[ { "method": "dfa", "L": 4, "d": 256, "seed": 42, "warmup_ratio": 0.2, "term_weight": 1.0, "M": 4, "test_acc": 0.3116, "mean_gamma": 0.10076353600015864, "mean_rho": -0.00476757250726223, "mean_nudge": -3.260793164372444e-07, "per_layer_gamma": [ 0.42987197637557983, 0.002452872460708022, -0.014306485652923584, -0.014964219182729721 ], "per_layer_rho": [ -0.01907029002904892, 0.0, 0.0, 0.0 ] }, { "method": "vec_eT_M4", "L": 4, "d": 256, "seed": 42, "warmup_ratio": 0.0, "term_weight": 1.0, "M": 4, "test_acc": 0.1585, "mean_gamma": 0.0070959172576294804, "mean_rho": 0.004696090705692768, "mean_nudge": 1.6298145055770874e-09, "per_layer_gamma": [ 0.02838953770697117, -1.1787886251113378e-05, 1.5722671378171071e-06, 4.346942660049535e-06 ], "per_layer_rho": [ 0.018784362822771072, 0.0, 0.0, 0.0 ] }, { "method": "vec_eT_M4", "L": 4, "d": 256, "seed": 42, "warmup_ratio": 0.0, "term_weight": 4.0, "M": 4, "test_acc": 0.1547, "mean_gamma": -0.004175424227696567, "mean_rho": 0.007184227928519249, "mean_nudge": -3.725290298461914e-09, "per_layer_gamma": [ -0.016694847494363785, -1.839158903749194e-05, 5.305797913024435e-06, 6.2363747019844595e-06 ], "per_layer_rho": [ 0.028736911714076996, 0.0, 0.0, 0.0 ] }, { "method": "vec_eT_M4", "L": 4, "d": 256, "seed": 42, "warmup_ratio": 0.05, "term_weight": 1.0, "M": 4, "test_acc": 0.1301, "mean_gamma": 0.0003392250334854907, "mean_rho": -0.006982688792049885, "mean_nudge": 2.561137080192566e-09, "per_layer_gamma": [ 0.0014179275603964925, 2.4803875930956565e-06, -5.080455230199732e-05, -1.270326174562797e-05 ], "per_layer_rho": [ -0.02793075516819954, 0.0, 0.0, 0.0 ] }, { "method": "vec_eT_M4", "L": 4, "d": 256, "seed": 42, "warmup_ratio": 0.05, "term_weight": 4.0, "M": 4, "test_acc": 0.1585, "mean_gamma": 0.001455232677017193, "mean_rho": -0.004349564202129841, "mean_nudge": 2.3283064365386963e-10, "per_layer_gamma": [ 0.005935228429734707, -4.2289950215490535e-05, -2.683553066162858e-05, -4.517224078881554e-05 ], "per_layer_rho": [ -0.017398256808519363, 0.0, 0.0, 0.0 ] }, { "method": "vec_eT_M4", "L": 4, "d": 256, "seed": 42, "warmup_ratio": 0.2, "term_weight": 1.0, "M": 4, "test_acc": 0.2425, "mean_gamma": 0.0011495156340970425, "mean_rho": 0.00014320318587124348, "mean_nudge": -8.847564458847046e-09, "per_layer_gamma": [ 0.004161187447607517, 8.946975140133873e-05, 0.00017050666792783886, 0.00017689866945147514 ], "per_layer_rho": [ 0.0005728127434849739, 0.0, 0.0, 0.0 ] }, { "method": "vec_eT_M4", "L": 4, "d": 256, "seed": 42, "warmup_ratio": 0.2, "term_weight": 4.0, "M": 4, "test_acc": 0.1985, "mean_gamma": 0.0039006864826660603, "mean_rho": 0.0011303124483674765, "mean_nudge": -1.234002411365509e-08, "per_layer_gamma": [ 0.014948057942092419, 0.0001860432676039636, 0.00021447567269206047, 0.00025416904827579856 ], "per_layer_rho": [ 0.004521249793469906, 0.0, 0.0, 0.0 ] } ]