{ "mode": "marginal", "args": { "mode": "marginal", "task": "belief_line", "seed": 23, "modulus": 11, "op_family": "full", "train_min_len": 1, "train_max_len": 6, "eval_lengths": "3,6,9,12", "eval_k": "0,1,2,3,6,9,12", "transition_temperature": 1.0, "dim": 128, "heads": 4, "compiler_layers": 2, "dropout": 0.05, "train_steps": 300, "batch_size": 512, "eval_batch_size": 512, "eval_examples": 1024, "lr": 0.03, "weight_decay": 0.0, "max_grad_norm": 1.0, "log_every": 25, "eval_every": 100, "output_dir": "experiments/joint_register_executor/runs/control_belief_marginal_mod11", "checkpoint_dir": "large_artifacts/joint_register_executor/checkpoints/control_belief_marginal_mod11" }, "env": { "python": "3.12.3", "torch": "2.8.0+cu128", "device": "cuda", "cuda_name": "NVIDIA RTX 6000 Ada Generation" }, "train": [ { "step": 1, "loss": 4.795785903930664, "pair_acc": 0.125, "grad_norm": 0.0024538147263228893, "transition_entropy": 2.397849361101786 }, { "step": 25, "loss": 4.798885345458984, "pair_acc": 0.08984375, "grad_norm": 0.0027714166790246964, "transition_entropy": 2.3969095945358276 }, { "step": 50, "loss": 4.7967848777771, "pair_acc": 0.078125, "grad_norm": 0.002315132413059473, "transition_entropy": 2.3971646626790366 }, { "step": 75, "loss": 4.79749059677124, "pair_acc": 0.08203125, "grad_norm": 0.0024698241613805294, "transition_entropy": 2.397071282068888 }, { "step": 100, "loss": 4.797521114349365, "pair_acc": 0.09375, "grad_norm": 0.0029555645305663347, "transition_entropy": 2.3969992796579995 }, { "step": 125, "loss": 4.797061920166016, "pair_acc": 0.083984375, "grad_norm": 0.0025547503028064966, "transition_entropy": 2.3971678415934243 }, { "step": 150, "loss": 4.797024726867676, "pair_acc": 0.072265625, "grad_norm": 0.0029474636539816856, "transition_entropy": 2.396950125694275 }, { "step": 175, "loss": 4.796481132507324, "pair_acc": 0.080078125, "grad_norm": 0.002674966584891081, "transition_entropy": 2.396949132283529 }, { "step": 200, "loss": 4.7977447509765625, "pair_acc": 0.111328125, "grad_norm": 0.00251887203194201, "transition_entropy": 2.3970172007878623 }, { "step": 225, "loss": 4.7966485023498535, "pair_acc": 0.0703125, "grad_norm": 0.002432065550237894, "transition_entropy": 2.396958867708842 }, { "step": 250, "loss": 4.797397136688232, "pair_acc": 0.083984375, "grad_norm": 0.0026773589197546244, "transition_entropy": 2.396987716356913 }, { "step": 275, "loss": 4.798551082611084, "pair_acc": 0.078125, "grad_norm": 0.0025181712117046118, "transition_entropy": 2.3969995578130088 }, { "step": 300, "loss": 4.796629905700684, "pair_acc": 0.09765625, "grad_norm": 0.0027341418899595737, "transition_entropy": 2.3969560464223227 } ], "eval": [ { "step": 100, "rows": [ { "model": "marginal", "length": 3, "k": 0, "n": 1024, "pair_accuracy": 0.0966796875, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0909090859375, "target_nll": 4.7957911484375, "top1_on_support": 0.0966796875 }, { "model": "marginal", "length": 3, "k": 1, "n": 1024, "pair_accuracy": 0.087890625, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0909065048828125, "target_nll": 4.797151088867188, "top1_on_support": 0.087890625 }, { "model": "marginal", "length": 3, "k": 2, "n": 1024, "pair_accuracy": 0.0947265625, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0908829052734375, "target_nll": 4.79812884375, "top1_on_support": 0.0947265625 }, { "model": "marginal", "length": 3, "k": 3, "n": 1024, "pair_accuracy": 0.0908203125, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0907360693359375, "target_nll": 4.800102233398437, "top1_on_support": 0.0908203125 }, { "model": "marginal", "length": 3, "k": 6, "n": 1024, "pair_accuracy": 0.0908203125, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0907360693359375, "target_nll": 4.800102233398437, "top1_on_support": 0.0908203125 }, { "model": "marginal", "length": 3, "k": 9, "n": 1024, "pair_accuracy": 0.0908203125, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.090736072265625, "target_nll": 4.800102233398437, "top1_on_support": 0.0908203125 }, { "model": "marginal", "length": 3, "k": 12, "n": 1024, "pair_accuracy": 0.0908203125, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0907360693359375, "target_nll": 4.800102233398437, "top1_on_support": 0.0908203125 }, { "model": "marginal", "length": 6, "k": 0, "n": 