diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000125_test__val_iceplains__0__000044/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000125_test__val_iceplains__0__000044/case.json new file mode 100644 index 0000000000000000000000000000000000000000..6348435c7f41a151f6e2f5f76731b6a1472e1c6f --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000125_test__val_iceplains__0__000044/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "125", + "global_test_index": "125", + "dataset_index": "358", + "relative_path": "test/val_iceplains/0/000044.mp4", + "clip_id": "test/val_iceplains/0/000044", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000125, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.56510119698942, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000125_test__val_iceplains__0__000044/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000125_test__val_iceplains__0__000044/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..d261dd3c6c520dc93d89c77d6c2fddc7ddd75f8f --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000125_test__val_iceplains__0__000044/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b811c35c2d3535d357e8db41b59e856247ad065cdf78f816ca7dfba55c853d3d +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000125_test__val_iceplains__0__000044/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000125_test__val_iceplains__0__000044/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..ea73101f867253b31489132bc654581bbe8e321c --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000125_test__val_iceplains__0__000044/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "125", + "global_test_index": "125", + "dataset_index": "358", + "relative_path": "test/val_iceplains/0/000044.mp4", + "clip_id": "test/val_iceplains/0/000044", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000125, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/0/000044_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0008869441808201373, + "psnr": 30.563405990600586, + "lpips": 0.01116867270320654 + }, + "frames_100_200": { + "mse": 0.0008632521494291723, + "psnr": 30.660537719726562, + "lpips": 0.011587833054363728 + }, + "frames_200_300": { + "mse": 0.0008356715552508831, + "psnr": 30.796979904174805, + "lpips": 0.011783415451645851 + }, + "frames_300_400": { + "mse": 0.0011603445746004581, + "psnr": 30.075002670288086, + "lpips": 0.017254747450351715 + }, + "frames_400_500": { + "mse": 0.0008149634813889861, + "psnr": 30.907333374023438, + "lpips": 0.013248861767351627 + }, + "overall": { + "mse": 0.0009122352348640561, + "psnr": 30.60065269470215, + "lpips": 0.013008706271648407 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000126_test__val_desert__1__000020/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000126_test__val_desert__1__000020/case.json new file mode 100644 index 0000000000000000000000000000000000000000..7eb8518a34febeeb6d89883e49a0249342f111a7 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000126_test__val_desert__1__000020/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "126", + "global_test_index": "126", + "dataset_index": "50", + "relative_path": "test/val_desert/1/000020.mp4", + "clip_id": "test/val_desert/1/000020", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000126, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 223.92392826499417, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000126_test__val_desert__1__000020/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000126_test__val_desert__1__000020/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..bd6801bb54dc7193591252d7b63e4e19782a0363 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000126_test__val_desert__1__000020/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ee148a7f6e4eda6ca86c8a4c815e296f07412acf0fe5f913edffde3d64fabccf +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000126_test__val_desert__1__000020/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000126_test__val_desert__1__000020/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..79dbf9e072a28392e3a8f1503b3dbb4aa053de48 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000126_test__val_desert__1__000020/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "126", + "global_test_index": "126", + "dataset_index": "50", + "relative_path": "test/val_desert/1/000020.mp4", + "clip_id": "test/val_desert/1/000020", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000126, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_desert/1/000020_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0015150229446589947, + "psnr": 29.627111434936523, + "lpips": 0.018530523404479027 + }, + "frames_100_200": { + "mse": 0.0007773140096105635, + "psnr": 31.101455688476562, + "lpips": 0.010587131604552269 + }, + "frames_200_300": { + "mse": 0.0008125437889248133, + "psnr": 30.920879364013672, + "lpips": 0.011569960042834282 + }, + "frames_300_400": { + "mse": 0.002293758327141404, + "psnr": 28.208471298217773, + "lpips": 0.027469586580991745 + }, + "frames_400_500": { + "mse": 0.008861529640853405, + "psnr": 23.146879196166992, + "lpips": 0.06957513093948364 + }, + "overall": { + "mse": 0.0028520335908979177, + "psnr": 28.60095977783203, + "lpips": 0.027546467259526253 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000127_test__val_savanna__0__000036/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000127_test__val_savanna__0__000036/case.json new file mode 100644 index 0000000000000000000000000000000000000000..077d6a3b930af9e9a0d9dcc627be7bc9ea858129 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000127_test__val_savanna__0__000036/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "127", + "global_test_index": "127", + "dataset_index": "207", + "relative_path": "test/val_savanna/0/000036.mp4", + "clip_id": "test/val_savanna/0/000036", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000127, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.1422406550264, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000127_test__val_savanna__0__000036/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000127_test__val_savanna__0__000036/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..eb65ac62313cf74292c79896dbdf7d58da53c154 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000127_test__val_savanna__0__000036/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1641914b897cb04070a98ae157a1cd1873a337a77307ef13fa1ac4656fc29597 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000127_test__val_savanna__0__000036/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000127_test__val_savanna__0__000036/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..faad1f1cb92ca69d1e0b946c973024caade78f99 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000127_test__val_savanna__0__000036/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "127", + "global_test_index": "127", + "dataset_index": "207", + "relative_path": "test/val_savanna/0/000036.mp4", + "clip_id": "test/val_savanna/0/000036", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000127, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_savanna/0/000036_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.006850932724773884, + "psnr": 22.02444076538086, + "lpips": 0.34593069553375244 + }, + "frames_100_200": { + "mse": 0.008210565894842148, + "psnr": 21.02499008178711, + "lpips": 0.44283801317214966 + }, + "frames_200_300": { + "mse": 0.007101738825440407, + "psnr": 21.556961059570312, + "lpips": 0.4233298599720001 + }, + "frames_300_400": { + "mse": 0.016229726374149323, + "psnr": 19.134199142456055, + "lpips": 0.48480424284935 + }, + "frames_400_500": { + "mse": 0.015277471393346786, + "psnr": 18.69312286376953, + "lpips": 0.536292552947998 + }, + "overall": { + "mse": 0.010734088718891144, + "psnr": 20.48674201965332, + "lpips": 0.4466390609741211 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000128_test__val_iceplains__2__000034/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000128_test__val_iceplains__2__000034/case.json new file mode 100644 index 0000000000000000000000000000000000000000..e458971502ed2a098b293103bd9581f9c3a9a354 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000128_test__val_iceplains__2__000034/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "128", + "global_test_index": "128", + "dataset_index": "330", + "relative_path": "test/val_iceplains/2/000034.mp4", + "clip_id": "test/val_iceplains/2/000034", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000128, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.86871701898053, + "peak_generation_vram_bytes": 107565475840, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000128_test__val_iceplains__2__000034/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000128_test__val_iceplains__2__000034/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..7203494d6a6851fd45dd861c68ebecc6e017a6da --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000128_test__val_iceplains__2__000034/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:07026d8e83b8a30e008a61d83ad60ee573eef257e1633738f52ee1172f156fec +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000128_test__val_iceplains__2__000034/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000128_test__val_iceplains__2__000034/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..0b9f53a4516d496ef97e443f3195cb6bbc2d0852 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000128_test__val_iceplains__2__000034/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "128", + "global_test_index": "128", + "dataset_index": "330", + "relative_path": "test/val_iceplains/2/000034.mp4", + "clip_id": "test/val_iceplains/2/000034", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000128, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/2/000034_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0009578756871633232, + "psnr": 30.307897567749023, + "lpips": 0.01631380245089531 + }, + "frames_100_200": { + "mse": 0.0011427260469645262, + "psnr": 30.443988800048828, + "lpips": 0.014719050377607346 + }, + "frames_200_300": { + "mse": 0.012856044806540012, + "psnr": 24.947832107543945, + "lpips": 0.08157982677221298 + }, + "frames_300_400": { + "mse": 0.13289088010787964, + "psnr": 9.220673561096191, + "lpips": 0.5322049856185913 + }, + "frames_400_500": { + "mse": 0.11107451468706131, + "psnr": 10.005661964416504, + "lpips": 0.5256508588790894 + }, + "overall": { + "mse": 0.05178440362215042, + "psnr": 20.985210418701172, + "lpips": 0.23409372568130493 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000129_test__val_savanna__0__000001/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000129_test__val_savanna__0__000001/case.json new file mode 100644 index 0000000000000000000000000000000000000000..41fa465c56bdafc93a300cb5f04f2e353ce3acb3 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000129_test__val_savanna__0__000001/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "129", + "global_test_index": "129", + "dataset_index": "132", + "relative_path": "test/val_savanna/0/000001.mp4", + "clip_id": "test/val_savanna/0/000001", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000129, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.37514830904547, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000129_test__val_savanna__0__000001/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000129_test__val_savanna__0__000001/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..2bf218abd396a99d2b1616135fd10bc81b5d4d37 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000129_test__val_savanna__0__000001/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:966d0fded792a4674fce9829d153b3b442f529efd9625e8f488a16aa5d41d266 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000129_test__val_savanna__0__000001/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000129_test__val_savanna__0__000001/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..698faa74113a24ef0be3c2c73f4703855871ab66 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000129_test__val_savanna__0__000001/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "129", + "global_test_index": "129", + "dataset_index": "132", + "relative_path": "test/val_savanna/0/000001.mp4", + "clip_id": "test/val_savanna/0/000001", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000129, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_savanna/0/000001_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.001356962020508945, + "psnr": 28.945104598999023, + "lpips": 0.021553361788392067 + }, + "frames_100_200": { + "mse": 0.001741823274642229, + "psnr": 28.603599548339844, + "lpips": 0.02574244886636734 + }, + "frames_200_300": { + "mse": 0.007995889522135258, + "psnr": 21.866762161254883, + "lpips": 0.18454474210739136 + }, + "frames_300_400": { + "mse": 0.014076569117605686, + "psnr": 18.72548484802246, + "lpips": 0.2875327169895172 + }, + "frames_400_500": { + "mse": 0.005448769778013229, + "psnr": 23.1820011138916, + "lpips": 0.08340458571910858 + }, + "overall": { + "mse": 0.006124002393335104, + "psnr": 24.264589309692383, + "lpips": 0.1205555722117424 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000130_test__val_plains__3__000021/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000130_test__val_plains__3__000021/case.json new file mode 100644 index 0000000000000000000000000000000000000000..613405c0efc2c99ec78ec56a8c962d6814b33dc7 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000130_test__val_plains__3__000021/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "130", + "global_test_index": "130", + "dataset_index": "423", + "relative_path": "test/val_plains/3/000021.mp4", + "clip_id": "test/val_plains/3/000021", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000130, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.32027850009035, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000130_test__val_plains__3__000021/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000130_test__val_plains__3__000021/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..9c8dae992a0c9f283f1b5aa9705a3ceac0430d64 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000130_test__val_plains__3__000021/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c390529ea4906fc303e4f7d5c70833d17fd7c53cfcf364a95868afd8173e4f04 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000130_test__val_plains__3__000021/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000130_test__val_plains__3__000021/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..402b0dc0ea603681036c554e61929e8decb255d5 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000130_test__val_plains__3__000021/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "130", + "global_test_index": "130", + "dataset_index": "423", + "relative_path": "test/val_plains/3/000021.mp4", + "clip_id": "test/val_plains/3/000021", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000130, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_plains/3/000021_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.003252933267503977, + "psnr": 25.030717849731445, + "lpips": 0.2486245185136795 + }, + "frames_100_200": { + "mse": 0.004135854076594114, + "psnr": 23.96880340576172, + "lpips": 0.3133731782436371 + }, + "frames_200_300": { + "mse": 0.007658843416720629, + "psnr": 21.572031021118164, + "lpips": 0.3982016444206238 + }, + "frames_300_400": { + "mse": 0.0129533801227808, + "psnr": 21.206653594970703, + "lpips": 0.3382457494735718 + }, + "frames_400_500": { + "mse": 0.04915771633386612, + "psnr": 14.327862739562988, + "lpips": 0.36338722705841064 + }, + "overall": { + "mse": 0.015431745909154415, + "psnr": 21.22121238708496, + "lpips": 0.3323664963245392 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000131_test__val_desert__2__000025/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000131_test__val_desert__2__000025/case.json new file mode 100644 index 0000000000000000000000000000000000000000..20c6d9efe590151f9a9c4afd2e3259b0b916b121 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000131_test__val_desert__2__000025/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "131", + "global_test_index": "131", + "dataset_index": "66", + "relative_path": "test/val_desert/2/000025.mp4", + "clip_id": "test/val_desert/2/000025", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000131, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.4913472509943, + "peak_generation_vram_bytes": 107565475840, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000131_test__val_desert__2__000025/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000131_test__val_desert__2__000025/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..b17b607b074f4177cfd7c6ca2a34ddb317e31aec --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000131_test__val_desert__2__000025/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:34114a4406a11bdd72205f2dc9f41dcdd69c2b759fed3d6c5ee4c330d84259e2 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000131_test__val_desert__2__000025/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000131_test__val_desert__2__000025/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..616cdd8f7b709710472912facac0554cca050fea --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000131_test__val_desert__2__000025/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "131", + "global_test_index": "131", + "dataset_index": "66", + "relative_path": "test/val_desert/2/000025.mp4", + "clip_id": "test/val_desert/2/000025", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000131, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_desert/2/000025_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.012859283946454525, + "psnr": 19.553508758544922, + "lpips": 0.2718072235584259 + }, + "frames_100_200": { + "mse": 0.015745092183351517, + "psnr": 18.143659591674805, + "lpips": 0.29671427607536316 + }, + "frames_200_300": { + "mse": 0.01682784967124462, + "psnr": 18.066320419311523, + "lpips": 0.35653263330459595 + }, + "frames_300_400": { + "mse": 0.017631614580750465, + "psnr": 17.88900375366211, + "lpips": 0.4226989448070526 + }, + "frames_400_500": { + "mse": 0.019850967451930046, + "psnr": 17.446786880493164, + "lpips": 0.43988990783691406 + }, + "overall": { + "mse": 0.01658296026289463, + "psnr": 18.21985626220703, + "lpips": 0.3575286269187927 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000132_test__val_savanna__3__000009/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000132_test__val_savanna__3__000009/case.json new file mode 100644 index 0000000000000000000000000000000000000000..1f31019d435e3295b8137768aeec4cb6c4cba9cc --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000132_test__val_savanna__3__000009/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "132", + "global_test_index": "132", + "dataset_index": "153", + "relative_path": "test/val_savanna/3/000009.mp4", + "clip_id": "test/val_savanna/3/000009", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000132, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.96212022204418, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000132_test__val_savanna__3__000009/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000132_test__val_savanna__3__000009/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..b43f6e1a370b5bbcc9b3fde0683bf6038203243c --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000132_test__val_savanna__3__000009/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6c6c8f0c829e6ccd9a0c76404d6268381a57e0b4be3a4fce371c9e8b1e3f5983 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000132_test__val_savanna__3__000009/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000132_test__val_savanna__3__000009/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..4825ecc14e5dd679e21f63699ecb4ac0df15e881 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000132_test__val_savanna__3__000009/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "132", + "global_test_index": "132", + "dataset_index": "153", + "relative_path": "test/val_savanna/3/000009.mp4", + "clip_id": "test/val_savanna/3/000009", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000132, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_savanna/3/000009_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.01800626888871193, + "psnr": 18.803478240966797, + "lpips": 0.16434983909130096 + }, + "frames_100_200": { + "mse": 0.01665370911359787, + "psnr": 19.057180404663086, + "lpips": 0.1242644339799881 + }, + "frames_200_300": { + "mse": 0.015363054350018501, + "psnr": 18.428556442260742, + "lpips": 0.13614197075366974 + }, + "frames_300_400": { + "mse": 0.03010706976056099, + "psnr": 16.75568389892578, + "lpips": 0.14504624903202057 + }, + "frames_400_500": { + "mse": 0.12154603749513626, + "psnr": 9.465490341186523, + "lpips": 0.29975029826164246 + }, + "overall": { + "mse": 0.04033522680401802, + "psnr": 16.502079010009766, + "lpips": 0.17391057312488556 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000133_test__val_plains__0__000004/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000133_test__val_plains__0__000004/case.json new file mode 100644 index 0000000000000000000000000000000000000000..878235efd2a2ec7c5a021684cec4def13fcbd2f4 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000133_test__val_plains__0__000004/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "133", + "global_test_index": "133", + "dataset_index": "382", + "relative_path": "test/val_plains/0/000004.mp4", + "clip_id": "test/val_plains/0/000004", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000133, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 