Download maze_unique/source_evaluation.json from helblazer811/FlowReasoningModel: direct link, hf CLI and curl.
- Browser
- Download file 9.35 kB
-
https://huggingface.co/helblazer811/FlowReasoningModel/resolve/main/maze_unique/source_evaluation.json
- Command line
-
hf download hf://helblazer811/FlowReasoningModel/maze_unique/source_evaluation.json
-
curl -L -o source_evaluation.json https://huggingface.co/helblazer811/FlowReasoningModel/resolve/main/maze_unique/source_evaluation.json
9.35 kB
| { | |
| "schema_version": 1, | |
| "git_commit": null, | |
| "checkpoint": "/proj/grp-flow-map/LoopedFlows/experiments/public_simplified_repro/outputs/runs/maze_unique_20261016_v1/full/s0/stage_b/best.pt", | |
| "checkpoint_sha256": "b3115f2954f2d4d92daf7a893a80e1980fdb26f0bc813fb4944bd6f2abb8f698", | |
| "benchmark": "/proj/grp-flow-map/LoopedFlows/experiments/public_simplified_repro/outputs/benchmarks/maze_unique_20261016/test.npz", | |
| "benchmark_sha256": "84c7e4b6ad469a095b53468173ebfe66fbe0542add6be3cb2fbcf0f459ce57a3", | |
| "policy_config": { | |
| "max_nfe": 257, | |
| "max_breadth": 16, | |
| "held_t": 0.3, | |
| "renoise": 1.0, | |
| "check_every": 1, | |
| "residual_window": 16, | |
| "residual_metric": "euclidean", | |
| "confirmation_checks": 2, | |
| "require_prediction_stability": true, | |
| "min_nfe": 4, | |
| "batch_size": 64, | |
| "seed": 0, | |
| "restart_seed_offset": 1000003, | |
| "flops_per_forward": 11336290304 | |
| }, | |
| "residual": "trailing-window mean of per-puzzle euclidean distance between consecutive denoiser distributions, averaged over non-clue cells", | |
| "oracle_used_for_test_selection": false, | |
| "kind": "residual_guided_inference_evaluation", | |
| "run_fingerprint": "750c7e8b18d3bc7d1664e527b91edabb07389212fa6d347262459f43696433ea", | |
| "num_puzzles": 1000, | |
| "max_samples_override": null, | |
| "model_step": 30000, | |
| "calibration": "/proj/grp-flow-map/LoopedFlows/experiments/public_simplified_repro/outputs/runs/maze_unique_20261016_v1/full/s0/calibration/threshold.json", | |
| "calibration_sha256": "e598d82008b8f5e04082f7ec3509f0d1c006b77f14aa2ec8be1e96d21961149d", | |
| "calibration_benchmark_sha256": "3f0d8261ef2977c8a8d219dcb3f6cd146d3e7195438cfbe8939e774d1d8ff0dc", | |
| "threshold": 0.010557100176811218, | |
| "threshold_selection": { | |
| "threshold": 0.010557100176811218, | |
| "accuracy": 1.0, | |
| "solved": 256, | |
| "mean_nfe": 5.0, | |
| "accepted_fraction": 1.0, | |
| "fixed_accuracy": 1.0, | |
| "max_accuracy_drop": 0.001, | |
| "selection_objective": "minimize mean NFE subject to accuracy >= fixed-depth accuracy - max_accuracy_drop; tie-break by higher accuracy then lower threshold" | |
| }, | |
| "policies": [ | |
| { | |
| "policy": "Fixed-depth FRM", | |
| "solved": 996, | |
| "num_puzzles": 1000, | |
| "solve_rate": 0.996, | |
| "mean_nfe": 257.0, | |
| "mean_tflops": 2.913426608128 | |
| }, | |
| { | |
| "policy": "+ Residual early stopping", | |
| "solved": 996, | |
| "num_puzzles": 1000, | |
| "solve_rate": 0.996, | |
| "mean_nfe": 9.734, | |
| "mean_tflops": 0.110347449819136 | |
| }, | |
| { | |
| "policy": "+ Sequential restarts", | |
| "solved": 999, | |
| "num_puzzles": 1000, | |
| "solve_rate": 0.999, | |
| "mean_nfe": 15.708, | |
| "mean_tflops": 0.17807044809523198 | |
| } | |
| ], | |
| "accepted_by_threshold": 999, | |
| "unresolved_at_breadth_cap": 1, | |