1024, "pair_accuracy": 0.091796875, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0909090859375, "target_nll": 4.7957911484375, "top1_on_support": 0.091796875 }, { "model": "marginal", "length": 6, "k": 1, "n": 1024, "pair_accuracy": 0.1015625, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.090886146484375, "target_nll": 4.797472953125, "top1_on_support": 0.1015625 }, { "model": "marginal", "length": 6, "k": 2, "n": 1024, "pair_accuracy": 0.0712890625, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0908990654296875, "target_nll": 4.797838926757812, "top1_on_support": 0.0712890625 }, { "model": "marginal", "length": 6, "k": 3, "n": 1024, "pair_accuracy": 0.072265625, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.090867609375, "target_nll": 4.798536061523437, "top1_on_support": 0.072265625 }, { "model": "marginal", "length": 6, "k": 6, "n": 1024, "pair_accuracy": 0.0830078125, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0908988447265625, "target_nll": 4.798659563476562, "top1_on_support": 0.0830078125 }, { "model": "marginal", "length": 6, "k": 9, "n": 1024, "pair_accuracy": 0.0830078125, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0908988447265625, "target_nll": 4.798659325195312, "top1_on_support": 0.0830078125 }, { "model": "marginal", "length": 6, "k": 12, "n": 1024, "pair_accuracy": 0.0830078125, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0908988447265625, "target_nll": 4.798659325195312, "top1_on_support": 0.0830078125 }, { "model": "marginal", "length": 9, "k": 0, "n": 1024, "pair_accuracy": 0.0927734375, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0909090859375, "target_nll": 4.7957911484375, "top1_on_support": 0.0927734375 }, { "model": "marginal", "length": 9, "k": 1, "n": 1024, "pair_accuracy": 0.0927734375, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.090908650390625, "target_nll": 4.797157764648437, "top1_on_support": 0.0927734375 }, { "model": "marginal", "length": 9, "k": 2, "n": 1024, "pair_accuracy": 0.083984375, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0908371025390625, "target_nll": 4.798607349609375, "top1_on_support": 0.083984375 }, { "model": "marginal", "length": 9, "k": 3, "n": 1024, "pair_accuracy": 0.08984375, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0908429921875, "target_nll": 4.7988531591796875, "top1_on_support": 0.08984375 }, { "model": "marginal", "length": 9, "k": 6, "n": 1024, "pair_accuracy": 0.0908203125, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0909140625, "target_nll": 4.798410416015625, "top1_on_support": 0.0908203125 }, { "model": "marginal", "length": 9, "k": 9, "n": 1024, "pair_accuracy": 0.08984375, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0908342158203125, "target_nll": 4.79932284375, "top1_on_support": 0.08984375 }, { "model": "marginal", "length": 9, "k": 12, "n": 1024, "pair_accuracy": 0.08984375, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0908342197265625, "target_nll": 4.79932284375, "top1_on_support": 0.08984375 }, { "model": "marginal", "length": 12, "k": 0, "n": 1024, "pair_accuracy": 0.07421875, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0909090859375, "target_nll": 4.7957911484375, "top1_on_support": 0.07421875 }, { "model": "marginal", "length": 12, "k": 1, "n": 1024, "pair_accuracy": 0.0908203125, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0909481044921875, "target_nll": 4.796639204101562, "top1_on_support": 0.0908203125 }, { "model": "marginal", "length": 12, "k": 2, "n": 1024, "pair_accuracy": 0.083984375, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.09085569140625, "target_nll": 4.7983684541015625, "top1_on_support": 0.083984375 }, { "model": "marginal", "length": 12, "k": 3, "n": 1024, "pair_accuracy": 0.076171875, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0908739013671875, "target_nll": 4.798529625, "top1_on_support": 0.076171875 }, { "model": "marginal", "length": 12, "k": 6, "n": 1024, "pair_accuracy": 0.1005859375, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0910003037109375, "target_nll": 4.797370434570312, "top1_on_support": 0.1005859375 }, { "model": "marginal", "length": 12, "k": 9, "n": 1024, "pair_accuracy": 0.0986328125, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.090799115234375, "target_nll": 