223.99282652896363, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000133_test__val_plains__0__000004/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000133_test__val_plains__0__000004/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..4cb02a57726eebfe83756c5508a6006f3fece552 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000133_test__val_plains__0__000004/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5687de7afb6ea6854326df135c973aa312fac9c118f5653df30df3176f12de0d +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000133_test__val_plains__0__000004/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000133_test__val_plains__0__000004/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..3ef6414d1317b3fc437f8fbbfd55145ed9f76ec1 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000133_test__val_plains__0__000004/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "133", + "global_test_index": "133", + "dataset_index": "382", + "relative_path": "test/val_plains/0/000004.mp4", + "clip_id": "test/val_plains/0/000004", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000133, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_plains/0/000004_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.006476230453699827, + "psnr": 22.102195739746094, + "lpips": 0.2624092400074005 + }, + "frames_100_200": { + "mse": 0.003971607889980078, + "psnr": 24.036714553833008, + "lpips": 0.3388296961784363 + }, + "frames_200_300": { + "mse": 0.004514392465353012, + "psnr": 23.63833999633789, + "lpips": 0.36804714798927307 + }, + "frames_300_400": { + "mse": 0.007099134847521782, + "psnr": 21.781776428222656, + "lpips": 0.3833627998828888 + }, + "frames_400_500": { + "mse": 0.004460825119167566, + "psnr": 23.611814498901367, + "lpips": 0.3442753553390503 + }, + "overall": { + "mse": 0.005304438527673483, + "psnr": 23.034168243408203, + "lpips": 0.3393848240375519 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000134_test__val_savanna__2__000005/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000134_test__val_savanna__2__000005/case.json new file mode 100644 index 0000000000000000000000000000000000000000..52a88418e8d47b0ab59b3dd45a1bb3884e315eea --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000134_test__val_savanna__2__000005/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "134", + "global_test_index": "134", + "dataset_index": "143", + "relative_path": "test/val_savanna/2/000005.mp4", + "clip_id": "test/val_savanna/2/000005", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000134, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.29743606899865, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000134_test__val_savanna__2__000005/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000134_test__val_savanna__2__000005/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..fb99a6e980f6b29162186b7a1e72c30e3234d09e --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000134_test__val_savanna__2__000005/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3a6cdccc8069b0968dbf43f53ab002b95bae45090f568e66253b277c2b58b835 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000134_test__val_savanna__2__000005/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000134_test__val_savanna__2__000005/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..a14b6beee214e5be92d7333f0ab884b407525517 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000134_test__val_savanna__2__000005/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "134", + "global_test_index": "134", + "dataset_index": "143", + "relative_path": "test/val_savanna/2/000005.mp4", + "clip_id": "test/val_savanna/2/000005", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000134, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_savanna/2/000005_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.021355681121349335, + "psnr": 17.576454162597656, + "lpips": 0.3634519875049591 + }, + "frames_100_200": { + "mse": 0.031553272157907486, + "psnr": 15.716588020324707, + "lpips": 0.4247947931289673 + }, + "frames_200_300": { + "mse": 0.017042307183146477, + "psnr": 19.017942428588867, + "lpips": 0.4350762665271759 + }, + "frames_300_400": { + "mse": 0.03582213819026947, + "psnr": 14.889634132385254, + "lpips": 0.5112087726593018 + }, + "frames_400_500": { + "mse": 0.07696304470300674, + "psnr": 11.411201477050781, + "lpips": 0.5409610867500305 + }, + "overall": { + "mse": 0.03654728829860687, + "psnr": 15.722365379333496, + "lpips": 0.45509862899780273 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000135_test__val_iceplains__1__000040/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000135_test__val_iceplains__1__000040/case.json new file mode 100644 index 0000000000000000000000000000000000000000..ec1c451b277ef742b02d8bf6f5b418e629807a60 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000135_test__val_iceplains__1__000040/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "135", + "global_test_index": "135", + "dataset_index": "348", + "relative_path": "test/val_iceplains/1/000040.mp4", + "clip_id": "test/val_iceplains/1/000040", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000135, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 223.94741691695526, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000135_test__val_iceplains__1__000040/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000135_test__val_iceplains__1__000040/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..e41414d50ed7ac6f137c34f68a3521f6d6be6efb --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000135_test__val_iceplains__1__000040/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a10ca449b3ea159b3a8213dafb8dba072da40e467eac79bbfcf4eb3c6c79a0e5 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000135_test__val_iceplains__1__000040/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000135_test__val_iceplains__1__000040/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..cc44f93d3e4a3e167f4d62b555d8db89a5f7a6af --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000135_test__val_iceplains__1__000040/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "135", + "global_test_index": "135", + "dataset_index": "348", + "relative_path": "test/val_iceplains/1/000040.mp4", + "clip_id": "test/val_iceplains/1/000040", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000135, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/1/000040_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.02679532580077648, + "psnr": 16.233144760131836, + "lpips": 0.22653232514858246 + }, + "frames_100_200": { + "mse": 0.022928809747099876, + "psnr": 17.30470848083496, + "lpips": 0.20770177245140076 + }, + "frames_200_300": { + "mse": 0.030709320679306984, + "psnr": 15.864934921264648, + "lpips": 0.22127622365951538 + }, + "frames_300_400": { + "mse": 0.027509674429893494, + "psnr": 17.74846649169922, + "lpips": 0.28099873661994934 + }, + "frames_400_500": { + "mse": 0.025864748284220695, + "psnr": 16.176921844482422, + "lpips": 0.19212226569652557 + }, + "overall": { + "mse": 0.026761578395962715, + "psnr": 16.665634155273438, + "lpips": 0.22572626173496246 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000136_test__val_plains__2__000005/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000136_test__val_plains__2__000005/case.json new file mode 100644 index 0000000000000000000000000000000000000000..0fda776b5e8f5cae2b37fdadce10cbf215aeda40 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000136_test__val_plains__2__000005/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "136", + "global_test_index": "136", + "dataset_index": "384", + "relative_path": "test/val_plains/2/000005.mp4", + "clip_id": "test/val_plains/2/000005", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000136, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 223.949406081927, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000136_test__val_plains__2__000005/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000136_test__val_plains__2__000005/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..00527d247116ebb806c2448111adb11272620fb1 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000136_test__val_plains__2__000005/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1c20768678209be56a811f022ab4138f36544abfc1455122e6c690c6f89f7a1a +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000136_test__val_plains__2__000005/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000136_test__val_plains__2__000005/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..0374c76fee9f590ba210ebe487dbf8c4fbe59546 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000136_test__val_plains__2__000005/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "136", + "global_test_index": "136", + "dataset_index": "384", + "relative_path": "test/val_plains/2/000005.mp4", + "clip_id": "test/val_plains/2/000005", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000136, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_plains/2/000005_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.003942803014069796, + "psnr": 24.08060073852539, + "lpips": 0.3473111391067505 + }, + "frames_100_200": { + "mse": 0.004088427871465683, + "psnr": 23.92580795288086, + "lpips": 0.36751025915145874 + }, + "frames_200_300": { + "mse": 0.016496507450938225, + "psnr": 18.773754119873047, + "lpips": 0.43271446228027344 + }, + "frames_300_400": { + "mse": 0.01982503943145275, + "psnr": 18.24018096923828, + "lpips": 0.3442988991737366 + }, + "frames_400_500": { + "mse": 0.00846190843731165, + "psnr": 23.310779571533203, + "lpips": 0.04988172650337219 + }, + "overall": { + "mse": 0.010562936775386333, + "psnr": 21.666223526000977, + "lpips": 0.3083432912826538 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000137_test__val_desert__0__000012/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000137_test__val_desert__0__000012/case.json new file mode 100644 index 0000000000000000000000000000000000000000..fcbf30acad70adc55dd010ac1ae9e69278f8ef7e --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000137_test__val_desert__0__000012/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "137", + "global_test_index": "137", + "dataset_index": "30", + "relative_path": "test/val_desert/0/000012.mp4", + "clip_id": "test/val_desert/0/000012", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000137, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.11715133406688, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000137_test__val_desert__0__000012/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000137_test__val_desert__0__000012/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..4ebe28d4bed0fd89e7e49f5c115ad5926e8b1818 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000137_test__val_desert__0__000012/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dcb018ddaf6e2908293a573105f3e04a2051b98cfdf8b15af6bf7db31657f55e +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000137_test__val_desert__0__000012/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000137_test__val_desert__0__000012/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..531e6bf53397031caf9ddf6d83f98b89d3082adf --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000137_test__val_desert__0__000012/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "137", + "global_test_index": "137", + "dataset_index": "30", + "relative_path": "test/val_desert/0/000012.mp4", + "clip_id": "test/val_desert/0/000012", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000137, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_desert/0/000012_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0007586944848299026, + "psnr": 31.212602615356445, + "lpips": 0.009996679611504078 + }, + "frames_100_200": { + "mse": 0.0007581472164019942, + "psnr": 31.214637756347656, + "lpips": 0.010570616461336613 + }, + "frames_200_300": { + "mse": 0.0012523633195087314, + "psnr": 29.58875274658203, + "lpips": 0.022255517542362213 + }, + "frames_300_400": { + "mse": 0.0007854649447835982, + "psnr": 31.06338119506836, + "lpips": 0.010646120645105839 + }, + "frames_400_500": { + "mse": 0.000788546574767679, + "psnr": 31.04404067993164, + "lpips": 0.011787815019488335 + }, + "overall": { + "mse": 0.0008686432265676558, + "psnr": 30.824682235717773, + "lpips": 0.0130513496696949 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000138_test__val_savanna__2__000034/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000138_test__val_savanna__2__000034/case.json new file mode 100644 index 0000000000000000000000000000000000000000..0cf31854bc2961212e41bebb16890caddd4fbda5 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000138_test__val_savanna__2__000034/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "138", + "global_test_index": "138", + "dataset_index": "203", + "relative_path": "test/val_savanna/2/000034.mp4", + "clip_id": "test/val_savanna/2/000034", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000138, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.59760828397702, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000138_test__val_savanna__2__000034/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000138_test__val_savanna__2__000034/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..0e45194016deb2e1453b0f049be1548b8d6d7704 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000138_test__val_savanna__2__000034/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5412f16516a52260cbc653c811f9ad4c5fc4f3e3ec560a0cca14722752aee8e1 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000138_test__val_savanna__2__000034/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000138_test__val_savanna__2__000034/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..a486873f0c50ed6c1c7e25b58399b15b28520267 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000138_test__val_savanna__2__000034/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "138", + "global_test_index": "138", + "dataset_index": "203", + "relative_path": "test/val_savanna/2/000034.mp4", + "clip_id": "test/val_savanna/2/000034", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000138, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_savanna/2/000034_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0022180741652846336, + "psnr": 27.611387252807617, + "lpips": 0.018016261979937553 + }, + "frames_100_200": { + "mse": 0.0016668871976435184, + "psnr": 29.274194717407227, + "lpips": 0.01426667906343937 + }, + "frames_200_300": { + "mse": 0.002634746488183737, + "psnr": 28.3170223236084, + "lpips": 0.0214780792593956 + }, + "frames_300_400": { + "mse": 0.003064398653805256, + "psnr": 27.00786781311035, + "lpips": 0.028961291536688805 + }, + "frames_400_500": { + "mse": 0.01282583363354206, + "psnr": 22.841415405273438, + "lpips": 0.060877710580825806 + }, + "overall": { + "mse": 0.004481988027691841, + "psnr": 27.010377883911133, + "lpips": 0.028720004484057426 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000139_test__val_savanna__0__000022/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000139_test__val_savanna__0__000022/case.json new file mode 100644 index 0000000000000000000000000000000000000000..7eae887cb573c4a48705bd7cd24408e562517ad8 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000139_test__val_savanna__0__000022/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "139", + "global_test_index": "139", + "dataset_index": "179", + "relative_path": "test/val_savanna/0/000022.mp4", + "clip_id": "test/val_savanna/0/000022", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000139, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.41809354000725, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000139_test__val_savanna__0__000022/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000139_test__val_savanna__0__000022/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..ed789c20942e5cdd865b662f8f0a2940d4394a8a --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000139_test__val_savanna__0__000022/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8a7c6581b00c6a430efc566a5d4c7684319dd708d53a2a08df5b18baa143418f +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000139_test__val_savanna__0__000022/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000139_test__val_savanna__0__000022/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..179e10f383c6fcff03327421a6ad51be063f122d --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000139_test__val_savanna__0__000022/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "139", + "global_test_index": "139", + "dataset_index": "179", + "relative_path": "test/val_savanna/0/000022.mp4", + "clip_id": "test/val_savanna/0/000022", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000139, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_savanna/0/000022_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0018945896299555898, + "psnr": 27.78436851501465, + "lpips": 0.14001664519309998 + }, + "frames_100_200": { + "mse": 0.00278268288820982, + "psnr": 25.628538131713867, + "lpips": 0.1555185467004776 + }, + "frames_200_300": { + "mse": 0.003113382961601019, + "psnr": 25.243968963623047, + "lpips": 0.12817919254302979 + }, + "frames_300_400": { + "mse": 0.006654243916273117, + "psnr": 22.103769302368164, + "lpips": 0.10962539911270142 + }, + "frames_400_500": { + "mse": 0.009970344603061676, + "psnr": 20.111858367919922, + "lpips": 0.11824092268943787 + }, + "overall": { + "mse": 0.004883049055933952, + "psnr": 24.17449951171875, + "lpips": 0.13031615316867828 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000140_test__val_iceplains__2__000003/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000140_test__val_iceplains__2__000003/case.json new file mode 100644 index 0000000000000000000000000000000000000000..1d56fd255616ecee44deedab64ec1e0f7c5dd6ab --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000140_test__val_iceplains__2__000003/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "140", + "global_test_index": "140", + "dataset_index": "253", + "relative_path": "test/val_iceplains/2/000003.mp4", + "clip_id": "test/val_iceplains/2/000003", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000140, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 223.66869594098534, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000140_test__val_iceplains__2__000003/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000140_test__val_iceplains__2__000003/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..51eb894d37eba0625c3a3d6a626297395322a0e5 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000140_test__val_iceplains__2__000003/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:af43ce1f6cacc4687e8e716b2d08e5bed195cddc5b36b03d1121e62469bcbe21 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000140_test__val_iceplains__2__000003/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000140_test__val_iceplains__2__000003/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..61f0b5db1d5db3df431de291f611ce157dee6bab --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000140_test__val_iceplains__2__000003/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "140", + "global_test_index": "140", + "dataset_index": "253", + "relative_path": "test/val_iceplains/2/000003.mp4", + "clip_id": "test/val_iceplains/2/000003", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000140, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/2/000003_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.011309000663459301, + "psnr": 20.107036590576172, + "lpips": 0.17270749807357788 + }, + "frames_100_200": { + "mse": 0.029256626963615417, + "psnr": 16.186426162719727, + "lpips": 0.23936496675014496 + }, + "frames_200_300": { + "mse": 0.06650839000940323, + "psnr": 13.023462295532227, + "lpips": 0.30355483293533325 + }, + "frames_300_400": { + "mse": 0.029666537418961525, + "psnr": 16.928707122802734, + "lpips": 0.2705179452896118 + }, + "frames_400_500": { + "mse": 0.0086685661226511, + "psnr": 21.536945343017578, + "lpips": 0.2776356041431427 + }, + "overall": { + "mse": 0.02908182702958584, + "psnr": 17.556516647338867, + "lpips": 0.2527562081813812 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000141_test__val_iceplains__2__000016/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000141_test__val_iceplains__2__000016/case.json new file mode 100644 index 0000000000000000000000000000000000000000..4ef65ec91e202cda105d7db9228dae9482803da8 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000141_test__val_iceplains__2__000016/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "141", + "global_test_index": "141", + "dataset_index": "289", + "relative_path": "test/val_iceplains/2/000016.mp4", + "clip_id": "test/val_iceplains/2/000016", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000141, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.99197798105888, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000141_test__val_iceplains__2__000016/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000141_test__val_iceplains__2__000016/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..c557464826a203f539507a59ee138d90f1394da7 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000141_test__val_iceplains__2__000016/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:45bb9c9d93bf737adda5aacd673cba954956ae97b20275be50509be2434ed1bd +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000141_test__val_iceplains__2__000016/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000141_test__val_iceplains__2__000016/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..ffac96a65fe713a99acd23cbb5eaf3c2d6f923d3 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000141_test__val_iceplains__2__000016/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "141", + "global_test_index": "141", + "dataset_index": "289", + "relative_path": "test/val_iceplains/2/000016.mp4", + "clip_id": "test/val_iceplains/2/000016", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000141, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/2/000016_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0007339908042922616, + "psnr": 31.34716796875, + "lpips": 0.009938481263816357 + }, + "frames_100_200": { + "mse": 0.004234290681779385, + "psnr": 27.926925659179688, + "lpips": 0.039519838988780975 + }, + "frames_200_300": { + "mse": 0.004052239470183849, + "psnr": 26.160995483398438, + "lpips": 0.039715565741062164 + }, + "frames_300_400": { + "mse": 0.004018112551420927, + "psnr": 25.307069778442383, + "lpips": 0.04146423190832138 + }, + "frames_400_500": { + "mse": 0.010619823820888996, + "psnr": 20.906478881835938, + "lpips": 0.0942152887582779 + }, + "overall": { + "mse": 0.004731691442430019, + "psnr": 26.329729080200195, + "lpips": 0.04497068375349045 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000142_test__val_plains__2__000020/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000142_test__val_plains__2__000020/case.json new file mode 100644 index 0000000000000000000000000000000000000000..6aab41c120708808c42096743e70f0383f9d5ed1 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000142_test__val_plains__2__000020/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "142", + "global_test_index": "142", + "dataset_index": "422", + "relative_path": "test/val_plains/2/000020.mp4", + "clip_id": "test/val_plains/2/000020", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000142, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.50500138907228, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000142_test__val_plains__2__000020/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000142_test__val_plains__2__000020/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..f26cb687abcaec96b8cf3c835765c6562e8a9796 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000142_test__val_plains__2__000020/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0d058adfd345b1cb49313c417155ef95a52754e0dea85996301af73e617f4d00 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000142_test__val_plains__2__000020/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000142_test__val_plains__2__000020/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..5c4cac4dc339abdd2db36437b9f89997e93ad370 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000142_test__val_plains__2__000020/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "142", + "global_test_index": "142", + "dataset_index": "422", + "relative_path": "test/val_plains/2/000020.mp4", + "clip_id": "test/val_plains/2/000020", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000142, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_plains/2/000020_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.005179745610803366, + "psnr": 23.393457412719727, + "lpips": 0.06411735713481903 + }, + "frames_100_200": { + "mse": 0.012120652012526989, + "psnr": 19.923017501831055, + "lpips": 0.2423017919063568 + }, + "frames_200_300": { + "mse": 0.013727474957704544, + "psnr": 20.06647491455078, + "lpips": 0.20633307099342346 + }, + "frames_300_400": { + "mse": 0.012624633498489857, + "psnr": 20.136577606201172, + "lpips": 0.13318783044815063 + }, + "frames_400_500": { + "mse": 0.004954768810421228, + "psnr": 24.394250869750977, + "lpips": 0.0685715302824974 + }, + "overall": { + "mse": 0.009721454232931137, + "psnr": 21.582754135131836, + "lpips": 0.14290231466293335 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000143_test__val_desert__0__000008/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000143_test__val_desert__0__000008/case.json new file mode 100644 index 0000000000000000000000000000000000000000..af419084a845a50a8a2afc56e138438565852bd6 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000143_test__val_desert__0__000008/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "143", + "global_test_index": "143", + "dataset_index": "20", + "relative_path": "test/val_desert/0/000008.mp4", + "clip_id": "test/val_desert/0/000008", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000143, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.14481051301118, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000143_test__val_desert__0__000008/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000143_test__val_desert__0__000008/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..cdd2bd5d8dd9694f16440087eb834ec4522a5984 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000143_test__val_desert__0__000008/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:31c692f539ef21a6b336966da68a38fc738b69003ca43345dd9e46a46b633b60 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000143_test__val_desert__0__000008/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000143_test__val_desert__0__000008/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..1dbb29478f335e8f21eae9e00065608586d3de82 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000143_test__val_desert__0__000008/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "143", + "global_test_index": "143", + "dataset_index": "20", + "relative_path": "test/val_desert/0/000008.mp4", + "clip_id": "test/val_desert/0/000008", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000143, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_desert/0/000008_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0021608571987599134, + "psnr": 29.054771423339844, + "lpips": 0.03201935067772865 + }, + "frames_100_200": { + "mse": 0.009991553612053394, + "psnr": 20.40047836303711, + "lpips": 0.18706092238426208 + }, + "frames_200_300": { + "mse": 0.009414276108145714, + "psnr": 20.423229217529297, + "lpips": 0.2686465382575989 + }, + "frames_300_400": { + "mse": 0.010549664497375488, + "psnr": 19.919300079345703, + "lpips": 0.18018601834774017 + }, + "frames_400_500": { + "mse": 0.0025424158666282892, + "psnr": 28.081045150756836, + "lpips": 0.03402670472860336 + }, + "overall": { + "mse": 0.006931753363460302, + "psnr": 23.575763702392578, + "lpips": 0.1403879076242447 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000144_test__val_savanna__0__000000/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000144_test__val_savanna__0__000000/case.json new file mode 100644 index 0000000000000000000000000000000000000000..0413d974192b5f9915d45b1d4ede9c46f2922bf7 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000144_test__val_savanna__0__000000/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "144", + "global_test_index": "144", + "dataset_index": "130", + "relative_path": "test/val_savanna/0/000000.mp4", + "clip_id": "test/val_savanna/0/000000", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000144, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.06946055300068, + "peak_generation_vram_bytes": 107565475840, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000144_test__val_savanna__0__000000/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000144_test__val_savanna__0__000000/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..21326d815454b1642c94d448bbcddefb156f1bcd --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000144_test__val_savanna__0__000000/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:53d60ba5ea5d2adb18167d7228887dd910887d44895a0e292f4148068e66d036 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000144_test__val_savanna__0__000000/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000144_test__val_savanna__0__000000/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..1c96a7f144623f713802e2b59dd38777ebc60abb --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000144_test__val_savanna__0__000000/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "144", + "global_test_index": "144", + "dataset_index": "130", + "relative_path": "test/val_savanna/0/000000.mp4", + "clip_id": "test/val_savanna/0/000000", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000144, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_savanna/0/000000_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.01432833168655634, + "psnr": 18.81938362121582, + "lpips": 0.1847548633813858 + }, + "frames_100_200": { + "mse": 0.02096223644912243, + "psnr": 16.971940994262695, + "lpips": 0.1620934158563614 + }, + "frames_200_300": { + "mse": 0.022076774388551712, + "psnr": 18.011005401611328, + "lpips": 0.15795347094535828 + }, + "frames_300_400": { + "mse": 0.033116985112428665, + "psnr": 15.83737564086914, + "lpips": 0.18332207202911377 + }, + "frames_400_500": { + "mse": 0.05101054534316063, + "psnr": 13.004201889038086, + "lpips": 0.25339677929878235 + }, + "overall": { + "mse": 0.02829897589981556, + "psnr": 16.52878189086914, + "lpips": 0.1883041262626648 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000145_test__val_iceplains__1__000020/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000145_test__val_iceplains__1__000020/case.json new file mode 100644 index 0000000000000000000000000000000000000000..09e48d481a5bf6c2c2b3f008ab0618b34ddc8671 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000145_test__val_iceplains__1__000020/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "145", + "global_test_index": "145", + "dataset_index": "296", + "relative_path": "test/val_iceplains/1/000020.mp4", + "clip_id": "test/val_iceplains/1/000020", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000145, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.3877320329193, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000145_test__val_iceplains__1__000020/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000145_test__val_iceplains__1__000020/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..182291fbb24779783b73f335f0d2b6fa11d5063c --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000145_test__val_iceplains__1__000020/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e776d9bf97a82407be97a23b3d35cf9430a75af37a48553b997b812ee05ae0fe +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000145_test__val_iceplains__1__000020/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000145_test__val_iceplains__1__000020/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..11ca58f3417b2a313a0d93b0e5a52285a2a7d03e --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000145_test__val_iceplains__1__000020/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "145", + "global_test_index": "145", + "dataset_index": "296", + "relative_path": "test/val_iceplains/1/000020.mp4", + "clip_id": "test/val_iceplains/1/000020", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000145, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/1/000020_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.018520131707191467, + "psnr": 17.437271118164062, + "lpips": 0.1893509328365326 + }, + "frames_100_200": { + "mse": 0.020824110135436058, + "psnr": 17.064260482788086, + "lpips": 0.19231382012367249 + }, + "frames_200_300": { + "mse": 0.05233718454837799, + "psnr": 12.990819931030273, + "lpips": 0.3208114504814148 + }, + "frames_300_400": { + "mse": 0.12671983242034912, + "psnr": 9.333992004394531, + "lpips": 0.5145698189735413 + }, + "frames_400_500": { + "mse": 0.10622066259384155, + "psnr": 9.9404935836792, + "lpips": 0.4689769744873047 + }, + "overall": { + "mse": 0.06492438167333603, + "psnr": 13.35336685180664, + "lpips": 0.33720463514328003 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000146_test__val_iceplains__3__000013/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000146_test__val_iceplains__3__000013/case.json new file mode 100644 index 0000000000000000000000000000000000000000..6580b1127d390a6952e11bd865f320c51b3c6917 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000146_test__val_iceplains__3__000013/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "146", + "global_test_index": "146", + "dataset_index": "278", + "relative_path": "test/val_iceplains/3/000013.mp4", + "clip_id": "test/val_iceplains/3/000013", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000146, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.32581361103803, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000146_test__val_iceplains__3__000013/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000146_test__val_iceplains__3__000013/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..afca14a77df605ba43324797c4da6fee2b51f706 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000146_test__val_iceplains__3__000013/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7e2c429efcdbc57d18065e699a90faf9ae431615169a958ed00009af212f3830 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000146_test__val_iceplains__3__000013/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000146_test__val_iceplains__3__000013/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..b94569de5c0e155d3d57abc1a6135e1897d07689 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000146_test__val_iceplains__3__000013/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "146", + "global_test_index": "146", + "dataset_index": "278", + "relative_path": "test/val_iceplains/3/000013.mp4", + "clip_id": "test/val_iceplains/3/000013", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000146, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/3/000013_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.017529720440506935, + "psnr": 18.878185272216797, + "lpips": 0.0991867184638977 + }, + "frames_100_200": { + "mse": 0.03677130863070488, + "psnr": 14.85781478881836, + "lpips": 0.22378647327423096 + }, + "frames_200_300": { + "mse": 0.09471405297517776, + "psnr": 10.633940696716309, + "lpips": 0.43479904532432556 + }, + "frames_300_400": { + "mse": 0.04496574401855469, + "psnr": 22.34035873413086, + "lpips": 0.221080482006073 + }, + "frames_400_500": { + "mse": 0.0009007162880152464, + "psnr": 30.556917190551758, + "lpips": 0.013845060020685196 + }, + "overall": { + "mse": 0.038976311683654785, + "psnr": 19.45344352722168, + "lpips": 0.19853955507278442 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000147_test__val_iceplains__0__000009/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000147_test__val_iceplains__0__000009/case.json new file mode 100644 index 0000000000000000000000000000000000000000..64e0e881033798d10686734f1e42478353423185 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000147_test__val_iceplains__0__000009/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "147", + "global_test_index": "147", + "dataset_index": "269", + "relative_path": "test/val_iceplains/0/000009.mp4", + "clip_id": "test/val_iceplains/0/000009", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000147, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.74690214102156, + "peak_generation_vram_bytes": 107565475840, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000147_test__val_iceplains__0__000009/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000147_test__val_iceplains__0__000009/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..821c1f6c667df0f0f22cf4873bcf98eb7803bb48 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000147_test__val_iceplains__0__000009/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:35799e503a410cbe6dcbc2ba5267c709f108b10cccbdaf6c8e3cb69e9d032768 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000147_test__val_iceplains__0__000009/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000147_test__val_iceplains__0__000009/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..ce4bd69af8a1d2ca9ab7ffe5504883b95152adcd --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000147_test__val_iceplains__0__000009/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "147", + "global_test_index": "147", + "dataset_index": "269", + "relative_path": "test/val_iceplains/0/000009.mp4", + "clip_id": "test/val_iceplains/0/000009", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000147, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/0/000009_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.006996840238571167, + "psnr": 21.609840393066406, + "lpips": 0.0959991067647934 + }, + "frames_100_200": { + "mse": 0.017328696325421333, + "psnr": 18.273849487304688, + "lpips": 0.20317289233207703 + }, + "frames_200_300": { + "mse": 0.015211950987577438, + "psnr": 19.46734046936035, + "lpips": 0.11414287239313126 + }, + "frames_300_400": { + "mse": 0.009908264502882957, + "psnr": 21.438011169433594, + "lpips": 0.06699898838996887 + }, + "frames_400_500": { + "mse": 0.007745485287159681, + "psnr": 22.149085998535156, + "lpips": 0.06910018622875214 + }, + "overall": { + "mse": 0.011438246816396713, + "psnr": 20.587627410888672, + "lpips": 0.10988280922174454 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000148_test__val_iceplains__0__000028/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000148_test__val_iceplains__0__000028/case.json new file mode 100644 index 0000000000000000000000000000000000000000..03dba479dd6d17694a0204319ded0ea31375e16c --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000148_test__val_iceplains__0__000028/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "148", + "global_test_index": "148", + "dataset_index": "314", + "relative_path": "test/val_iceplains/0/000028.mp4", + "clip_id": "test/val_iceplains/0/000028", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000148, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.3261045180261, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000148_test__val_iceplains__0__000028/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000148_test__val_iceplains__0__000028/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..75e6ae008e0c5c72616c6885f13964af1ca59948 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000148_test__val_iceplains__0__000028/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:90b1e932d515bb62947dbd128365c9ec6a8b3173e9cb72dc117961035c767cd8 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000148_test__val_iceplains__0__000028/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000148_test__val_iceplains__0__000028/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..be74919c76b5dbf348f4ccfad91be8941e6dab5d --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000148_test__val_iceplains__0__000028/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "148", + "global_test_index": "148", + "dataset_index": "314", + "relative_path": "test/val_iceplains/0/000028.mp4", + "clip_id": "test/val_iceplains/0/000028", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000148, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/0/000028_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0007885827217251062, + "psnr": 31.050010681152344, + "lpips": 0.010359418578445911 + }, + "frames_100_200": { + "mse": 0.0007732067024335265, + "psnr": 31.123899459838867, + "lpips": 0.009723990224301815 + }, + "frames_200_300": { + "mse": 0.0023505049757659435, + "psnr": 29.667678833007812, + "lpips": 0.022091791033744812 + }, + "frames_300_400": { + "mse": 0.05597124993801117, + "psnr": 12.895293235778809, + "lpips": 0.3342854976654053 + }, + "frames_400_500": { + "mse": 0.09654715657234192, + "psnr": 10.320138931274414, + "lpips": 0.4615365266799927 + }, + "overall": { + "mse": 0.03128613904118538, + "psnr": 23.011404037475586, + "lpips": 0.16759942471981049 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000149_test__val_iceplains__0__000014/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000149_test__val_iceplains__0__000014/case.json new file mode 100644 index 0000000000000000000000000000000000000000..0c3463c53a4fb2bcc26eea362bedd3ecf4d317c3 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000149_test__val_iceplains__0__000014/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "149", + "global_test_index": "149", + "dataset_index": "281", + "relative_path": "test/val_iceplains/0/000014.mp4", + "clip_id": "test/val_iceplains/0/000014", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000149, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 226.0105718289269, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000149_test__val_iceplains__0__000014/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000149_test__val_iceplains__0__000014/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..60f28e893795c7a2e89ed715d3568fb60cb4479c --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000149_test__val_iceplains__0__000014/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c0d572245b4f4efef916dd878f9eb9f9414d71e5aad50ad3171bd389e3b6e432 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000149_test__val_iceplains__0__000014/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000149_test__val_iceplains__0__000014/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..75d3abec747d6109ec34e250a0130505fa02a747 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000149_test__val_iceplains__0__000014/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "149", + "global_test_index": "149", + "dataset_index": "281", + "relative_path": "test/val_iceplains/0/000014.mp4", + "clip_id": "test/val_iceplains/0/000014", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000149, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/0/000014_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.04739540442824364, + "psnr": 