Add precomputed results for p=5
Browse files- .gitattributes +6 -0
- precomputed_results/p_005/p005_dft_heatmap_in.png +0 -0
- precomputed_results/p_005/p005_dft_heatmap_out.png +0 -0
- precomputed_results/p_005/p005_lineplot_in.png +3 -0
- precomputed_results/p_005/p005_lineplot_out.png +3 -0
- precomputed_results/p_005/p005_logits_interactive.json +1 -0
- precomputed_results/p_005/p005_lottery_beta_contour.png +0 -0
- precomputed_results/p_005/p005_lottery_mech_magnitude.png +0 -0
- precomputed_results/p_005/p005_lottery_mech_phase.png +0 -0
- precomputed_results/p_005/p005_magnitude_distribution.png +0 -0
- precomputed_results/p_005/p005_metadata.json +82 -0
- precomputed_results/p_005/p005_neuron_spectra.json +1 -0
- precomputed_results/p_005/p005_output_logits.png +0 -0
- precomputed_results/p_005/p005_overview.json +1 -0
- precomputed_results/p_005/p005_overview_loss_ipr.png +0 -0
- precomputed_results/p_005/p005_overview_phase_scatter.png +0 -0
- precomputed_results/p_005/p005_phase_align_approx1.png +3 -0
- precomputed_results/p_005/p005_phase_align_approx2.png +3 -0
- precomputed_results/p_005/p005_phase_align_quad.png +0 -0
- precomputed_results/p_005/p005_phase_align_relu.png +0 -0
- precomputed_results/p_005/p005_phase_distribution.png +0 -0
- precomputed_results/p_005/p005_phase_relationship.png +0 -0
- precomputed_results/p_005/p005_single_freq_quad.png +3 -0
- precomputed_results/p_005/p005_single_freq_relu.png +3 -0
- precomputed_results/p_005/p005_training_log.json +0 -0
.gitattributes
CHANGED
|
@@ -147,3 +147,9 @@ precomputed_results/p_007/p007_phase_align_approx1.png filter=lfs diff=lfs merge
|
|
| 147 |
precomputed_results/p_007/p007_phase_align_approx2.png filter=lfs diff=lfs merge=lfs -text
|
| 148 |
precomputed_results/p_007/p007_single_freq_quad.png filter=lfs diff=lfs merge=lfs -text
|
| 149 |
precomputed_results/p_007/p007_single_freq_relu.png filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 147 |
precomputed_results/p_007/p007_phase_align_approx2.png filter=lfs diff=lfs merge=lfs -text
|
| 148 |
precomputed_results/p_007/p007_single_freq_quad.png filter=lfs diff=lfs merge=lfs -text
|
| 149 |
precomputed_results/p_007/p007_single_freq_relu.png filter=lfs diff=lfs merge=lfs -text
|
| 150 |
+
precomputed_results/p_005/p005_lineplot_in.png filter=lfs diff=lfs merge=lfs -text
|
| 151 |
+
precomputed_results/p_005/p005_lineplot_out.png filter=lfs diff=lfs merge=lfs -text
|
| 152 |
+
precomputed_results/p_005/p005_phase_align_approx1.png filter=lfs diff=lfs merge=lfs -text
|
| 153 |
+
precomputed_results/p_005/p005_phase_align_approx2.png filter=lfs diff=lfs merge=lfs -text
|
| 154 |
+
precomputed_results/p_005/p005_single_freq_quad.png filter=lfs diff=lfs merge=lfs -text
|
| 155 |
+
precomputed_results/p_005/p005_single_freq_relu.png filter=lfs diff=lfs merge=lfs -text
|
precomputed_results/p_005/p005_dft_heatmap_in.png
ADDED
|
precomputed_results/p_005/p005_dft_heatmap_out.png
ADDED
|
precomputed_results/p_005/p005_lineplot_in.png
ADDED
|
Git LFS Details
|