4.799602508789063, "top1_on_support": 0.0986328125 }, { "model": "marginal", "length": 12, "k": 12, "n": 1024, "pair_accuracy": 0.080078125, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.090869359375, "target_nll": 4.798846721679688, "top1_on_support": 0.080078125 } ] }, { "step": 200, "rows": [ { "model": "marginal", "length": 3, "k": 0, "n": 1024, "pair_accuracy": 0.091796875, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0909090859375, "target_nll": 4.7957911484375, "top1_on_support": 0.091796875 }, { "model": "marginal", "length": 3, "k": 1, "n": 1024, "pair_accuracy": 0.0927734375, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.090840376953125, "target_nll": 4.797792911132812, "top1_on_support": 0.0927734375 }, { "model": "marginal", "length": 3, "k": 2, "n": 1024, "pair_accuracy": 0.09375, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.09081066015625, "target_nll": 4.798792838867188, "top1_on_support": 0.09375 }, { "model": "marginal", "length": 3, "k": 3, "n": 1024, "pair_accuracy": 0.087890625, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0908528271484375, "target_nll": 4.798640250976563, "top1_on_support": 0.087890625 }, { "model": "marginal", "length": 3, "k": 6, "n": 1024, "pair_accuracy": 0.087890625, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0908528271484375, "target_nll": 4.798640250976563, "top1_on_support": 0.087890625 }, { "model": "marginal", "length": 3, "k": 9, "n": 1024, "pair_accuracy": 0.087890625, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0908528271484375, "target_nll": 4.798640250976563, "top1_on_support": 0.087890625 }, { "model": "marginal", "length": 3, "k": 12, "n": 1024, "pair_accuracy": 0.087890625, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0908528271484375, "target_nll": 4.798640250976563, "top1_on_support": 0.087890625 }, { "model": "marginal", "length": 6, "k": 0, "n": 1024, "pair_accuracy": 0.083984375, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0909090859375, "target_nll": 4.7957911484375, "top1_on_support": 0.083984375 }, { "model": "marginal", "length": 6, "k": 1, "n": 1024, "pair_accuracy": 0.0830078125, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0909052265625, "target_nll": 4.797087907226563, "top1_on_support": 0.0830078125 }, { "model": "marginal", "length": 6, "k": 2, "n": 1024, "pair_accuracy": 0.08984375, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.090980890625, "target_nll": 4.796915291992187, "top1_on_support": 0.08984375 }, { "model": "marginal", "length": 6, "k": 3, "n": 1024, "pair_accuracy": 0.09765625, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.090992767578125, "target_nll": 4.797054290039062, "top1_on_support": 0.09765625 }, { "model": "marginal", "length": 6, "k": 6, "n": 1024, "pair_accuracy": 0.0703125, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0908336748046875, "target_nll": 4.798982143554688, "top1_on_support": 0.0703125 }, { "model": "marginal", "length": 6, "k": 9, "n": 1024, "pair_accuracy": 0.0703125, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0908336748046875, "target_nll": 4.798982143554688, "top1_on_support": 0.0703125 }, { "model": "marginal", "length": 6, "k": 12, "n": 1024, "pair_accuracy": 0.0703125, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0908336748046875, "target_nll": 4.798982143554688, "top1_on_support": 0.0703125 }, { "model": "marginal", "length": 9, "k": 0, "n": 1024, "pair_accuracy": 0.103515625, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0909090859375, "target_nll": 4.7957911484375, "top1_on_support": 0.103515625 }, { "model": "marginal", "length": 9, "k": 1, "n": 1024, "pair_accuracy": 0.103515625, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0909148603515625, "target_nll": 4.796957016601563, "top1_on_support": 0.103515625 }, { "model": "marginal", "length": 9, "k": 2, "n": 1024, "pair_accuracy": 0.099609375, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.09100406640625, "target_nll": 4.796620845703125, "top1_on_support": 0.099609375 }, { "model": "marginal", "length": 9, "k": 3, "n": 1024, "pair_accuracy": 0.0859375, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.090914271484375, "target_nll": 4.797916412109375, "top1_on_support": 0.0859375 }, { "model": "marginal", "length": 9, "k": 6, "n": 1024, "pair_accuracy": 