15.685415267944336, + "lpips": 0.21638593077659607 + }, + "frames_100_200": { + "mse": 0.08712805807590485, + "psnr": 10.938648223876953, + "lpips": 0.3583805561065674 + }, + "frames_200_300": { + "mse": 0.12507909536361694, + "psnr": 9.387578010559082, + "lpips": 0.46924901008605957 + }, + "frames_300_400": { + "mse": 0.08562373369932175, + "psnr": 11.049691200256348, + "lpips": 0.35876065492630005 + }, + "frames_400_500": { + "mse": 0.017072414979338646, + "psnr": 18.8704776763916, + "lpips": 0.12838011980056763 + }, + "overall": { + "mse": 0.07245973497629166, + "psnr": 13.186362266540527, + "lpips": 0.3062312602996826 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000150_test__val_savanna__0__000008/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000150_test__val_savanna__0__000008/case.json new file mode 100644 index 0000000000000000000000000000000000000000..5784fe543d14b018cdb6535074a9a9c6e696f41d --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000150_test__val_savanna__0__000008/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "150", + "global_test_index": "150", + "dataset_index": "149", + "relative_path": "test/val_savanna/0/000008.mp4", + "clip_id": "test/val_savanna/0/000008", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000150, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.7526781199267, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000150_test__val_savanna__0__000008/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000150_test__val_savanna__0__000008/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..14301caf729de29ba61727224734545c0b0d569b --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000150_test__val_savanna__0__000008/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:38d54795434915836c50e61cedb7f366a4aa101fc9d1c2d1a20c7ad6913926f7 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000150_test__val_savanna__0__000008/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000150_test__val_savanna__0__000008/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..761e8b15f10d7ade75973532e5b68e294db473a6 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000150_test__val_savanna__0__000008/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "150", + "global_test_index": "150", + "dataset_index": "149", + "relative_path": "test/val_savanna/0/000008.mp4", + "clip_id": "test/val_savanna/0/000008", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000150, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_savanna/0/000008_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.01636192575097084, + "psnr": 18.004098892211914, + "lpips": 0.1549384891986847 + }, + "frames_100_200": { + "mse": 0.019252950325608253, + "psnr": 17.360183715820312, + "lpips": 0.15803487598896027 + }, + "frames_200_300": { + "mse": 0.026423189789056778, + "psnr": 16.051828384399414, + "lpips": 0.3776746392250061 + }, + "frames_300_400": { + "mse": 0.007680749520659447, + "psnr": 21.295442581176758, + "lpips": 0.39117467403411865 + }, + "frames_400_500": { + "mse": 0.023196540772914886, + "psnr": 16.734790802001953, + "lpips": 0.45798367261886597 + }, + "overall": { + "mse": 0.01858307048678398, + "psnr": 17.889266967773438, + "lpips": 0.30796128511428833 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000151_test__val_savanna__0__000024/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000151_test__val_savanna__0__000024/case.json new file mode 100644 index 0000000000000000000000000000000000000000..aba7d363b6cbb408d644cf72c09b5dd8935e8202 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000151_test__val_savanna__0__000024/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "151", + "global_test_index": "151", + "dataset_index": "184", + "relative_path": "test/val_savanna/0/000024.mp4", + "clip_id": "test/val_savanna/0/000024", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000151, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.67810513090808, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000151_test__val_savanna__0__000024/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000151_test__val_savanna__0__000024/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..5fda8a33e8b04a6ef0c3327b10935395438ad175 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000151_test__val_savanna__0__000024/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3973e572bfba0452c690d2a12d25866b939a1407b3d64f71731f32cdb9ac7fb6 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000151_test__val_savanna__0__000024/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000151_test__val_savanna__0__000024/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..5a507668b09e0d18037b9bcac51ab6161e181f73 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000151_test__val_savanna__0__000024/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "151", + "global_test_index": "151", + "dataset_index": "184", + "relative_path": "test/val_savanna/0/000024.mp4", + "clip_id": "test/val_savanna/0/000024", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000151, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_savanna/0/000024_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.02016874961555004, + "psnr": 18.598539352416992, + "lpips": 0.3252335488796234 + }, + "frames_100_200": { + "mse": 0.02743268944323063, + "psnr": 16.873592376708984, + "lpips": 0.37241190671920776 + }, + "frames_200_300": { + "mse": 0.035358354449272156, + "psnr": 16.305774688720703, + "lpips": 0.39203834533691406 + }, + "frames_300_400": { + "mse": 0.012590973637998104, + "psnr": 19.652446746826172, + "lpips": 0.37331831455230713 + }, + "frames_400_500": { + "mse": 0.00914668757468462, + "psnr": 20.669275283813477, + "lpips": 0.41927146911621094 + }, + "overall": { + "mse": 0.02093948982656002, + "psnr": 18.419925689697266, + "lpips": 0.3764547109603882 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000152_test__val_plains__1__000014/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000152_test__val_plains__1__000014/case.json new file mode 100644 index 0000000000000000000000000000000000000000..ee0d77e519d687bc689ac8708fb46f77fbd3f9fa --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000152_test__val_plains__1__000014/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "152", + "global_test_index": "152", + "dataset_index": "407", + "relative_path": "test/val_plains/1/000014.mp4", + "clip_id": "test/val_plains/1/000014", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000152, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 223.6361303080339, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000152_test__val_plains__1__000014/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000152_test__val_plains__1__000014/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..9c165eea8b0804a3f2aa82dc9c039872ddebd0d6 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000152_test__val_plains__1__000014/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:43644449da8290ea99a62305c5a7fbd507ac07b170a7b2f3d262d25b3323375f +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000152_test__val_plains__1__000014/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000152_test__val_plains__1__000014/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..8c201cb8c00d72257fc41af9a8c43fcdde7a44ed --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000152_test__val_plains__1__000014/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "152", + "global_test_index": "152", + "dataset_index": "407", + "relative_path": "test/val_plains/1/000014.mp4", + "clip_id": "test/val_plains/1/000014", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000152, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_plains/1/000014_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.004154089838266373, + "psnr": 25.98583221435547, + "lpips": 0.03376161307096481 + }, + "frames_100_200": { + "mse": 0.0008194279507733881, + "psnr": 30.87285804748535, + "lpips": 0.01385972835123539 + }, + "frames_200_300": { + "mse": 0.0009062540484592319, + "psnr": 30.430971145629883, + "lpips": 0.01645892858505249 + }, + "frames_300_400": { + "mse": 0.016795752570033073, + "psnr": 18.671642303466797, + "lpips": 0.206385537981987 + }, + "frames_400_500": { + "mse": 0.020015614107251167, + "psnr": 17.561010360717773, + "lpips": 0.19880813360214233 + }, + "overall": { + "mse": 0.008538228459656239, + "psnr": 24.704463958740234, + "lpips": 0.09385479986667633 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000153_test__val_iceplains__2__000023/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000153_test__val_iceplains__2__000023/case.json new file mode 100644 index 0000000000000000000000000000000000000000..8aca77aa356ede2abd13ae0362ed592c648eb2b5 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000153_test__val_iceplains__2__000023/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "153", + "global_test_index": "153", + "dataset_index": "304", + "relative_path": "test/val_iceplains/2/000023.mp4", + "clip_id": "test/val_iceplains/2/000023", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000153, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.0451825130731, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000153_test__val_iceplains__2__000023/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000153_test__val_iceplains__2__000023/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..f4a3074756eb500dfe28c8b6800751be426fd425 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000153_test__val_iceplains__2__000023/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a3cbebc07674fdbf027a247c65e177b31880a954d14445ad4a47b66c7b8e1685 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000153_test__val_iceplains__2__000023/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000153_test__val_iceplains__2__000023/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..efd48d31365e86f93a2b224903e10b3e1a416cf6 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000153_test__val_iceplains__2__000023/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "153", + "global_test_index": "153", + "dataset_index": "304", + "relative_path": "test/val_iceplains/2/000023.mp4", + "clip_id": "test/val_iceplains/2/000023", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000153, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/2/000023_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.001208552741445601, + "psnr": 29.956308364868164, + "lpips": 0.013506072573363781 + }, + "frames_100_200": { + "mse": 0.006902288645505905, + "psnr": 22.614727020263672, + "lpips": 0.0661943331360817 + }, + "frames_200_300": { + "mse": 0.027931327000260353, + "psnr": 16.237600326538086, + "lpips": 0.22020532190799713 + }, + "frames_300_400": { + "mse": 0.050451572984457016, + "psnr": 13.465534210205078, + "lpips": 0.35356396436691284 + }, + "frames_400_500": { + "mse": 0.05939914658665657, + "psnr": 13.074418067932129, + "lpips": 0.44269701838493347 + }, + "overall": { + "mse": 0.029178574681282043, + "psnr": 19.069717407226562, + "lpips": 0.21923333406448364 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000154_test__val_savanna__3__000037/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000154_test__val_savanna__3__000037/case.json new file mode 100644 index 0000000000000000000000000000000000000000..1416a5b4a010891e0c21d9c599f369717719f5ce --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000154_test__val_savanna__3__000037/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "154", + "global_test_index": "154", + "dataset_index": "208", + "relative_path": "test/val_savanna/3/000037.mp4", + "clip_id": "test/val_savanna/3/000037", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000154, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.41810397803783, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000154_test__val_savanna__3__000037/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000154_test__val_savanna__3__000037/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..240cd817b10413d05c1cdb0c19b6515963121943 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000154_test__val_savanna__3__000037/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4a9618fbd7873e90aad002756df04a3572333eb9859bd0e0aae5df685b55ecbf +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000154_test__val_savanna__3__000037/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000154_test__val_savanna__3__000037/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..627f346cefb796b56e568eb85aaedafea51fd828 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000154_test__val_savanna__3__000037/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "154", + "global_test_index": "154", + "dataset_index": "208", + "relative_path": "test/val_savanna/3/000037.mp4", + "clip_id": "test/val_savanna/3/000037", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000154, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_savanna/3/000037_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.012125239707529545, + "psnr": 19.92957305908203, + "lpips": 0.2845328450202942 + }, + "frames_100_200": { + "mse": 0.006556785199791193, + "psnr": 22.380020141601562, + "lpips": 0.40014877915382385 + }, + "frames_200_300": { + "mse": 0.0068070171400904655, + "psnr": 21.82123565673828, + "lpips": 0.4008292853832245 + }, + "frames_300_400": { + "mse": 0.007113582920283079, + "psnr": 21.552873611450195, + "lpips": 0.3991297781467438 + }, + "frames_400_500": { + "mse": 0.01006106287240982, + "psnr": 20.391435623168945, + "lpips": 0.43605759739875793 + }, + "overall": { + "mse": 0.008532737381756306, + "psnr": 21.215024948120117, + "lpips": 0.38413965702056885 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000155_test__val_savanna__3__000023/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000155_test__val_savanna__3__000023/case.json new file mode 100644 index 0000000000000000000000000000000000000000..c44b723fa49c430131e7e028e24ec0d6e983a064 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000155_test__val_savanna__3__000023/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "155", + "global_test_index": "155", + "dataset_index": "182", + "relative_path": "test/val_savanna/3/000023.mp4", + "clip_id": "test/val_savanna/3/000023", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000155, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.7775760430377, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000155_test__val_savanna__3__000023/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000155_test__val_savanna__3__000023/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..360882fee90f29c9281d73235c4e93b2fcd37948 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000155_test__val_savanna__3__000023/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:afead7e507cda9999c70ccf0b0d519ccb104f1fc1a18de195d5773344d1acb67 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000155_test__val_savanna__3__000023/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000155_test__val_savanna__3__000023/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..4092bd3cdd2df48ee04f743d3188c43569a592ba --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000155_test__val_savanna__3__000023/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "155", + "global_test_index": "155", + "dataset_index": "182", + "relative_path": "test/val_savanna/3/000023.mp4", + "clip_id": "test/val_savanna/3/000023", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000155, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_savanna/3/000023_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.003888083389028907, + "psnr": 24.230520248413086, + "lpips": 0.28781530261039734 + }, + "frames_100_200": { + "mse": 0.006445824634283781, + "psnr": 21.948415756225586, + "lpips": 0.36910000443458557 + }, + "frames_200_300": { + "mse": 0.005860466510057449, + "psnr": 22.387897491455078, + "lpips": 0.3716805577278137 + }, + "frames_300_400": { + "mse": 0.005950090941041708, + "psnr": 22.350486755371094, + "lpips": 0.38043221831321716 + }, + "frames_400_500": { + "mse": 0.007178816013038158, + "psnr": 21.498716354370117, + "lpips": 0.4050551652908325 + }, + "overall": { + "mse": 0.005864656530320644, + "psnr": 22.48320960998535, + "lpips": 0.3628166615962982 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000156_test__val_plains__1__000009/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000156_test__val_plains__1__000009/case.json new file mode 100644 index 0000000000000000000000000000000000000000..ff1099cb445bea47da28566852cb77cf331487db --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000156_test__val_plains__1__000009/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "156", + "global_test_index": "156", + "dataset_index": "394", + "relative_path": "test/val_plains/1/000009.mp4", + "clip_id": "test/val_plains/1/000009", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000156, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 223.77454236801714, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000156_test__val_plains__1__000009/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000156_test__val_plains__1__000009/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..9cb9ef476015d46b33500d59f1dbfed666a8abce --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000156_test__val_plains__1__000009/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d3c51a4f765a42e5ec8e236beee33fd65b0f21e7ced0fe8567f6db25e216f35c +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000156_test__val_plains__1__000009/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000156_test__val_plains__1__000009/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..f19d722d1af35627bf44cfd8d36f5a0f66936643 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000156_test__val_plains__1__000009/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "156", + "global_test_index": "156", + "dataset_index": "394", + "relative_path": "test/val_plains/1/000009.mp4", + "clip_id": "test/val_plains/1/000009", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000156, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_plains/1/000009_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0019891317933797836, + "psnr": 27.246416091918945, + "lpips": 0.17555823922157288 + }, + "frames_100_200": { + "mse": 0.004146236460655928, + "psnr": 24.576278686523438, + "lpips": 0.26506486535072327 + }, + "frames_200_300": { + "mse": 0.003526050830259919, + "psnr": 24.609071731567383, + "lpips": 0.26947087049484253 + }, + "frames_300_400": { + "mse": 0.003443977329879999, + "psnr": 24.93905258178711, + "lpips": 0.239695206284523 + }, + "frames_400_500": { + "mse": 0.002820749534294009, + "psnr": 25.558700561523438, + "lpips": 0.24021081626415253 + }, + "overall": { + "mse": 0.003185229143127799, + "psnr": 25.385906219482422, + "lpips": 0.23800000548362732 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000157_test__val_desert__0__000006/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000157_test__val_desert__0__000006/case.json new file mode 100644 index 0000000000000000000000000000000000000000..ca38c75ae09baef4f26a67b64519eb0ca36bf4ca --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000157_test__val_desert__0__000006/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "157", + "global_test_index": "157", + "dataset_index": "14", + "relative_path": "test/val_desert/0/000006.mp4", + "clip_id": "test/val_desert/0/000006", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000157, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 226.07496935105883, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000157_test__val_desert__0__000006/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000157_test__val_desert__0__000006/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..4c57f903f85a1e8336db62f1b93f19bbbf7047e4 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000157_test__val_desert__0__000006/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c4ecc5027352ac277592f5a7ae246dcbe43cf0a0083074aae276d2d125451147 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000157_test__val_desert__0__000006/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000157_test__val_desert__0__000006/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..9ecfa9b088c4fa444b1546415909a006b4f3705c --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000157_test__val_desert__0__000006/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "157", + "global_test_index": "157", + "dataset_index": "14", + "relative_path": "test/val_desert/0/000006.mp4", + "clip_id": "test/val_desert/0/000006", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000157, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_desert/0/000006_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.00341719388961792, + "psnr": 24.88055992126465, + "lpips": 0.07364889234304428 + }, + "frames_100_200": { + "mse": 0.004918004386126995, + "psnr": 23.128835678100586, + "lpips": 0.11859641224145889 + }, + "frames_200_300": { + "mse": 0.0045169866643846035, + "psnr": 23.6483211517334, + "lpips": 0.13063378632068634 + }, + "frames_300_400": { + "mse": 0.007304962258785963, + "psnr": 22.347393035888672, + "lpips": 0.16555781662464142 + }, + "frames_400_500": { + "mse": 0.00849438738077879, + "psnr": 