precomputed_results/p_005/p005_lineplot_out.png
ADDED
|
Git LFS Details
|
precomputed_results/p_005/p005_logits_interactive.json
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"pairs": [[0, 0], [1, 0], [2, 0], [3, 0], [4, 0]], "correct_answers": [0, 1, 2, 3, 4], "logits": [[6.911810874938965, 0.1288711130619049, -0.09793221950531006, -0.2052120566368103, -0.3570995330810547], [-0.03658895194530487, 6.876556873321533, -0.132428377866745, -1.8679537773132324, -2.2972207069396973], [-0.70485520362854, -2.7452077865600586, 6.343863487243652, -2.4648513793945312, -0.7085124850273132], [-0.46954765915870667, -0.40340685844421387, -2.44149112701416, 6.718780040740967, -1.6097832918167114], [0.041380375623703, -2.54884672164917, -2.3647713661193848, -0.0392208993434906, 7.032649517059326]], "output_classes": [0, 1, 2, 3, 4]}
|
precomputed_results/p_005/p005_lottery_beta_contour.png
ADDED
|
precomputed_results/p_005/p005_lottery_mech_magnitude.png
ADDED
|
precomputed_results/p_005/p005_lottery_mech_phase.png
ADDED
|
precomputed_results/p_005/p005_magnitude_distribution.png
ADDED
|
precomputed_results/p_005/p005_metadata.json
ADDED
|
@@ -0,0 +1,82 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"prime": 5,
|
| 3 |
+
"d_mlp": 512,
|
| 4 |
+
"training_runs": {
|
| 5 |
+
"standard": {
|
| 6 |
+
"act_type": "ReLU",
|
| 7 |
+
"lr": 5e-05,
|
| 8 |
+
"weight_decay": 0,
|
| 9 |
+
"num_epochs": 5000,
|
| 10 |
+
"frac_train": 1.0,
|
| 11 |
+
"init_type": "random",
|
| 12 |
+
"init_scale": 0.1,
|
| 13 |
+
"optimizer": "AdamW"
|
| 14 |
+
},
|
| 15 |
+
"grokking": {
|
| 16 |
+
"act_type": "ReLU",
|
| 17 |
+
"lr": 0.0001,
|
| 18 |
+
"weight_decay": 2.0,
|
| 19 |
+
"num_epochs": 50000,
|
| 20 |
+
"frac_train": 0.75,
|
| 21 |
+
"init_type": "random",
|
| 22 |
+
"init_scale": 0.1,
|
| 23 |
+
"optimizer": "AdamW"
|
| 24 |
+
},
|
| 25 |
+
"quad_random": {
|
| 26 |
+
"act_type": "Quad",
|
| 27 |
+
"lr": 5e-05,
|
| 28 |
+
"weight_decay": 0,
|
| 29 |
+
"num_epochs": 5000,
|
| 30 |
+
"frac_train": 1.0,
|
| 31 |
+
"init_type": "random",
|
| 32 |
+
"init_scale": 0.1,
|
| 33 |
+
"optimizer": "AdamW"
|
| 34 |
+
},
|
| 35 |
+
"quad_single_freq": {
|
| 36 |
+
"act_type": "Quad",
|
| 37 |
+
"lr": 0.1,
|
| 38 |
+
"weight_decay": 0,
|
| 39 |
+
"num_epochs": 10000,
|
| 40 |
+
"frac_train": 1.0,
|
| 41 |
+
"init_type": "single-freq",
|
| 42 |
+
"init_scale": 0.02,
|
| 43 |
+
"optimizer": "SGD"
|
| 44 |
+
},
|
| 45 |
+
"relu_single_freq": {
|
| 46 |
+
"act_type": "ReLU",
|
| 47 |
+
"lr": 0.01,
|
| 48 |
+
"weight_decay": 0,
|
| 49 |
+
"num_epochs": 10000,
|
| 50 |
+
"frac_train": 1.0,
|
| 51 |
+
"init_type": "single-freq",
|
| 52 |
+
"init_scale": 0.002,
|
| 53 |
+
"optimizer": "SGD"
|
| 54 |
+
}
|
| 55 |
+
},
|
| 56 |
+
"final_metrics": {
|
| 57 |
+
"standard": {
|
| 58 |
+
"train_acc": 1.0,
|
| 59 |
+
"test_acc": 1.0,
|
| 60 |
+
"train_loss": 0.002307518618181348,
|
| 61 |
+
"test_loss": 0.002307518618181348
|
| 62 |
+
},
|
| 63 |
+
"quad_random": {
|
| 64 |
+
"train_acc": 1.0,
|
| 65 |
+
"test_acc": 1.0,
|
| 66 |
+
"train_loss": 0.00014215294504538178,
|
| 67 |