0.0986328125, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0909840126953125, "target_nll": 4.7973775859375, "top1_on_support": 0.0986328125 }, { "model": "marginal", "length": 9, "k": 9, "n": 1024, "pair_accuracy": 0.09375, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0909014423828125, "target_nll": 4.7984268671875, "top1_on_support": 0.09375 }, { "model": "marginal", "length": 9, "k": 12, "n": 1024, "pair_accuracy": 0.09375, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0909014423828125, "target_nll": 4.7984268671875, "top1_on_support": 0.09375 }, { "model": "marginal", "length": 12, "k": 0, "n": 1024, "pair_accuracy": 0.076171875, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0909090859375, "target_nll": 4.7957911484375, "top1_on_support": 0.076171875 }, { "model": "marginal", "length": 12, "k": 1, "n": 1024, "pair_accuracy": 0.0888671875, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0909364443359375, "target_nll": 4.79674530078125, "top1_on_support": 0.0888671875 }, { "model": "marginal", "length": 12, "k": 2, "n": 1024, "pair_accuracy": 0.0869140625, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.090917140625, "target_nll": 4.797536850585938, "top1_on_support": 0.0869140625 }, { "model": "marginal", "length": 12, "k": 3, "n": 1024, "pair_accuracy": 0.0791015625, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.09099336328125, "target_nll": 4.797007083984375, "top1_on_support": 0.0791015625 }, { "model": "marginal", "length": 12, "k": 6, "n": 1024, "pair_accuracy": 0.0849609375, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.090907140625, "target_nll": 4.798235416015625, "top1_on_support": 0.0849609375 }, { "model": "marginal", "length": 12, "k": 9, "n": 1024, "pair_accuracy": 0.0986328125, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0909301640625, "target_nll": 4.7981281279296875, "top1_on_support": 0.0986328125 }, { "model": "marginal", "length": 12, "k": 12, "n": 1024, "pair_accuracy": 0.099609375, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0909449609375, "target_nll": 4.797840833007813, "top1_on_support": 0.099609375 } ] }, { "step": 300, "rows": [ { "model": "marginal", "length": 3, "k": 0, "n": 1024, "pair_accuracy": 0.0859375, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0909090859375, "target_nll": 4.7957911484375, "top1_on_support": 0.0859375 }, { "model": "marginal", "length": 3, "k": 1, "n": 1024, "pair_accuracy": 0.0869140625, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0910174365234375, "target_nll": 4.7958855625, "top1_on_support": 0.0869140625 }, { "model": "marginal", "length": 3, "k": 2, "n": 1024, "pair_accuracy": 0.0869140625, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.091065142578125, "target_nll": 4.79595851953125, "top1_on_support": 0.0869140625 }, { "model": "marginal", "length": 3, "k": 3, "n": 1024, "pair_accuracy": 0.1015625, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.09111871875, "target_nll": 4.795761346679687, "top1_on_support": 0.1015625 }, { "model": "marginal", "length": 3, "k": 6, "n": 1024, "pair_accuracy": 0.1015625, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.09111872265625, "target_nll": 4.795761346679687, "top1_on_support": 0.1015625 }, { "model": "marginal", "length": 3, "k": 9, "n": 1024, "pair_accuracy": 0.1015625, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.09111872265625, "target_nll": 4.795761346679687, "top1_on_support": 0.1015625 }, { "model": "marginal", "length": 3, "k": 12, "n": 1024, "pair_accuracy": 0.1015625, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.09111872265625, "target_nll": 4.795761346679687, "top1_on_support": 0.1015625 }, { "model": "marginal", "length": 6, "k": 0, "n": 1024, "pair_accuracy": 0.091796875, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0909090859375, "target_nll": 4.7957911484375, "top1_on_support": 0.091796875 }, { "model": "marginal", "length": 6, "k": 1, "n": 1024, "pair_accuracy": 0.109375, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0909179521484375, "target_nll": 4.79699039453125, "top1_on_support": 0.109375 }, { "model": "marginal", "length": 6, "k": 2, "n": 1024, "pair_accuracy": 0.115234375, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.091044302734375, "target_nll": 4.796286344726562, "top1_on_support": 0.115234375 }, { "model": "marginal", "length": 