20.86764144897461, + "lpips": 0.1825994998216629 + }, + "overall": { + "mse": 0.005730306264013052, + "psnr": 22.97454833984375, + "lpips": 0.13420727849006653 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000158_test__val_desert__3__000021/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000158_test__val_desert__3__000021/case.json new file mode 100644 index 0000000000000000000000000000000000000000..d1171a9c2ea34f5c7685f5f013e7cb4c03eee4a2 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000158_test__val_desert__3__000021/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "158", + "global_test_index": "158", + "dataset_index": "55", + "relative_path": "test/val_desert/3/000021.mp4", + "clip_id": "test/val_desert/3/000021", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000158, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.49190915504005, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000158_test__val_desert__3__000021/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000158_test__val_desert__3__000021/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..faf4e25f24c1d9780bc04e343ac02416e286e223 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000158_test__val_desert__3__000021/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f648e2da4520dbddbe4ffee1a077b136c1289742317db0825d68d0f715a49561 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000158_test__val_desert__3__000021/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000158_test__val_desert__3__000021/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..5f734b46177b2881e28de1044819305e36513d43 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000158_test__val_desert__3__000021/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "158", + "global_test_index": "158", + "dataset_index": "55", + "relative_path": "test/val_desert/3/000021.mp4", + "clip_id": "test/val_desert/3/000021", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000158, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_desert/3/000021_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0013348831562325358, + "psnr": 29.131452560424805, + "lpips": 0.020271457731723785 + }, + "frames_100_200": { + "mse": 0.0008992105722427368, + "psnr": 30.597665786743164, + "lpips": 0.011846187524497509 + }, + "frames_200_300": { + "mse": 0.003027664963155985, + "psnr": 28.21802520751953, + "lpips": 0.02849796786904335 + }, + "frames_300_400": { + "mse": 0.0008370288414880633, + "psnr": 30.78664779663086, + "lpips": 0.011146942153573036 + }, + "frames_400_500": { + "mse": 0.002388040767982602, + "psnr": 29.268051147460938, + "lpips": 0.019816748797893524 + }, + "overall": { + "mse": 0.0016973658930510283, + "psnr": 29.600366592407227, + "lpips": 0.018315861001610756 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000159_test__val_desert__2__000028/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000159_test__val_desert__2__000028/case.json new file mode 100644 index 0000000000000000000000000000000000000000..0a6f87ca492bd34d980c633cf8e01f4544aa9483 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000159_test__val_desert__2__000028/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "159", + "global_test_index": "159", + "dataset_index": "75", + "relative_path": "test/val_desert/2/000028.mp4", + "clip_id": "test/val_desert/2/000028", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000159, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.50328751595225, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000159_test__val_desert__2__000028/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000159_test__val_desert__2__000028/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..7c9ea31af99e45a7973e42841fc0e9d255acd75f --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000159_test__val_desert__2__000028/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d85fb6cef6a7119756ff6b2e850b2cc680ae38843625c2796c10e86a40955042 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000159_test__val_desert__2__000028/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000159_test__val_desert__2__000028/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..45db418c24e70b7eca18aad40e0d3013c44faa3c --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000159_test__val_desert__2__000028/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "159", + "global_test_index": "159", + "dataset_index": "75", + "relative_path": "test/val_desert/2/000028.mp4", + "clip_id": "test/val_desert/2/000028", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000159, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_desert/2/000028_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.011148875579237938, + "psnr": 19.70049476623535, + "lpips": 0.3619179427623749 + }, + "frames_100_200": { + "mse": 0.013112234883010387, + "psnr": 18.944795608520508, + "lpips": 0.38696715235710144 + }, + "frames_200_300": { + "mse": 0.04134311154484749, + "psnr": 14.73762321472168, + "lpips": 0.4491835832595825 + }, + "frames_300_400": { + "mse": 0.027485845610499382, + "psnr": 15.996743202209473, + "lpips": 0.3735336661338806 + }, + "frames_400_500": { + "mse": 0.013043110258877277, + "psnr": 19.831369400024414, + "lpips": 0.20952443778514862 + }, + "overall": { + "mse": 0.021226637065410614, + "psnr": 17.842205047607422, + "lpips": 0.3562253415584564 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000160_test__val_iceplains__1__000031/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000160_test__val_iceplains__1__000031/case.json new file mode 100644 index 0000000000000000000000000000000000000000..9eaf688da76a16e687d42ea8348a9125478dd855 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000160_test__val_iceplains__1__000031/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "160", + "global_test_index": "160", + "dataset_index": "319", + "relative_path": "test/val_iceplains/1/000031.mp4", + "clip_id": "test/val_iceplains/1/000031", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000160, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 223.4046555699315, + "peak_generation_vram_bytes": 107565475840, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000160_test__val_iceplains__1__000031/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000160_test__val_iceplains__1__000031/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..ff4bd322b79436413b909f9da97162d7bee39bdb --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000160_test__val_iceplains__1__000031/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e9a384beb4bc31be0611eb7e97e522b53c105697cd7d18e8b419e1c8961e2e79 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000160_test__val_iceplains__1__000031/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000160_test__val_iceplains__1__000031/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..4ced47e88dba37a41d11c1b3322b5c5ae276aeca --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000160_test__val_iceplains__1__000031/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "160", + "global_test_index": "160", + "dataset_index": "319", + "relative_path": "test/val_iceplains/1/000031.mp4", + "clip_id": "test/val_iceplains/1/000031", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000160, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/1/000031_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.013877329416573048, + "psnr": 20.097023010253906, + "lpips": 0.19726957380771637 + }, + "frames_100_200": { + "mse": 0.011188950389623642, + "psnr": 19.786588668823242, + "lpips": 0.25397196412086487 + }, + "frames_200_300": { + "mse": 0.029003873467445374, + "psnr": 17.220870971679688, + "lpips": 0.34528103470802307 + }, + "frames_300_400": { + "mse": 0.039648495614528656, + "psnr": 14.23535442352295, + "lpips": 0.3318473696708679 + }, + "frames_400_500": { + "mse": 0.043801236897706985, + "psnr": 13.690010070800781, + "lpips": 0.31728890538215637 + }, + "overall": { + "mse": 0.027503976598381996, + "psnr": 17.005971908569336, + "lpips": 0.289131760597229 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000161_test__val_desert__1__000016/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000161_test__val_desert__1__000016/case.json new file mode 100644 index 0000000000000000000000000000000000000000..2dda5e9eb664c128f002d8938fdfbbd37d4c7f2a --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000161_test__val_desert__1__000016/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "161", + "global_test_index": "161", + "dataset_index": "39", + "relative_path": "test/val_desert/1/000016.mp4", + "clip_id": "test/val_desert/1/000016", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000161, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.6792197310133, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000161_test__val_desert__1__000016/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000161_test__val_desert__1__000016/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..ef4b871ae901c604aa1f5aa643a432c7322a68fb --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000161_test__val_desert__1__000016/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3a0e45b8a1a449b6607d4ff6ec1246e4e408680877ea53e88b82223b209863e3 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000161_test__val_desert__1__000016/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000161_test__val_desert__1__000016/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..a10b014e6914d81031f22cdd4538bfabb35edf76 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000161_test__val_desert__1__000016/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "161", + "global_test_index": "161", + "dataset_index": "39", + "relative_path": "test/val_desert/1/000016.mp4", + "clip_id": "test/val_desert/1/000016", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000161, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_desert/1/000016_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.020943736657500267, + "psnr": 18.581317901611328, + "lpips": 0.2684013843536377 + }, + "frames_100_200": { + "mse": 0.01986922137439251, + "psnr": 18.325883865356445, + "lpips": 0.38177037239074707 + }, + "frames_200_300": { + "mse": 0.007824622094631195, + "psnr": 21.1066951751709, + "lpips": 0.4380839169025421 + }, + "frames_300_400": { + "mse": 0.008290417492389679, + "psnr": 20.866697311401367, + "lpips": 0.39785686135292053 + }, + "frames_400_500": { + "mse": 0.007510737050324678, + "psnr": 21.26839828491211, + "lpips": 0.39896470308303833 + }, + "overall": { + "mse": 0.012887747026979923, + "psnr": 20.02979850769043, + "lpips": 0.37701544165611267 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000162_test__val_plains__1__000019/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000162_test__val_plains__1__000019/case.json new file mode 100644 index 0000000000000000000000000000000000000000..af7729ebeca925c875433a2e3608a20ee680cfb4 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000162_test__val_plains__1__000019/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "162", + "global_test_index": "162", + "dataset_index": "418", + "relative_path": "test/val_plains/1/000019.mp4", + "clip_id": "test/val_plains/1/000019", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000162, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.63981890294235, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000162_test__val_plains__1__000019/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000162_test__val_plains__1__000019/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..0504d073b9258af6e162843d3b84e11e6406462a --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000162_test__val_plains__1__000019/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3948759c6e86a9d49f75c7f8443b648dcd6e35284746336c224922381e1eee08 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000162_test__val_plains__1__000019/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000162_test__val_plains__1__000019/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..d25ca8048857c905a5a6e0914b7ca68f1cb8a4f6 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000162_test__val_plains__1__000019/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "162", + "global_test_index": "162", + "dataset_index": "418", + "relative_path": "test/val_plains/1/000019.mp4", + "clip_id": "test/val_plains/1/000019", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000162, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_plains/1/000019_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.007762417662888765, + "psnr": 24.263172149658203, + "lpips": 0.08087268471717834 + }, + "frames_100_200": { + "mse": 0.007414294872432947, + "psnr": 22.503538131713867, + "lpips": 0.05996355414390564 + }, + "frames_200_300": { + "mse": 0.01024309266358614, + "psnr": 22.1127872467041, + "lpips": 0.11518069356679916 + }, + "frames_300_400": { + "mse": 0.014860306866466999, + "psnr": 18.391315460205078, + "lpips": 0.2985695004463196 + }, + "frames_400_500": { + "mse": 0.012824838049709797, + "psnr": 19.7665958404541, + "lpips": 0.22873742878437042 + }, + "overall": { + "mse": 0.01062099076807499, + "psnr": 21.407480239868164, + "lpips": 0.15666477382183075 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000163_test__val_plains__3__000007/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000163_test__val_plains__3__000007/case.json new file mode 100644 index 0000000000000000000000000000000000000000..283ec79ee010540f5adf0843040ebefde86fd247 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000163_test__val_plains__3__000007/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "163", + "global_test_index": "163", + "dataset_index": "389", + "relative_path": "test/val_plains/3/000007.mp4", + "clip_id": "test/val_plains/3/000007", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000163, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.73282831301913, + "peak_generation_vram_bytes": 107565475840, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000163_test__val_plains__3__000007/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000163_test__val_plains__3__000007/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..c92b300209ae6e1d5b377b5f89dafb97c64f1eb9 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000163_test__val_plains__3__000007/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:048ace71847ac140308a9e07bc226b3f5c19dbbbc9fa79617e83f22a1d25ba03 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000163_test__val_plains__3__000007/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000163_test__val_plains__3__000007/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..1d7429ee37d2d3cdb334c4bc11ddb6dd44386675 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000163_test__val_plains__3__000007/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "163", + "global_test_index": "163", + "dataset_index": "389", + "relative_path": "test/val_plains/3/000007.mp4", + "clip_id": "test/val_plains/3/000007", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000163, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_plains/3/000007_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.006144950166344643, + "psnr": 22.602977752685547, + "lpips": 0.25757575035095215 + }, + "frames_100_200": { + "mse": 0.006809709127992392, + "psnr": 21.841859817504883, + "lpips": 0.3618086874485016 + }, + "frames_200_300": { + "mse": 0.014300215989351273, + "psnr": 19.4311580657959, + "lpips": 0.4289647936820984 + }, + "frames_300_400": { + "mse": 0.022751130163669586, + "psnr": 16.840465545654297, + "lpips": 0.4086460769176483 + }, + "frames_400_500": { + "mse": 0.011271731927990913, + "psnr": 20.237756729125977, + "lpips": 0.37748953700065613 + }, + "overall": { + "mse": 0.012255547568202019, + "psnr": 20.19084358215332, + "lpips": 0.36689698696136475 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000164_test__val_plains__2__000014/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000164_test__val_plains__2__000014/case.json new file mode 100644 index 0000000000000000000000000000000000000000..bd57484b2fd9f18ae0e91a02dc250dc375b40ab0 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000164_test__val_plains__2__000014/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "164", + "global_test_index": "164", + "dataset_index": "409", + "relative_path": "test/val_plains/2/000014.mp4", + "clip_id": "test/val_plains/2/000014", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000164, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.42979060392827, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000164_test__val_plains__2__000014/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000164_test__val_plains__2__000014/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..3c8011f9d1f6f0bb6abf3288e4c26801f84ab400 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000164_test__val_plains__2__000014/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d4a3752ba091787d583c354747ddde845146e73935dbf4380db7b00d5cbb3a88 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000164_test__val_plains__2__000014/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000164_test__val_plains__2__000014/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..810296dc59e0365ee225ccc81e05300633938cba --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000164_test__val_plains__2__000014/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "164", + "global_test_index": "164", + "dataset_index": "409", + "relative_path": "test/val_plains/2/000014.mp4", + "clip_id": "test/val_plains/2/000014", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000164, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_plains/2/000014_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.02146289311349392, + "psnr": 18.341766357421875, + "lpips": 0.32576969265937805 + }, + "frames_100_200": { + "mse": 0.010879461653530598, + "psnr": 20.739595413208008, + "lpips": 0.212301105260849 + }, + "frames_200_300": { + "mse": 0.005692845210433006, + "psnr": 22.55858612060547, + "lpips": 0.11756019294261932 + }, + "frames_300_400": { + "mse": 0.004374367184937, + "psnr": 23.782800674438477, + "lpips": 0.14108823239803314 + }, + "frames_400_500": { + "mse": 0.005384379532188177, + "psnr": 22.75505828857422, + "lpips": 0.2199276089668274 + }, + "overall": { + "mse": 0.009558789432048798, + "psnr": 21.635562896728516, + "lpips": 0.20332935452461243 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000165_test__val_iceplains__2__000038/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000165_test__val_iceplains__2__000038/case.json new file mode 100644 index 0000000000000000000000000000000000000000..fa0af857225dfb4be17070529532ef8ef581daaf --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000165_test__val_iceplains__2__000038/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "165", + "global_test_index": "165", + "dataset_index": "344", + "relative_path": "test/val_iceplains/2/000038.mp4", + "clip_id": "test/val_iceplains/2/000038", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000165, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.51617478800472, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000165_test__val_iceplains__2__000038/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000165_test__val_iceplains__2__000038/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..e9412c36ba1dee92e2c06cb3ae1ff89e6a7d251b --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000165_test__val_iceplains__2__000038/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:edcf98738a21c3fb96197c81063075296d5693f87ee39ab903ae9e5ed63a36ac +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000165_test__val_iceplains__2__000038/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000165_test__val_iceplains__2__000038/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..4e1a0bbfb1f08501ebc55286c4c33e34bed6f37f --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000165_test__val_iceplains__2__000038/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "165", + "global_test_index": "165", + "dataset_index": "344", + "relative_path": "test/val_iceplains/2/000038.mp4", + "clip_id": "test/val_iceplains/2/000038", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000165, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/2/000038_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.030555756762623787, + "psnr": 15.550201416015625, + "lpips": 0.2515853941440582 + }, + "frames_100_200": { + "mse": 0.03187628835439682, + "psnr": 16.576932907104492, + "lpips": 0.2817156910896301 + }, + "frames_200_300": { + "mse": 0.04216410964727402, + "psnr": 15.887704849243164, + "lpips": 0.3050660490989685 + }, + "frames_300_400": { + "mse": 0.006120733451098204, + "psnr": 23.239795684814453, + "lpips": 0.17241375148296356 + }, + "frames_400_500": { + "mse": 0.030842822045087814, + "psnr": 16.699542999267578, + "lpips": 0.20323333144187927 + }, + "overall": { + "mse": 0.028311941772699356, + "psnr": 17.590835571289062, + "lpips": 0.24280284345149994 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000166_test__val_plains__2__000009/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000166_test__val_plains__2__000009/case.json new file mode 100644 index 0000000000000000000000000000000000000000..b0d688697f6d95c888ae20f520bbecb3b1c27dcc --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000166_test__val_plains__2__000009/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "166", + "global_test_index": "166", + "dataset_index": "396", + "relative_path": "test/val_plains/2/000009.mp4", + "clip_id": "test/val_plains/2/000009", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000166, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.7257591029629, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000166_test__val_plains__2__000009/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000166_test__val_plains__2__000009/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..59871d9f318887eef4f018c9f43e0294f4f6d316 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000166_test__val_plains__2__000009/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7a2336ae219f37bb2854ed4bff0f5d33492a39818947cebd8a796ac8f6bf8379 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000166_test__val_plains__2__000009/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000166_test__val_plains__2__000009/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..e6976ae37090afb9fe5050e4d10e2cba68d5bb77 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000166_test__val_plains__2__000009/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "166", + "global_test_index": "166", + "dataset_index": "396", + "relative_path": "test/val_plains/2/000009.mp4", + "clip_id": "test/val_plains/2/000009", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000166, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_plains/2/000009_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0025471255648881197, + "psnr": 26.280105590820312, + "lpips": 0.023878369480371475 + }, + "frames_100_200": { + "mse": 0.0033187714871019125, + "psnr": 26.166616439819336, + "lpips": 0.02930053323507309 + }, + "frames_200_300": { + "mse": 0.02041269838809967, + "psnr": 18.873777389526367, + "lpips": 0.1614265888929367 + }, + "frames_300_400": { + "mse": 0.020580600947141647, + "psnr": 17.835220336914062, + "lpips": 0.2747064232826233 + }, + "frames_400_500": { + "mse": 0.017723724246025085, + "psnr": 17.91985511779785, + "lpips": 0.2888321280479431 + }, + "overall": { + "mse": 0.012916583567857742, + "psnr": 21.41511344909668, + "lpips": 0.15562881529331207 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000167_test__val_desert__1__000026/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000167_test__val_desert__1__000026/case.json new file mode 100644 index 0000000000000000000000000000000000000000..b9f69ff6135f6f8765c99a002724bf2985014f77 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000167_test__val_desert__1__000026/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "167", + "global_test_index": "167", + "dataset_index": "67", + "relative_path": "test/val_desert/1/000026.mp4", + "clip_id": "test/val_desert/1/000026", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000167, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.240451724967, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000167_test__val_desert__1__000026/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000167_test__val_desert__1__000026/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..1a9ba44697361c371e281343fc48fa545c740389 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000167_test__val_desert__1__000026/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7c31c438387d8e41c0a62b64e1d2166c87c89be5c4e0e92302f689c8fca0def5 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000167_test__val_desert__1__000026/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000167_test__val_desert__1__000026/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..f4e6634d6d88e2f581175ab3f66e2541032d4294 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000167_test__val_desert__1__000026/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "167", + "global_test_index": "167", + "dataset_index": "67", + "relative_path": "test/val_desert/1/000026.mp4", + "clip_id": "test/val_desert/1/000026", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000167, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_desert/1/000026_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.007905388250946999, + "psnr": 21.1109561920166, + "lpips": 0.29017987847328186 + }, + "frames_100_200": { + "mse": 0.008967417292296886, + "psnr": 20.553504943847656, + "lpips": 0.3557288646697998 + }, + "frames_200_300": { + "mse": 0.03313088044524193, + "psnr": 16.58196449279785, + "lpips": 0.43567633628845215 + }, + "frames_300_400": { + "mse": 0.03509125858545303, + "psnr": 15.312836647033691, + "lpips": 0.39672207832336426 + }, + "frames_400_500": { + "mse": 0.021745679900050163, + "psnr": 17.10478401184082, + "lpips": 0.305416464805603 + }, + "overall": { + "mse": 0.021368123590946198, + "psnr": 18.132808685302734, + "lpips": 0.3567447066307068 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000168_test__val_desert__3__000038/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000168_test__val_desert__3__000038/case.json new file mode 100644 index 0000000000000000000000000000000000000000..e61625f135977b073a14ed51327b315d2c925e07 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000168_test__val_desert__3__000038/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "168", + "global_test_index": "168", + "dataset_index": "99", + "relative_path": "test/val_desert/3/000038.mp4", + "clip_id": "test/val_desert/3/000038", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000168, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.0877070419956, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000168_test__val_desert__3__000038/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000168_test__val_desert__3__000038/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..2fac2a93f757be16a583f8404b168247f10abe73 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000168_test__val_desert__3__000038/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d0add41424a2517acf6fcc6b1ae79118c35628aad718509f565d4e10b71899ea +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000168_test__val_desert__3__000038/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000168_test__val_desert__3__000038/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..c94a09345479596aef54d7493001af1d5ddd6cb0 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000168_test__val_desert__3__000038/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "168", + "global_test_index": "168", + "dataset_index": "99", + "relative_path": "test/val_desert/3/000038.mp4", + "clip_id": "test/val_desert/3/000038", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000168, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_desert/3/000038_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.005888526327908039, + "psnr": 22.572546005249023, + "lpips": 0.2650488317012787 + }, + "frames_100_200": { + "mse": 0.0076377601362764835, + "psnr": 21.200742721557617, + "lpips": 0.35595375299453735 + }, + "frames_200_300": { + "mse": 0.007073129527270794, + "psnr": 21.576807022094727, + "lpips": 0.2999154329299927 + }, + "frames_300_400": { + "mse": 0.007439468055963516, + "psnr": 21.30449676513672, + "lpips": 0.3031652569770813 + }, + "frames_400_500": { + "mse": 0.00751463370397687, + "psnr": 21.261371612548828, + "lpips": 0.3575636148452759 + }, + "overall": { + "mse": 0.007110703736543655, + "psnr": 21.583192825317383, + "lpips": 0.3163294196128845 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000169_test__val_iceplains__3__000016/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000169_test__val_iceplains__3__000016/case.json new file mode 100644 index 0000000000000000000000000000000000000000..440fbdcb0b3f0ba514b4e394c2f086e5e868079c --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000169_test__val_iceplains__3__000016/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "169", + "global_test_index": "169", + "dataset_index": "288", + "relative_path": "test/val_iceplains/3/000016.mp4", + "clip_id": "test/val_iceplains/3/000016", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000169, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.47973686410114, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000169_test__val_iceplains__3__000016/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000169_test__val_iceplains__3__000016/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..27961103c60300ac9eea4824ca6bba3612544503 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000169_test__val_iceplains__3__000016/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f0da4b377311409b62772f45eb9f903fd97c53988ee7254d3df1733e9703e143 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000169_test__val_iceplains__3__000016/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000169_test__val_iceplains__3__000016/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..0560fd5609227f510b2e8c8694d72dc9ca05e19f --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000169_test__val_iceplains__3__000016/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "169", + "global_test_index": "169", + "dataset_index": "288", + "relative_path": "test/val_iceplains/3/000016.mp4", + "clip_id": "test/val_iceplains/3/000016", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000169, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/3/000016_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.006526840850710869, + "psnr": 22.513322830200195, + "lpips": 0.17153584957122803 + }, + "frames_100_200": { + "mse": 0.022824235260486603, + "psnr": 17.65245819091797, + "lpips": 0.2549095153808594 + }, + "frames_200_300": { + "mse": 0.019986964762210846, + "psnr": 18.019977569580078, + "lpips": 0.20184291899204254 + }, + "frames_300_400": { + "mse": 0.0320192314684391, + "psnr": 16.370119094848633, + "lpips": 0.24171602725982666 + }, + "frames_400_500": { + "mse": 0.011659253388643265, + "psnr": 19.76287078857422, + "lpips": 0.1566677689552307 + }, + "overall": { + "mse": 0.018603304401040077, + "psnr": 18.86374855041504, + "lpips": 0.20533441007137299 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000170_test__val_desert__1__000038/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000170_test__val_desert__1__000038/case.json new file mode 100644 index 0000000000000000000000000000000000000000..02a5545f270a1cad525e67d20f1c9bee172eec1c --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000170_test__val_desert__1__000038/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "170", + "global_test_index": "170", + "dataset_index": "97", + "relative_path": "test/val_desert/1/000038.mp4", + "clip_id": "test/val_desert/1/000038", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000170, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.50915506889578, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000170_test__val_desert__1__000038/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000170_test__val_desert__1__000038/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..a40c26d0a53eb67bb102238282c447436b9f4c2c --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000170_test__val_desert__1__000038/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3c11dcd33a226cbea3e82b738ffcf201a1d67019be0ead24900534f25baa6216 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000170_test__val_desert__1__000038/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000170_test__val_desert__1__000038/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..b3f7d37124695e029d8e1e08ec3a81b583e6b455 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000170_test__val_desert__1__000038/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "170", + "global_test_index": "170", + "dataset_index": "97", + "relative_path": "test/val_desert/1/000038.mp4", + "clip_id": "test/val_desert/1/000038", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000170, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_desert/1/000038_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.007497961632907391, + "psnr": 21.42159652709961, + "lpips": 0.19325967133045197 + }, + "frames_100_200": { + "mse": 0.011632507666945457, + "psnr": 19.442317962646484, + "lpips": 0.2533006966114044 + }, + "frames_200_300": { + "mse": 0.009847799316048622, + "psnr": 20.229087829589844, + "lpips": 0.2614911198616028 + }, + "frames_300_400": { + "mse": 0.008327375166118145, + "psnr": 21.473350524902344, + "lpips": 0.15794533491134644 + }, + "frames_400_500": { + "mse": 0.004891327116638422, + "psnr": 25.451576232910156, + "lpips": 0.07892022281885147 + }, + "overall": { + "mse": 0.00843939371407032, + "psnr": 21.603586196899414, + "lpips": 0.18898342549800873 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000171_test__val_savanna__2__000021/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000171_test__val_savanna__2__000021/case.json new file mode 100644 index 0000000000000000000000000000000000000000..a01967bb99e6df240be820569805b238787eb390 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000171_test__val_savanna__2__000021/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "171", + "global_test_index": "171", + "dataset_index": "178", + "relative_path": "test/val_savanna/2/000021.mp4", + "clip_id": "test/val_savanna/2/000021", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000171, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.0216604459565, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000171_test__val_savanna__2__000021/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000171_test__val_savanna__2__000021/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..4652c4ffef4365aeaad9b6be815fa336f1189d6d --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000171_test__val_savanna__2__000021/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cac2092c1df4542fab80e019c79735e07029feddb8a55eef78fb7b33948324b3 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000171_test__val_savanna__2__000021/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000171_test__val_savanna__2__000021/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..819c67b81de42887e0635d855d160fe1d534e692 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000171_test__val_savanna__2__000021/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "171", + "global_test_index": "171", + "dataset_index": "178", + "relative_path": "test/val_savanna/2/000021.mp4", + "clip_id": "test/val_savanna/2/000021", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000171, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_savanna/2/000021_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0008786466787569225, + "psnr": 30.766040802001953, + "lpips": 0.010845622979104519 + }, + "frames_100_200": { + "mse": 0.0213946383446455, + "psnr": 23.912841796875, + "lpips": 0.04662133380770683 + }, + "frames_200_300": { + "mse": 0.07565885782241821, + "psnr": 11.314199447631836, + "lpips": 0.15685062110424042 + }, + "frames_300_400": { + "mse": 0.0989752933382988, + "psnr": 10.088833808898926, + "lpips": 0.2498144507408142 + }, + "frames_400_500": { + "mse": 0.07032478600740433, + "psnr": 11.595925331115723, + "lpips": 0.264774888753891 + }, + "overall": { + "mse": 0.053446441888809204, + "psnr": 17.535568237304688, + "lpips": 0.14578138291835785 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000172_test__val_iceplains__1__000007/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000172_test__val_iceplains__1__000007/case.json new file mode 100644 index 0000000000000000000000000000000000000000..1ce3b814f38554d10dd0d3b62bb29fa8bbbdc588 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000172_test__val_iceplains__1__000007/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "172", + "global_test_index": "172", + "dataset_index": "264", + "relative_path": "test/val_iceplains/1/000007.mp4", + "clip_id": "test/val_iceplains/1/000007", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000172, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 223.82317619095556, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000172_test__val_iceplains__1__000007/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000172_test__val_iceplains__1__000007/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..2b093a1f4c256906cb4556637a532fcbc5b4259d --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000172_test__val_iceplains__1__000007/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8054d9b6f6ab8f17141bfb07d0a587644fe66eba1252ce0c8f0b212fbf1115bf +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000172_test__val_iceplains__1__000007/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000172_test__val_iceplains__1__000007/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..71b788d4bbee04c5e14126380aa725cd4ba2e816 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000172_test__val_iceplains__1__000007/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "172", + "global_test_index": "172", + "dataset_index": "264", + "relative_path": "test/val_iceplains/1/000007.mp4", + "clip_id": "test/val_iceplains/1/000007", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000172, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/1/000007_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0007551621529273689, + "psnr": 31.225072860717773, + "lpips": 0.010254962369799614 + }, + "frames_100_200": { + "mse": 0.0008774463785812259, + "psnr": 30.822877883911133, + "lpips": 0.013003608211874962 + }, + "frames_200_300": { + "mse": 0.0009248884161934257, + "psnr": 30.690767288208008, + "lpips": 0.01213188748806715 + }, + "frames_300_400": { + "mse": 0.012638522312045097, + "psnr": 21.897642135620117, + "lpips": 0.0816640555858612 + }, + "frames_400_500": { + "mse": 0.0037172185257077217, + "psnr": 26.936880111694336, + "lpips": 0.03832709416747093 + }, + "overall": { + "mse": 0.003782647894695401, + "psnr": 28.314647674560547, + "lpips": 0.031076323240995407 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000173_test__val_iceplains__0__000008/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000173_test__val_iceplains__0__000008/case.json new file mode 100644 index 0000000000000000000000000000000000000000..737119a58f8ab19bb11df421f68c8d74d33044c0 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000173_test__val_iceplains__0__000008/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "173", + "global_test_index": "173", + "dataset_index": "266", + "relative_path": "test/val_iceplains/0/000008.mp4", + "clip_id": "test/val_iceplains/0/000008", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000173, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.22142926498782, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000173_test__val_iceplains__0__000008/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000173_test__val_iceplains__0__000008/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..635aff80c87208a3eb8c35c0b9a3874ca245f182 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000173_test__val_iceplains__0__000008/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:526ae3d308b7a03f2e08617d694c76e46a887afca5f5d10aa7896d2ea51442bc +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000173_test__val_iceplains__0__000008/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000173_test__val_iceplains__0__000008/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..2b13f84afa72c22eb338002ec3154f297a61abc1 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000173_test__val_iceplains__0__000008/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "173", + "global_test_index": "173", + "dataset_index": "266", + "relative_path": "test/val_iceplains/0/000008.mp4", + "clip_id": "test/val_iceplains/0/000008", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000173, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/0/000008_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.022675883024930954, + "psnr": 17.290510177612305, + "lpips": 0.21909460425376892 + }, + "frames_100_200": { + "mse": 0.016656776890158653, + "psnr": 18.187223434448242, + "lpips": 0.20549213886260986 + }, + "frames_200_300": { + "mse": 0.036590203642845154, + "psnr": 14.63988971710205, + "lpips": 0.3330352008342743 + }, + "frames_300_400": { + "mse": 0.03161051496863365, + "psnr": 15.277080535888672, + "lpips": 0.35721874237060547 + }, + "frames_400_500": { + "mse": 0.06072431430220604, + "psnr": 12.921960830688477, + "lpips": 0.4346253573894501 + }, + "overall": { + "mse": 0.03365153819322586, + "psnr": 15.663331985473633, + "lpips": 0.3098932206630707 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000174_test__val_iceplains__1__000023/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000174_test__val_iceplains__1__000023/case.json new file mode 100644 index 0000000000000000000000000000000000000000..c1caccc2f4ecfb88c0c7a6f2850a2bf0bf8df4bd --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000174_test__val_iceplains__1__000023/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "174", + "global_test_index": "174", + "dataset_index": "303", + "relative_path": "test/val_iceplains/1/000023.mp4", + "clip_id": "test/val_iceplains/1/000023", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000174, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.78351361700334, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000174_test__val_iceplains__1__000023/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000174_test__val_iceplains__1__000023/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..9ca8197c381b1f0d418fb9535476c1312ffeb37b --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000174_test__val_iceplains__1__000023/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:638b6fca7d0f760bf19924452a12da7f411637ecf5f7e838939bbfdc33769b40 