{ "s_eT_tgw1.0_wr0.2": { "test_acc": 0.4935, "mean_bp_cosine": 0.49867621809244156, "mean_rho": 0.5327592641115189, "mean_nudge": -0.04268983006477356, "bp_cosine_per_layer": [ 0.21021124720573425, 0.5842739939689636, 0.5913121700286865, 0.6089074611663818 ], "rho_per_layer": [ 0.24787980318069458, 0.6222057342529297, 0.6273762583732605, 0.6335752606391907 ], "final_value_loss": 0.4651986366271973, "s_type": "eT", "term_grad_weight": 1.0, "warmup_ratio": 0.2 }, "s_deltaL_tgw1.0_wr0.2": { "test_acc": 0.4805, "mean_bp_cosine": 0.572130560874939, "mean_rho": 0.5290974006056786, "mean_nudge": -0.05305152013897896, "bp_cosine_per_layer": [ 0.3073272705078125, 0.6108690500259399, 0.6810163259506226, 0.6893095970153809 ], "rho_per_layer": [ 0.33536288142204285, 0.6113735437393188, 0.6066526174545288, 0.5630005598068237 ], "final_value_loss": 0.48287113523483277, "s_type": "deltaL", "term_grad_weight": 1.0, "warmup_ratio": 0.2 } }