+
"test_loss": 0.00014215294504538178
|
| 68 |
+
},
|
| 69 |
+
"quad_single_freq": {
|
| 70 |
+
"train_acc": 1.0,
|
| 71 |
+
"test_acc": 1.0,
|
| 72 |
+
"train_loss": 0.00030729224090464413,
|
| 73 |
+
"test_loss": 0.00030729224090464413
|
| 74 |
+
},
|
| 75 |
+
"relu_single_freq": {
|
| 76 |
+
"train_acc": 1.0,
|
| 77 |
+
"test_acc": 1.0,
|
| 78 |
+
"train_loss": 0.07391073554754257,
|
| 79 |
+
"test_loss": 0.07391073554754257
|
| 80 |
+
}
|
| 81 |
+
}
|
| 82 |
+
}
|
precomputed_results/p_005/p005_neuron_spectra.json
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"fourier_basis_names": ["Const", "cos 1", "sin 1", "cos 2", "sin 2"], "neurons": {"neuron_0": {"global_index": 309, "dominant_freq": 1, "fourier_magnitudes_in": [0.08683527261018753, 0.15921242535114288, 0.5680559277534485, 0.20496675372123718, 0.08499649912118912], "fourier_magnitudes_out": [0.10559431463479996, 0.41936710476875305, 0.344473272562027, 0.07086890190839767, 0.13639706373214722]}, "neuron_1": {"global_index": 165, "dominant_freq": 2, "fourier_magnitudes_in": [0.0964902862906456, 0.12699627876281738, 0.10945871472358704, 0.1354667693376541, 0.5761796832084656], "fourier_magnitudes_out": [0.09691566228866577, 0.002606252208352089, 0.0963432565331459, 0.44277000427246094, 0.2752063572406769]}, "neuron_2": {"global_index": 103, "dominant_freq": 1, "fourier_magnitudes_in": [0.1169174537062645, 0.1633247286081314, 0.522530198097229, 0.15160667896270752, 0.05688190832734108], "fourier_magnitudes_out": [0.18352776765823364, 0.43118932843208313, 0.2747894823551178, 0.04444558918476105, 0.09664081037044525]}, "neuron_3": {"global_index": 100, "dominant_freq": 2, "fourier_magnitudes_in": [0.10001371800899506, 0.17589667439460754, 0.002035110956057906, 0.18880105018615723, 0.5120527148246765], "fourier_magnitudes_out": [0.08281271904706955, 0.07412151992321014, 0.20204409956932068, 0.407164990901947, 0.32999488711357117]}, "neuron_4": {"global_index": 274, "dominant_freq": 1, "fourier_magnitudes_in": [0.0904177725315094, 0.5084242224693298, 0.011039209552109241, 0.18728874623775482, 0.02626693807542324], "fourier_magnitudes_out": [0.07059796154499054, 0.5275266766548157, 0.011537865735590458, 0.1595611721277237, 0.05362024903297424]}, "neuron_5": {"global_index": 73, "dominant_freq": 2, "fourier_magnitudes_in": [0.06564176827669144, 0.06385522335767746, 0.07292439043521881, 0.12749841809272766, 0.5017807483673096], "fourier_magnitudes_out": [0.10108558088541031, 0.06667263805866241, 0.16434097290039062, 0.410533607006073, 0.28931868076324463]}, "neuron_6": {"global_index": 483, "dominant_freq": 1, "fourier_magnitudes_in": [0.08611836284399033, 0.5039265155792236, 0.017540622502565384, 0.13896282017230988, 0.08160430192947388], "fourier_magnitudes_out": [0.11595603078603745, 0.5027888417243958, 0.021235240623354912, 0.14109770953655243, 0.027328623458743095]}, "neuron_7": {"global_index": 459, "dominant_freq": 2, "fourier_magnitudes_in": [0.12552514672279358, 0.08680035173892975, 0.10824543982744217, 0.1438632756471634, 