6, "k": 3, "n": 1024, "pair_accuracy": 0.0966796875, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0909652265625, "target_nll": 4.79754638671875, "top1_on_support": 0.0966796875 }, { "model": "marginal", "length": 6, "k": 6, "n": 1024, "pair_accuracy": 0.09375, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0908680595703125, "target_nll": 4.798712969726562, "top1_on_support": 0.09375 }, { "model": "marginal", "length": 6, "k": 9, "n": 1024, "pair_accuracy": 0.09375, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0908680595703125, "target_nll": 4.798712969726562, "top1_on_support": 0.09375 }, { "model": "marginal", "length": 6, "k": 12, "n": 1024, "pair_accuracy": 0.09375, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0908680595703125, "target_nll": 4.798712969726562, "top1_on_support": 0.09375 }, { "model": "marginal", "length": 9, "k": 0, "n": 1024, "pair_accuracy": 0.091796875, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0909090859375, "target_nll": 4.7957911484375, "top1_on_support": 0.091796875 }, { "model": "marginal", "length": 9, "k": 1, "n": 1024, "pair_accuracy": 0.091796875, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0909346484375, "target_nll": 4.796810388671875, "top1_on_support": 0.091796875 }, { "model": "marginal", "length": 9, "k": 2, "n": 1024, "pair_accuracy": 0.08203125, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.091003681640625, "target_nll": 4.7967131142578125, "top1_on_support": 0.08203125 }, { "model": "marginal", "length": 9, "k": 3, "n": 1024, "pair_accuracy": 0.1025390625, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0910280859375, "target_nll": 4.7967987060546875, "top1_on_support": 0.1025390625 }, { "model": "marginal", "length": 9, "k": 6, "n": 1024, "pair_accuracy": 0.091796875, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0908612421875, "target_nll": 4.798802375976562, "top1_on_support": 0.091796875 }, { "model": "marginal", "length": 9, "k": 9, "n": 1024, "pair_accuracy": 0.0751953125, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0908829833984375, "target_nll": 4.79863834375, "top1_on_support": 0.0751953125 }, { "model": "marginal", "length": 9, "k": 12, "n": 1024, "pair_accuracy": 0.0751953125, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0908829794921875, "target_nll": 4.79863810546875, "top1_on_support": 0.0751953125 }, { "model": "marginal", "length": 12, "k": 0, "n": 1024, "pair_accuracy": 0.099609375, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0909090859375, "target_nll": 4.7957911484375, "top1_on_support": 0.099609375 }, { "model": "marginal", "length": 12, "k": 1, "n": 1024, "pair_accuracy": 0.10546875, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.090962126953125, "target_nll": 4.796519041015625, "top1_on_support": 0.10546875 }, { "model": "marginal", "length": 12, "k": 2, "n": 1024, "pair_accuracy": 0.1064453125, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.090942259765625, "target_nll": 4.797345638671875, "top1_on_support": 0.1064453125 }, { "model": "marginal", "length": 12, "k": 3, "n": 1024, "pair_accuracy": 0.0986328125, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.090854931640625, "target_nll": 4.798605442382812, "top1_on_support": 0.0986328125 }, { "model": "marginal", "length": 12, "k": 6, "n": 1024, "pair_accuracy": 0.0986328125, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0909062919921875, "target_nll": 4.798448801757813, "top1_on_support": 0.0986328125 }, { "model": "marginal", "length": 12, "k": 9, "n": 1024, "pair_accuracy": 0.0966796875, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0908220556640625, "target_nll": 4.799487352539063, "top1_on_support": 0.0966796875 }, { "model": "marginal", "length": 12, "k": 12, "n": 1024, "pair_accuracy": 0.0927734375, "a_accuracy": NaN, "b_accuracy": NaN, "register_accuracy": NaN, "target_mass": 0.0908806396484375, "target_nll": 4.7988157265625, "top1_on_support": 0.0927734375 } ] } ], "checkpoints": [ "large_artifacts/joint_register_executor/checkpoints/control_belief_marginal_mod11/checkpoint_step00100.pt", "large_artifacts/joint_register_executor/checkpoints/control_belief_marginal_mod11/checkpoint_step00200.pt", "large_artifacts/joint_register_executor/checkpoints/control_belief_marginal_mod11/checkpoint_step00300.pt" ] }