negg": [0.3668328523635864, 0.518203616142273, 0.7497802972793579], "cos_apical_negg": [9.238254278898239e-06, 0.1693142205476761, 0.5549061298370361], "cos_Ac_negg": [0.36657118797302246, 0.5130937099456787, 0.7536516189575195]}}
|