| "attempts": [ | |
| { | |
| "attempt": 1, | |
| "attempted": 1000, | |
| "accepted": 996, | |
| "remaining": 4, | |
| "mean_nfe_among_attempted": 9.734 | |
| }, | |
| { | |
| "attempt": 2, | |
| "seed": 1000003, | |
| "attempted": 4, | |
| "accepted": 1, | |
| "remaining": 3, | |
| "mean_nfe_among_attempted": 197.75 | |
| }, | |
| { | |
| "attempt": 3, | |
| "seed": 2000006, | |
| "attempted": 3, | |
| "accepted": 0, | |
| "remaining": 3, | |
| "mean_nfe_among_attempted": 257.0 | |
| }, | |
| { | |
| "attempt": 4, | |
| "seed": 3000009, | |
| "attempted": 3, | |
| "accepted": 0, | |
| "remaining": 3, | |
| "mean_nfe_among_attempted": 257.0 | |
| }, | |
| { | |
| "attempt": 5, | |
| "seed": 4000012, | |
| "attempted": 3, | |
| "accepted": 1, | |
| "remaining": 2, | |
| "mean_nfe_among_attempted": 178.33333333333334 | |
| }, | |
| { | |
| "attempt": 6, | |
| "seed": 5000015, | |
| "attempted": 2, | |
| "accepted": 0, | |
| "remaining": 2, | |
| "mean_nfe_among_attempted": 257.0 | |
| }, | |
| { | |
| "attempt": 7, | |
| "seed": 6000018, | |
| "attempted": 2, | |
| "accepted": 1, | |
| "remaining": 1, | |
| "mean_nfe_among_attempted": 139.5 | |
| }, | |
| { | |
| "attempt": 8, | |
| "seed": 7000021, | |
| "attempted": 1, | |
| "accepted": 0, | |
| "remaining": 1, | |
| "mean_nfe_among_attempted": 257.0 | |
| }, | |
| { | |
| "attempt": 9, | |
| "seed": 8000024, | |
| "attempted": 1, | |
| "accepted": 0, | |
| "remaining": 1, | |
| "mean_nfe_among_attempted": 257.0 | |
| }, | |
| { | |
| "attempt": 10, | |
| "seed": 9000027, | |
| "attempted": 1, | |
| "accepted": 0, | |
| "remaining": 1, | |
| "mean_nfe_among_attempted": 257.0 | |
| }, | |
| { | |
| "attempt": 11, | |
| "seed": 10000030, | |
| "attempted": 1, | |
| "accepted": 0, | |
| "remaining": 1, | |
| "mean_nfe_among_attempted": 257.0 | |
| }, | |
| { | |
| "attempt": 12, | |
| "seed": 11000033, | |
| "attempted": 1, | |
| "accepted": 0, | |
| "remaining": 1, | |
| "mean_nfe_among_attempted": 257.0 | |
| }, | |
| { | |
| "attempt": 13, | |
| "seed": 12000036, | |
| "attempted": 1, | |
| "accepted": 0, | |
| "remaining": 1, | |
| "mean_nfe_among_attempted": 257.0 | |
| }, | |
| { | |
| "attempt": 14, | |
| "seed": 13000039, | |
| "attempted": 1, | |
| "accepted": 0, | |
| "remaining": 1, | |
| "mean_nfe_among_attempted": 257.0 | |
| }, | |
| { | |
| "attempt": 15, | |
| "seed": 14000042, | |
| "attempted": 1, | |
| "accepted": 0, | |
| "remaining": 1, | |
| "mean_nfe_among_attempted": 257.0 | |
| }, | |
| { | |
| "attempt": 16, | |
| "seed": 15000045, | |
| "attempted": 1, | |
| "accepted": 0, | |
| "remaining": 1, | |
| "mean_nfe_among_attempted": 257.0 | |
| } | |
| ], | |
| "restart_frontier": [ | |
| { | |
| "breadth_cap": 1, | |
| "solved": 996, | |
| "num_puzzles": 1000, | |
| "solve_rate": 0.996, | |
| "mean_nfe": 9.734, | |
| "mean_tflops": 0.110347449819136, | |
| "accepted": 996, | |
| "unresolved": 4 | |
| }, | |
| { | |
| "breadth_cap": 2, | |
| "solved": 997, | |
| "num_puzzles": 1000, | |
| "solve_rate": 0.997, | |
| "mean_nfe": 10.525, | |
| "mean_tflops": 0.11931445544960001, | |
| "accepted": 997, | |
| "unresolved": 3 | |
| }, | |
| { | |
| "breadth_cap": 3, | |
| "solved": 997, | |
| "num_puzzles": 1000, | |
| "solve_rate": 0.997, | |
| "mean_nfe": 11.296, | |
| "mean_tflops": 0.128054735273984, | |
| "accepted": 997, | |
| "unresolved": 3 | |