0.5015738010406494], "fourier_magnitudes_out": [0.038486167788505554, 0.056414779275655746, 0.10874727368354797, 0.44881340861320496, 0.3089643716812134]}, "neuron_8": {"global_index": 72, "dominant_freq": 1, "fourier_magnitudes_in": [0.014196483418345451, 0.00610319571569562, 0.4962560534477234, 0.01613822765648365, 0.03811746835708618], "fourier_magnitudes_out": [0.06270183622837067, 0.3995334804058075, 0.1276235729455948, 0.07181502133607864, 0.14153124392032623]}, "neuron_9": {"global_index": 318, "dominant_freq": 2, "fourier_magnitudes_in": [0.12204458564519882, 0.1388137936592102, 0.08629858493804932, 0.15970827639102936, 0.5006576776504517], "fourier_magnitudes_out": [0.021696599200367928, 0.0874486193060875, 0.010394024662673473, 0.38898777961730957, 0.16081035137176514]}, "neuron_10": {"global_index": 237, "dominant_freq": 1, "fourier_magnitudes_in": [0.1312435120344162, 0.15172745287418365, 0.496026873588562, 0.1568685621023178, 0.09524926543235779], "fourier_magnitudes_out": [0.1340775340795517, 0.51954185962677, 0.11751013994216919, 0.17218303680419922, 0.0372869074344635]}, "neuron_11": {"global_index": 401, "dominant_freq": 2, "fourier_magnitudes_in": [0.01433237548917532, 0.017167123034596443, 0.12486902624368668, 0.00703269662335515, 0.4982934296131134], "fourier_magnitudes_out": [0.03908177465200424, 0.0046936869621276855, 0.12982794642448425, 0.4312635660171509, 0.21357746422290802]}, "neuron_12": {"global_index": 206, "dominant_freq": 1, "fourier_magnitudes_in": [0.10658783465623856, 0.4908372759819031, 0.00023165940365288407, 0.18676185607910156, 0.0001426737871952355], "fourier_magnitudes_out": [0.07604879885911942, 0.5404461026191711, 0.019800249487161636, 0.15009360015392303, 0.07335415482521057]}, "neuron_13": {"global_index": 59, "dominant_freq": 2, "fourier_magnitudes_in": [0.008304246701300144, 0.11832792311906815, 0.05578143149614334, 0.49310770630836487, 0.03374133259057999], "fourier_magnitudes_out": [0.04245651885867119, 0.17832817137241364, 0.050221171230077744, 0.5386509299278259, 0.028482872992753983]}, "neuron_14": {"global_index": 259, "dominant_freq": 1, "fourier_magnitudes_in": [0.014181695878505707, 0.0029230420477688313, 0.48269304633140564, 0.012970007956027985, 0.09761106222867966], "fourier_magnitudes_out": [0.0014188888017088175, 0.5021384954452515, 0.16395965218544006, 0.018990054726600647, 0.032611772418022156]}, "neuron_15": {"global_index": 351, "dominant_freq": 2, "fourier_magnitudes_in": [0.10057320445775986, 0.15442830324172974, 0.10147861391305923, 0.17416693270206451, 0.4906236529350281], "fourier_magnitudes_out": [0.05129798874258995, 0.007373969070613384, 0.1536463499069214, 0.4994337856769562, 0.34374621510505676]}, "neuron_16": {"global_index": 458, "dominant_freq": 1, "fourier_magnitudes_in": [0.07627244293689728, 0.4801045060157776, 0.3187965154647827, 0.09155220538377762, 0.1805570423603058], "fourier_magnitudes_out": [0.06700605899095535, 0.16012659668922424, 0.5404700636863708, 0.143175408244133, 0.14613336324691772]}, "neuron_17": {"global_index": 379, "dominant_freq": 