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000174_test__val_iceplains__1__000023/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000174_test__val_iceplains__1__000023/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..e06b08fd63ec699a836a4b1342ef9c7bb81ec17b --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000174_test__val_iceplains__1__000023/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "174", + "global_test_index": "174", + "dataset_index": "303", + "relative_path": "test/val_iceplains/1/000023.mp4", + "clip_id": "test/val_iceplains/1/000023", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000174, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/1/000023_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0007651979685761034, + "psnr": 31.173986434936523, + "lpips": 0.010478497482836246 + }, + "frames_100_200": { + "mse": 0.000980709446594119, + "psnr": 30.430021286010742, + "lpips": 0.011289257556200027 + }, + "frames_200_300": { + "mse": 0.011850221082568169, + "psnr": 21.05548095703125, + "lpips": 0.09770283848047256 + }, + "frames_300_400": { + "mse": 0.00942776259034872, + "psnr": 22.593172073364258, + "lpips": 0.07468224316835403 + }, + "frames_400_500": { + "mse": 0.005023585632443428, + "psnr": 24.137712478637695, + "lpips": 0.04487298056483269 + }, + "overall": { + "mse": 0.005609494633972645, + "psnr": 25.878076553344727, + "lpips": 0.04780516400933266 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000175_test__val_plains__0__000031/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000175_test__val_plains__0__000031/case.json new file mode 100644 index 0000000000000000000000000000000000000000..cc3e949dcbe6121b338abf046b37d085b6d21016 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000175_test__val_plains__0__000031/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "175", + "global_test_index": "175", + "dataset_index": "442", + "relative_path": "test/val_plains/0/000031.mp4", + "clip_id": "test/val_plains/0/000031", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000175, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.99329431401566, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000175_test__val_plains__0__000031/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000175_test__val_plains__0__000031/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..296d90634b9b285422e7df86ea05dd381ca56241 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000175_test__val_plains__0__000031/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:35d286f8c2e264fcbfb6f1054e7032a144f4b311cd99665579609e8ad403c4cd +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000175_test__val_plains__0__000031/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000175_test__val_plains__0__000031/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..d5cbbe5454ef58bed984f17f72c0a48d870f86b6 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000175_test__val_plains__0__000031/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "175", + "global_test_index": "175", + "dataset_index": "442", + "relative_path": "test/val_plains/0/000031.mp4", + "clip_id": "test/val_plains/0/000031", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000175, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_plains/0/000031_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0008402083767578006, + "psnr": 30.925912857055664, + "lpips": 0.01091485284268856 + }, + "frames_100_200": { + "mse": 0.004285436123609543, + "psnr": 27.52774429321289, + "lpips": 0.025846654549241066 + }, + "frames_200_300": { + "mse": 0.019337503239512444, + "psnr": 18.012496948242188, + "lpips": 0.20624305307865143 + }, + "frames_300_400": { + "mse": 0.014785069972276688, + "psnr": 19.539186477661133, + "lpips": 0.16851383447647095 + }, + "frames_400_500": { + "mse": 0.04776793345808983, + "psnr": 13.952051162719727, + "lpips": 0.39155974984169006 + }, + "overall": { + "mse": 0.01740323193371296, + "psnr": 21.991477966308594, + "lpips": 0.16061563789844513 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000176_test__val_savanna__1__000047/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000176_test__val_savanna__1__000047/case.json new file mode 100644 index 0000000000000000000000000000000000000000..4a53874575130c9db8c48d5d2f5e41f53fcfde62 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000176_test__val_savanna__1__000047/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "176", + "global_test_index": "176", + "dataset_index": "233", + "relative_path": "test/val_savanna/1/000047.mp4", + "clip_id": "test/val_savanna/1/000047", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000176, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 223.79038262402173, + "peak_generation_vram_bytes": 107565475840, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000176_test__val_savanna__1__000047/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000176_test__val_savanna__1__000047/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..4f02777d3142e61f67a1fecbe0a5aa8a507a4822 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000176_test__val_savanna__1__000047/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:386af0074c3d2112ac1562bdcd0c87b88f3c748618d834e4610e9817187e5356 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000176_test__val_savanna__1__000047/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000176_test__val_savanna__1__000047/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..bd91caf52f89eacf80ed609b19300ab501eb76ec --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000176_test__val_savanna__1__000047/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "176", + "global_test_index": "176", + "dataset_index": "233", + "relative_path": "test/val_savanna/1/000047.mp4", + "clip_id": "test/val_savanna/1/000047", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000176, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_savanna/1/000047_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.009244482964277267, + "psnr": 21.246009826660156, + "lpips": 0.17948350310325623 + }, + "frames_100_200": { + "mse": 0.00862476509064436, + "psnr": 23.010337829589844, + "lpips": 0.09171100705862045 + }, + "frames_200_300": { + "mse": 0.02116173319518566, + "psnr": 18.239503860473633, + "lpips": 0.18384279310703278 + }, + "frames_300_400": { + "mse": 0.01155412569642067, + "psnr": 19.86504364013672, + "lpips": 0.13987821340560913 + }, + "frames_400_500": { + "mse": 0.014536329545080662, + "psnr": 18.66570472717285, + "lpips": 0.1976795643568039 + }, + "overall": { + "mse": 0.01302428636699915, + "psnr": 20.205320358276367, + "lpips": 0.15851901471614838 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000177_test__val_savanna__1__000049/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000177_test__val_savanna__1__000049/case.json new file mode 100644 index 0000000000000000000000000000000000000000..82828be9265164fddb1031dfd47f12e27dbcdc74 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000177_test__val_savanna__1__000049/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "177", + "global_test_index": "177", + "dataset_index": "240", + "relative_path": "test/val_savanna/1/000049.mp4", + "clip_id": "test/val_savanna/1/000049", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000177, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.2052869000472, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000177_test__val_savanna__1__000049/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000177_test__val_savanna__1__000049/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..f7b11a5f1e64d52b136cb9281b0b6f1487949c61 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000177_test__val_savanna__1__000049/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2bf329641600a58e9ddfb44029aca12944a44f805ce24f9f46bd7fe14ad5cd82 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000177_test__val_savanna__1__000049/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000177_test__val_savanna__1__000049/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..8fc0feb2d4be8ff9165bc73d32929df5607b4101 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000177_test__val_savanna__1__000049/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "177", + "global_test_index": "177", + "dataset_index": "240", + "relative_path": "test/val_savanna/1/000049.mp4", + "clip_id": "test/val_savanna/1/000049", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000177, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_savanna/1/000049_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.010845852084457874, + "psnr": 20.125513076782227, + "lpips": 0.11073470860719681 + }, + "frames_100_200": { + "mse": 0.013574395328760147, + "psnr": 19.060667037963867, + "lpips": 0.1314743310213089 + }, + "frames_200_300": { + "mse": 0.0077837985008955, + "psnr": 21.63711166381836, + "lpips": 0.06382851302623749 + }, + "frames_300_400": { + "mse": 0.018131351098418236, + "psnr": 18.146766662597656, + "lpips": 0.11146997660398483 + }, + "frames_400_500": { + "mse": 0.029169127345085144, + "psnr": 15.748600959777832, + "lpips": 0.2782854437828064 + }, + "overall": { + "mse": 0.015900906175374985, + "psnr": 18.9437313079834, + "lpips": 0.13915860652923584 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000178_test__val_desert__0__000017/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000178_test__val_desert__0__000017/case.json new file mode 100644 index 0000000000000000000000000000000000000000..5dab4550cf9f1c33784a5c301af5fbc487129873 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000178_test__val_desert__0__000017/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "178", + "global_test_index": "178", + "dataset_index": "42", + "relative_path": "test/val_desert/0/000017.mp4", + "clip_id": "test/val_desert/0/000017", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000178, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.25551047595218, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000178_test__val_desert__0__000017/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000178_test__val_desert__0__000017/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..3047f1775cb6f0d94aa6bb7e53241de8b8c567de --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000178_test__val_desert__0__000017/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b28f9a3b7fabfdda166f8cc725086a2a3c1247bb1287cc082810363ad7478f4a +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000178_test__val_desert__0__000017/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000178_test__val_desert__0__000017/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..077a69c7e03a61a1cca71891262e8a675a25c43b --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000178_test__val_desert__0__000017/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "178", + "global_test_index": "178", + "dataset_index": "42", + "relative_path": "test/val_desert/0/000017.mp4", + "clip_id": "test/val_desert/0/000017", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000178, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_desert/0/000017_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0015952019020915031, + "psnr": 28.519847869873047, + "lpips": 0.0252693984657526 + }, + "frames_100_200": { + "mse": 0.0031888133380562067, + "psnr": 26.102079391479492, + "lpips": 0.052372779697179794 + }, + "frames_200_300": { + "mse": 0.0021283309906721115, + "psnr": 27.572708129882812, + "lpips": 0.027202138677239418 + }, + "frames_300_400": { + "mse": 0.0007770633092150092, + "psnr": 31.10116958618164, + "lpips": 0.010572722181677818 + }, + "frames_400_500": { + "mse": 0.005769789218902588, + "psnr": 23.974885940551758, + "lpips": 0.13155139982700348 + }, + "overall": { + "mse": 0.002691839588806033, + "psnr": 27.454137802124023, + "lpips": 0.04939369112253189 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000179_test__val_iceplains__3__000048/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000179_test__val_iceplains__3__000048/case.json new file mode 100644 index 0000000000000000000000000000000000000000..08582e44aa47e62e16567d96f92625caee6c2cbc --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000179_test__val_iceplains__3__000048/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "179", + "global_test_index": "179", + "dataset_index": "369", + "relative_path": "test/val_iceplains/3/000048.mp4", + "clip_id": "test/val_iceplains/3/000048", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000179, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.9945957220625, + "peak_generation_vram_bytes": 107565475840, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000179_test__val_iceplains__3__000048/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000179_test__val_iceplains__3__000048/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..ef59ddbf9f816e56e9345f661fa7209efd820f9a --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000179_test__val_iceplains__3__000048/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:62f990002b1a06da36fbe2a060b6ea3df214cca63d19b6652c2197f577a429de +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000179_test__val_iceplains__3__000048/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000179_test__val_iceplains__3__000048/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..c6e5b8b4900b5f07b069eb9a791ed32361942c36 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000179_test__val_iceplains__3__000048/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "179", + "global_test_index": "179", + "dataset_index": "369", + "relative_path": "test/val_iceplains/3/000048.mp4", + "clip_id": "test/val_iceplains/3/000048", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000179, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/3/000048_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.017984941601753235, + "psnr": 17.59564971923828, + "lpips": 0.13755445182323456 + }, + "frames_100_200": { + "mse": 0.010967586189508438, + "psnr": 21.64156723022461, + "lpips": 0.10474232584238052 + }, + "frames_200_300": { + "mse": 0.0031651898752897978, + "psnr": 26.06787109375, + "lpips": 0.028647687286138535 + }, + "frames_300_400": { + "mse": 0.016557348892092705, + "psnr": 21.418529510498047, + "lpips": 0.13025067746639252 + }, + "frames_400_500": { + "mse": 0.0720861405134201, + "psnr": 11.661589622497559, + "lpips": 0.3902378976345062 + }, + "overall": { + "mse": 0.0241522416472435, + "psnr": 19.677043914794922, + "lpips": 0.15828660130500793 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000180_test__val_plains__3__000010/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000180_test__val_plains__3__000010/case.json new file mode 100644 index 0000000000000000000000000000000000000000..2746fc8c52f07967aa3388ff722f0b979935efec --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000180_test__val_plains__3__000010/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "180", + "global_test_index": "180", + "dataset_index": "398", + "relative_path": "test/val_plains/3/000010.mp4", + "clip_id": "test/val_plains/3/000010", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000180, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 223.4082445659442, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000180_test__val_plains__3__000010/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000180_test__val_plains__3__000010/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..90a4d6c7865f9932a4324c29499fee6889b6528c --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000180_test__val_plains__3__000010/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8ca61c7617841a77997554ceeec2a341fbce4fc5120d39c00314cefdc1c09895 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000180_test__val_plains__3__000010/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000180_test__val_plains__3__000010/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..420d86b0b8b856ed48c288877364f8c2ded655cb --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000180_test__val_plains__3__000010/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "180", + "global_test_index": "180", + "dataset_index": "398", + "relative_path": "test/val_plains/3/000010.mp4", + "clip_id": "test/val_plains/3/000010", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000180, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_plains/3/000010_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0020876412745565176, + "psnr": 28.676366806030273, + "lpips": 0.024585451930761337 + }, + "frames_100_200": { + "mse": 0.0008928640745580196, + "psnr": 30.494976043701172, + "lpips": 0.015709780156612396 + }, + "frames_200_300": { + "mse": 0.0009522492182441056, + "psnr": 30.2262020111084, + "lpips": 0.017527952790260315 + }, + "frames_300_400": { + "mse": 0.0015720180235803127, + "psnr": 29.304044723510742, + "lpips": 0.029319092631340027 + }, + "frames_400_500": { + "mse": 0.014845812693238258, + "psnr": 19.00160026550293, + "lpips": 0.18638235330581665 + }, + "overall": { + "mse": 0.004070117138326168, + "psnr": 27.54063606262207, + "lpips": 0.054704926908016205 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000181_test__val_plains__2__000027/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000181_test__val_plains__2__000027/case.json new file mode 100644 index 0000000000000000000000000000000000000000..8ec8c89673c0b013ad474468770c9c72abab1c8b --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000181_test__val_plains__2__000027/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "181", + "global_test_index": "181", + "dataset_index": "435", + "relative_path": "test/val_plains/2/000027.mp4", + "clip_id": "test/val_plains/2/000027", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000181, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.90831308497582, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000181_test__val_plains__2__000027/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000181_test__val_plains__2__000027/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..577babf53e36fdf06e86b275a5b3d27be5b74963 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000181_test__val_plains__2__000027/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8ed478d7d859568cd0bc3825cb34c0371f2da72364d09c1d79a1107a9e6d7262 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000181_test__val_plains__2__000027/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000181_test__val_plains__2__000027/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..cfd58fe9090f425b53872c703bd6b0af6848767a --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000181_test__val_plains__2__000027/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "181", + "global_test_index": "181", + "dataset_index": "435", + "relative_path": "test/val_plains/2/000027.mp4", + "clip_id": "test/val_plains/2/000027", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000181, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_plains/2/000027_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0024310562293976545, + "psnr": 28.022838592529297, + "lpips": 0.020260488614439964 + }, + "frames_100_200": { + "mse": 0.0009284248808398843, + "psnr": 30.347705841064453, + "lpips": 0.016543924808502197 + }, + "frames_200_300": { + "mse": 0.004536602180451155, + "psnr": 24.634197235107422, + "lpips": 0.03372075408697128 + }, + "frames_300_400": { + "mse": 0.002662894083186984, + "psnr": 28.19304847717285, + "lpips": 0.025938764214515686 + }, + "frames_400_500": { + "mse": 0.008006047457456589, + "psnr": 23.014057159423828, + "lpips": 0.14379407465457916 + }, + "overall": { + "mse": 0.003713004756718874, + "psnr": 26.842370986938477, + "lpips": 0.048051606863737106 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000182_test__val_savanna__2__000049/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000182_test__val_savanna__2__000049/case.json new file mode 100644 index 0000000000000000000000000000000000000000..5a7943209fe7bfad27cebaefdba77e36b9584c63 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000182_test__val_savanna__2__000049/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "182", + "global_test_index": "182", + "dataset_index": "242", + "relative_path": "test/val_savanna/2/000049.mp4", + "clip_id": "test/val_savanna/2/000049", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000182, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.8744477930013, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000182_test__val_savanna__2__000049/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000182_test__val_savanna__2__000049/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..5bb04299e28c44287b240649808c0e26c9fe3c6d --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000182_test__val_savanna__2__000049/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a2420c5398ba3e245e1bd067f5d9def002b68e162aaa09d4a228431fd0e33bb2 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000182_test__val_savanna__2__000049/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000182_test__val_savanna__2__000049/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..933fec8db445c744b8bf9eb11d1b358d4871125f --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000182_test__val_savanna__2__000049/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "182", + "global_test_index": "182", + "dataset_index": "242", + "relative_path": "test/val_savanna/2/000049.mp4", + "clip_id": "test/val_savanna/2/000049", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000182, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_savanna/2/000049_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.003578542498871684, + "psnr": 