| }, | |
| { | |
| "breadth_cap": 4, | |
| "solved": 997, | |
| "num_puzzles": 1000, | |
| "solve_rate": 0.997, | |
| "mean_nfe": 12.067, | |
| "mean_tflops": 0.136795015098368, | |
| "accepted": 997, | |
| "unresolved": 3 | |
| }, | |
| { | |
| "breadth_cap": 5, | |
| "solved": 998, | |
| "num_puzzles": 1000, | |
| "solve_rate": 0.998, | |
| "mean_nfe": 12.602, | |
| "mean_tflops": 0.142859930411008, | |
| "accepted": 998, | |
| "unresolved": 2 | |
| }, | |
| { | |
| "breadth_cap": 6, | |
| "solved": 998, | |
| "num_puzzles": 1000, | |
| "solve_rate": 0.998, | |
| "mean_nfe": 13.116, | |
| "mean_tflops": 0.148686783627264, | |
| "accepted": 998, | |
| "unresolved": 2 | |
| }, | |
| { | |
| "breadth_cap": 7, | |
| "solved": 999, | |
| "num_puzzles": 1000, | |
| "solve_rate": 0.999, | |
| "mean_nfe": 13.395, | |
| "mean_tflops": 0.15184960862208, | |
| "accepted": 999, | |
| "unresolved": 1 | |
| }, | |
| { | |
| "breadth_cap": 8, | |
| "solved": 999, | |
| "num_puzzles": 1000, | |
| "solve_rate": 0.999, | |
| "mean_nfe": 13.652, | |
| "mean_tflops": 0.15476303523020798, | |
| "accepted": 999, | |
| "unresolved": 1 | |
| }, | |
| { | |
| "breadth_cap": 9, | |
| "solved": 999, | |
| "num_puzzles": 1000, | |
| "solve_rate": 0.999, | |
| "mean_nfe": 13.909, | |
| "mean_tflops": 0.157676461838336, | |
| "accepted": 999, | |
| "unresolved": 1 | |
| }, | |
| { | |
| "breadth_cap": 10, | |
| "solved": 999, | |
| "num_puzzles": 1000, | |
| "solve_rate": 0.999, | |
| "mean_nfe": 14.166, | |
| "mean_tflops": 0.160589888446464, | |
| "accepted": 999, | |
| "unresolved": 1 | |
| }, | |
| { | |
| "breadth_cap": 11, | |
| "solved": 999, | |
| "num_puzzles": 1000, | |
| "solve_rate": 0.999, | |
| "mean_nfe": 14.423, | |
| "mean_tflops": 0.163503315054592, | |
| "accepted": 999, | |
| "unresolved": 1 | |
| }, | |
| { | |
| "breadth_cap": 12, | |
| "solved": 999, | |
| "num_puzzles": 1000, | |
| "solve_rate": 0.999, | |
| "mean_nfe": 14.68, | |
| "mean_tflops": 0.16641674166272, | |
| "accepted": 999, | |
| "unresolved": 1 | |
| }, | |
| { | |
| "breadth_cap": 13, | |
| "solved": 999, | |
| "num_puzzles": 1000, | |
| "solve_rate": 0.999, | |
| "mean_nfe": 14.937, | |
| "mean_tflops": 0.16933016827084799, | |
| "accepted": 999, | |
| "unresolved": 1 | |
| }, | |
| { | |
| "breadth_cap": 14, | |
| "solved": 999, | |
| "num_puzzles": 1000, | |
| "solve_rate": 0.999, | |
| "mean_nfe": 15.194, | |
| "mean_tflops": 0.172243594878976, | |
| "accepted": 999, | |
| "unresolved": 1 | |
| }, | |
| { | |
| "breadth_cap": 15, | |
| "solved": 999, | |
| "num_puzzles": 1000, | |
| "solve_rate": 0.999, | |
| "mean_nfe": 15.451, | |
| "mean_tflops": 0.175157021487104, | |
| "accepted": 999, | |
| "unresolved": 1 | |
| }, | |
| { | |
| "breadth_cap": 16, | |
| "solved": 999, | |
| "num_puzzles": 1000, | |
| "solve_rate": 0.999, | |
| "mean_nfe": 15.708, | |
| "mean_tflops": 0.17807044809523198, | |
| "accepted": 999, | |
| "unresolved": 1 | |
| } | |
| ], | |
| "selection_rule": "stop each trajectory once the trailing-mean residual remains <= threshold for 2 checks and the decoded grid remains unchanged; restart only trajectories that reach max_nfe without acceptance; if max_breadth is exhausted, return the attempted terminal prediction with minimum residual", | |
| "labels_used_for": "final scoring only; never stopping, restart, or candidate selection" | |
| } | |