{ "s_eT_tgw1.0_wr0.2": { "test_acc": 0.5135, "mean_bp_cosine": 0.3966578356921673, "mean_rho": 0.46284765750169754, "mean_nudge": -0.03856433369219303, "bp_cosine_per_layer": [ 0.1582626849412918, 0.46924665570259094, 0.4738205075263977, 0.48530149459838867 ], "rho_per_layer": [ 0.20893594622612, 0.5339754819869995, 0.5519976019859314, 0.5564815998077393 ], "final_value_loss": 0.588364347743988, "s_type": "eT", "term_grad_weight": 1.0, "warmup_ratio": 0.2 }, "s_deltaL_tgw1.0_wr0.2": { "test_acc": 0.422, "mean_bp_cosine": 0.556720569729805, "mean_rho": 0.5030678957700729, "mean_nudge": -0.055364650674164295, "bp_cosine_per_layer": [ 0.25414949655532837, 0.6241912841796875, 0.6748597621917725, 0.6736817359924316 ], "rho_per_layer": [ 0.2788790464401245, 0.5604095458984375, 0.5989287495613098, 0.5740542411804199 ], "final_value_loss": 0.6335030339241028, "s_type": "deltaL", "term_grad_weight": 1.0, "warmup_ratio": 0.2 } }