2, "fourier_magnitudes_in": [0.11650610715150833, 0.11074958741664886, 0.138760045170784, 0.1441315859556198, 0.4850243330001831], "fourier_magnitudes_out": [0.20369304716587067, 0.0641430988907814, 0.18225109577178955, 0.39435094594955444, 0.30214613676071167]}, "neuron_18": {"global_index": 322, "dominant_freq": 1, "fourier_magnitudes_in": [0.028615474700927734, 0.4798811376094818, 0.1681438833475113, 0.07095656543970108, 0.04339901730418205], "fourier_magnitudes_out": [0.059505004435777664, 0.43107718229293823, 0.12261204421520233, 0.1029493659734726, 0.015014735981822014]}, "neuron_19": {"global_index": 275, "dominant_freq": 2, "fourier_magnitudes_in": [8.879166557562712e-07, 2.1836196992808254e-06, 0.11432433873414993, 7.948009397296119e-07, 0.4843239486217499], "fourier_magnitudes_out": [0.04735521599650383, 0.09558973461389542, 0.051857028156518936, 0.4821547269821167, 0.1215662732720375]}}}
|
precomputed_results/p_005/p005_output_logits.png
ADDED
|
precomputed_results/p_005/p005_overview.json
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"std_epochs": [0, 200, 400, 600, 800, 1000, 1200, 1400, 1600, 1800, 2000, 2200, 2400, 2600, 2800, 3000, 3200, 3400, 3600, 3800, 4000, 4200, 4400, 4600, 4800], "std_ipr": [0.6679549813270569, 0.6757190227508545, 0.6858887672424316, 0.7011680603027344, 0.7171341180801392, 0.7317519783973694, 0.7443163394927979, 0.7545392513275146, 0.7632024884223938, 0.7700366377830505, 0.775721549987793, 0.7804027795791626, 0.7843307256698608, 0.7876507043838501, 0.7904716730117798, 0.7927570343017578, 0.7945860028266907, 0.7960714101791382, 0.7973570823669434, 0.7984920144081116, 0.7993782758712769, 0.8001348972320557, 0.8007866740226746, 0.8013813495635986, 0.8018988370895386], "std_train_loss": [1.6213630437850952, 1.5519102811813354, 1.4744852781295776, 1.3869051933288574, 1.2878625392913818, 1.178138017654419, 1.0597352981567383, 0.9359710216522217, 0.8100465536117554, 0.6858271956443787, 0.5675457715988159, 0.4584864377975464, 0.36111724376678467, 0.2772042453289032, 0.20750892162322998, 0.1516403704881668, 0.10829538106918335, 0.07570212334394455, 0.0518789142370224, 0.03489707410335541, 0.023066148161888123, 0.014997381716966629, 0.009599926881492138, 0.006051937118172646, 0.003758826060220599]}
|
precomputed_results/p_005/p005_overview_loss_ipr.png
ADDED
|
precomputed_results/p_005/p005_overview_phase_scatter.png
ADDED
|
precomputed_results/p_005/p005_phase_align_approx1.png
ADDED
|
Git LFS Details
|
precomputed_results/p_005/p005_phase_align_approx2.png
ADDED
|
Git LFS Details
|
precomputed_results/p_005/p005_phase_align_quad.png
ADDED
|
precomputed_results/p_005/p005_phase_align_relu.png
ADDED
|
precomputed_results/p_005/p005_phase_distribution.png
ADDED
|
precomputed_results/p_005/p005_phase_relationship.png
ADDED
|
precomputed_results/p_005/p005_single_freq_quad.png
ADDED
|
Git LFS Details
|
precomputed_results/p_005/p005_single_freq_relu.png
ADDED
|
Git LFS Details
|
precomputed_results/p_005/p005_training_log.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|