25.81414031982422, + "lpips": 0.026364071294665337 + }, + "frames_100_200": { + "mse": 0.027184929698705673, + "psnr": 17.480812072753906, + "lpips": 0.1621544510126114 + }, + "frames_200_300": { + "mse": 0.08217776566743851, + "psnr": 10.907691955566406, + "lpips": 0.3711894154548645 + }, + "frames_300_400": { + "mse": 0.09891282021999359, + "psnr": 10.155101776123047, + "lpips": 0.38610729575157166 + }, + "frames_400_500": { + "mse": 0.07500674575567245, + "psnr": 12.037632942199707, + "lpips": 0.3051260709762573 + }, + "overall": { + "mse": 0.057372163981199265, + "psnr": 15.279075622558594, + "lpips": 0.25018826127052307 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000183_test__val_iceplains__2__000045/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000183_test__val_iceplains__2__000045/case.json new file mode 100644 index 0000000000000000000000000000000000000000..41b1b887329325282c430957d8c5430b22d41947 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000183_test__val_iceplains__2__000045/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "183", + "global_test_index": "183", + "dataset_index": "361", + "relative_path": "test/val_iceplains/2/000045.mp4", + "clip_id": "test/val_iceplains/2/000045", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000183, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.47174006595742, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000183_test__val_iceplains__2__000045/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000183_test__val_iceplains__2__000045/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..bce1a62e381cf5688e5e416599e780b424c61acf --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000183_test__val_iceplains__2__000045/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2e2a9b5ce7821732eaf997df4179330a56317ad687a780aef86f364f4ef9043a +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000183_test__val_iceplains__2__000045/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000183_test__val_iceplains__2__000045/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..ba5cabc720e5ac07977a03a8943d42260cf37c26 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000183_test__val_iceplains__2__000045/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "183", + "global_test_index": "183", + "dataset_index": "361", + "relative_path": "test/val_iceplains/2/000045.mp4", + "clip_id": "test/val_iceplains/2/000045", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000183, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/2/000045_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.018255706876516342, + "psnr": 20.997419357299805, + "lpips": 0.10469307750463486 + }, + "frames_100_200": { + "mse": 0.02528175339102745, + "psnr": 18.593446731567383, + "lpips": 0.1438060998916626 + }, + "frames_200_300": { + "mse": 0.02097984030842781, + "psnr": 21.307756423950195, + "lpips": 0.09578806906938553 + }, + "frames_300_400": { + "mse": 0.01850239560008049, + "psnr": 19.756668090820312, + "lpips": 0.08759112656116486 + }, + "frames_400_500": { + "mse": 0.007119438145309687, + "psnr": 26.35297393798828, + "lpips": 0.03947019577026367 + }, + "overall": { + "mse": 0.018027829006314278, + "psnr": 21.401653289794922, + "lpips": 0.09426971524953842 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000184_test__val_plains__2__000047/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000184_test__val_plains__2__000047/case.json new file mode 100644 index 0000000000000000000000000000000000000000..86594832760df7bb76e147ab2aed6179a2e0bc72 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000184_test__val_plains__2__000047/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "184", + "global_test_index": "184", + "dataset_index": "470", + "relative_path": "test/val_plains/2/000047.mp4", + "clip_id": "test/val_plains/2/000047", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000184, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.4635016239481, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000184_test__val_plains__2__000047/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000184_test__val_plains__2__000047/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..4ad50e6748192a8f0ccc413a27123f60634c6e01 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000184_test__val_plains__2__000047/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a0b9769562f65344d7001db59bffb8c138d467acdcb375ba25c5807e4b228743 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000184_test__val_plains__2__000047/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000184_test__val_plains__2__000047/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..42dc73f6b6ca635232005568d6cfb85afb125f31 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000184_test__val_plains__2__000047/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "184", + "global_test_index": "184", + "dataset_index": "470", + "relative_path": "test/val_plains/2/000047.mp4", + "clip_id": "test/val_plains/2/000047", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000184, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_plains/2/000047_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.011540573090314865, + "psnr": 19.521507263183594, + "lpips": 0.2972938120365143 + }, + "frames_100_200": { + "mse": 0.008716238662600517, + "psnr": 20.760339736938477, + "lpips": 0.3898153305053711 + }, + "frames_200_300": { + "mse": 0.005321638658642769, + "psnr": 22.874351501464844, + "lpips": 0.3587180972099304 + }, + "frames_300_400": { + "mse": 0.007412498816847801, + "psnr": 21.589706420898438, + "lpips": 0.4197494089603424 + }, + "frames_400_500": { + "mse": 0.007423045113682747, + "psnr": 21.5477237701416, + "lpips": 0.42416083812713623 + }, + "overall": { + "mse": 0.008082798682153225, + "psnr": 21.258726119995117, + "lpips": 0.37794750928878784 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000185_test__val_iceplains__2__000042/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000185_test__val_iceplains__2__000042/case.json new file mode 100644 index 0000000000000000000000000000000000000000..7d887baed0913e1c2a4f08fefe129b64a95b4ffe --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000185_test__val_iceplains__2__000042/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "185", + "global_test_index": "185", + "dataset_index": "356", + "relative_path": "test/val_iceplains/2/000042.mp4", + "clip_id": "test/val_iceplains/2/000042", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000185, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.78526628098916, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000185_test__val_iceplains__2__000042/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000185_test__val_iceplains__2__000042/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..08bec22eab6d49337f691de8ff6885d5481b2c19 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000185_test__val_iceplains__2__000042/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ba228debf347e02c82e906a4d78f2cf89328606ffc80f642dde049fe92b264da +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000185_test__val_iceplains__2__000042/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000185_test__val_iceplains__2__000042/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..a168c465667a552386bf5d7b6565d33f04ff2cb2 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000185_test__val_iceplains__2__000042/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "185", + "global_test_index": "185", + "dataset_index": "356", + "relative_path": "test/val_iceplains/2/000042.mp4", + "clip_id": "test/val_iceplains/2/000042", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000185, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/2/000042_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.021334590390324593, + "psnr": 16.98296546936035, + "lpips": 0.21918617188930511 + }, + "frames_100_200": { + "mse": 0.04072931781411171, + "psnr": 15.812261581420898, + "lpips": 0.2945006191730499 + }, + "frames_200_300": { + "mse": 0.018620116636157036, + "psnr": 19.64048194885254, + "lpips": 0.11230625957250595 + }, + "frames_300_400": { + "mse": 0.005292389541864395, + "psnr": 26.306415557861328, + "lpips": 0.038150716572999954 + }, + "frames_400_500": { + "mse": 0.0014625228941440582, + "psnr": 29.6754093170166, + "lpips": 0.01820795238018036 + }, + "overall": { + "mse": 0.0174877867102623, + "psnr": 21.683507919311523, + "lpips": 0.13647034764289856 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000186_test__val_plains__2__000046/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000186_test__val_plains__2__000046/case.json new file mode 100644 index 0000000000000000000000000000000000000000..dd374c0bbbf16cbc1e24fd8c25cfa6a7d9a87911 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000186_test__val_plains__2__000046/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "186", + "global_test_index": "186", + "dataset_index": "468", + "relative_path": "test/val_plains/2/000046.mp4", + "clip_id": "test/val_plains/2/000046", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000186, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.95720790396444, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000186_test__val_plains__2__000046/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000186_test__val_plains__2__000046/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..f6a8d4016bbeffbec3bb8aaf5fb0f0052e1e0e1e --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000186_test__val_plains__2__000046/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:db43824f443320c74ca2af2634179b29a42f10bc8b66e2c097a2698740a67390 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000186_test__val_plains__2__000046/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000186_test__val_plains__2__000046/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..e2949b01f0f24dbd2b7a0b81f8909cf70988acf2 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000186_test__val_plains__2__000046/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "186", + "global_test_index": "186", + "dataset_index": "468", + "relative_path": "test/val_plains/2/000046.mp4", + "clip_id": "test/val_plains/2/000046", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000186, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_plains/2/000046_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.010129456408321857, + "psnr": 20.68238067626953, + "lpips": 0.2698359191417694 + }, + "frames_100_200": { + "mse": 0.006234454922378063, + "psnr": 22.20311737060547, + "lpips": 0.3002992868423462 + }, + "frames_200_300": { + "mse": 0.03860742226243019, + "psnr": 14.780054092407227, + "lpips": 0.40778931975364685 + }, + "frames_300_400": { + "mse": 0.05115716531872749, + "psnr": 13.117304801940918, + "lpips": 0.5189842581748962 + }, + "frames_400_500": { + "mse": 0.016680696979165077, + "psnr": 18.36639976501465, + "lpips": 0.5181167125701904 + }, + "overall": { + "mse": 0.024561839178204536, + "psnr": 17.829851150512695, + "lpips": 0.40300512313842773 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000187_test__val_desert__0__000043/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000187_test__val_desert__0__000043/case.json new file mode 100644 index 0000000000000000000000000000000000000000..8760cd2a4567cc2b2c830d7ee1829ee85260dafb --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000187_test__val_desert__0__000043/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "187", + "global_test_index": "187", + "dataset_index": "113", + "relative_path": "test/val_desert/0/000043.mp4", + "clip_id": "test/val_desert/0/000043", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000187, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.15593550808262, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000187_test__val_desert__0__000043/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000187_test__val_desert__0__000043/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..45ecbac01e9fec8c4d660a1df6003513a316a24f --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000187_test__val_desert__0__000043/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9329ff7df0c5f1480819b87b17d700af3e354b2d9b606db7a1664ec3bfaeaf8e +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000187_test__val_desert__0__000043/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000187_test__val_desert__0__000043/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..3f8841a9929026c9358b0e72286c9832a3279806 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000187_test__val_desert__0__000043/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "187", + "global_test_index": "187", + "dataset_index": "113", + "relative_path": "test/val_desert/0/000043.mp4", + "clip_id": "test/val_desert/0/000043", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000187, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_desert/0/000043_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.006877042818814516, + "psnr": 21.693653106689453, + "lpips": 0.34879299998283386 + }, + "frames_100_200": { + "mse": 0.008673430420458317, + "psnr": 20.959335327148438, + "lpips": 0.4165762662887573 + }, + "frames_200_300": { + "mse": 0.00687105767428875, + "psnr": 21.668256759643555, + "lpips": 0.3981080651283264 + }, + "frames_300_400": { + "mse": 0.021079914644360542, + "psnr": 17.437488555908203, + "lpips": 0.33952194452285767 + }, + "frames_400_500": { + "mse": 0.020769134163856506, + "psnr": 16.966642379760742, + "lpips": 0.3398689329624176 + }, + "overall": { + "mse": 0.012854115106165409, + "psnr": 19.745075225830078, + "lpips": 0.3685736656188965 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000188_test__val_desert__1__000008/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000188_test__val_desert__1__000008/case.json new file mode 100644 index 0000000000000000000000000000000000000000..4828da2751612ecba3e5d00f67f1aac1deb0d30d --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000188_test__val_desert__1__000008/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "188", + "global_test_index": "188", + "dataset_index": "19", + "relative_path": "test/val_desert/1/000008.mp4", + "clip_id": "test/val_desert/1/000008", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000188, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.10881033493206, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000188_test__val_desert__1__000008/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000188_test__val_desert__1__000008/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..44f5ec03aa6460f0cbfb47a2a47d92cb447645b8 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000188_test__val_desert__1__000008/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:de32edf1369f75ae5d7b9d8fd33aa8371e3e59f9199efe6ee62162ffa8381902 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000188_test__val_desert__1__000008/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000188_test__val_desert__1__000008/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..36e12e7d29ccb279db4af75fcc204a4338b43d1d --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000188_test__val_desert__1__000008/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "188", + "global_test_index": "188", + "dataset_index": "19", + "relative_path": "test/val_desert/1/000008.mp4", + "clip_id": "test/val_desert/1/000008", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000188, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_desert/1/000008_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.007863547652959824, + "psnr": 21.323667526245117, + "lpips": 0.2782555818557739 + }, + "frames_100_200": { + "mse": 0.009824289940297604, + "psnr": 20.633020401000977, + "lpips": 0.3955698013305664 + }, + "frames_200_300": { + "mse": 0.008274193853139877, + "psnr": 21.26275634765625, + "lpips": 0.3635425567626953 + }, + "frames_300_400": { + "mse": 0.007465423550456762, + "psnr": 21.304912567138672, + "lpips": 0.3031291663646698 + }, + "frames_400_500": { + "mse": 0.012528922408819199, + "psnr": 19.445959091186523, + "lpips": 0.30654624104499817 + }, + "overall": { + "mse": 0.009191274642944336, + "psnr": 20.794063568115234, + "lpips": 0.3294087052345276 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000189_test__val_savanna__0__000033/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000189_test__val_savanna__0__000033/case.json new file mode 100644 index 0000000000000000000000000000000000000000..299bfe6b8190d8efcb7d14104418bc121479eedc --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000189_test__val_savanna__0__000033/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "189", + "global_test_index": "189", + "dataset_index": "199", + "relative_path": "test/val_savanna/0/000033.mp4", + "clip_id": "test/val_savanna/0/000033", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000189, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 226.12324601900764, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000189_test__val_savanna__0__000033/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000189_test__val_savanna__0__000033/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..4e1b0062d31476e2d8f95025941b2ca163ede05c --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000189_test__val_savanna__0__000033/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1bc84ab48295b29b7be7d941bd78aff17253650e990d70f4a8a182183e5678d6 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000189_test__val_savanna__0__000033/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000189_test__val_savanna__0__000033/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..25c55a51dbee0cfb1e3800d3023707488e2d9048 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000189_test__val_savanna__0__000033/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "189", + "global_test_index": "189", + "dataset_index": "199", + "relative_path": "test/val_savanna/0/000033.mp4", + "clip_id": "test/val_savanna/0/000033", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000189, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_savanna/0/000033_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.016611862927675247, + "psnr": 18.349485397338867, + "lpips": 0.18943263590335846 + }, + "frames_100_200": { + "mse": 0.011281444691121578, + "psnr": 19.7685489654541, + "lpips": 0.11389850825071335 + }, + "frames_200_300": { + "mse": 0.022955436259508133, + "psnr": 16.486827850341797, + "lpips": 0.3150939643383026 + }, + "frames_300_400": { + "mse": 0.027115005999803543, + "psnr": 15.79104232788086, + "lpips": 0.35977500677108765 + }, + "frames_400_500": { + "mse": 0.026369502767920494, + "psnr": 15.874588966369629, + "lpips": 0.35851871967315674 + }, + "overall": { + "mse": 0.020866649225354195, + "psnr": 17.25409698486328, + "lpips": 0.26734378933906555 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000190_test__val_desert__2__000045/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000190_test__val_desert__2__000045/case.json new file mode 100644 index 0000000000000000000000000000000000000000..18d988b040949a7ee7ab6bd9d46c88aec64098b4 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000190_test__val_desert__2__000045/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "190", + "global_test_index": "190", + "dataset_index": "118", + "relative_path": "test/val_desert/2/000045.mp4", + "clip_id": "test/val_desert/2/000045", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000190, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.28946694801562, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000190_test__val_desert__2__000045/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000190_test__val_desert__2__000045/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..2158cc52dcd5c567bed8f70040ac09f593190e5e --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000190_test__val_desert__2__000045/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e23afe96b6dbead87f12269596c2ca31f11be00bfa94d9e1f44a9ba902b941b5 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000190_test__val_desert__2__000045/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000190_test__val_desert__2__000045/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..c153e5f46c46a1b9d83739ffc1f97945c1629281 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000190_test__val_desert__2__000045/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "190", + "global_test_index": "190", + "dataset_index": "118", + "relative_path": "test/val_desert/2/000045.mp4", + "clip_id": "test/val_desert/2/000045", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000190, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_desert/2/000045_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0007878431933932006, + "psnr": 31.26030731201172, + "lpips": 0.020858345553278923 + }, + "frames_100_200": { + "mse": 0.0008701477781869471, + "psnr": 30.65471649169922, + "lpips": 0.01602707803249359 + }, + "frames_200_300": { + "mse": 0.0010529746068641543, + "psnr": 30.493974685668945, + "lpips": 0.018241794779896736 + }, + "frames_300_400": { + "mse": 0.0010127623099833727, + "psnr": 29.999818801879883, + "lpips": 0.014919712208211422 + }, + "frames_400_500": { + "mse": 0.0014915235806256533, + "psnr": 28.43225860595703, + "lpips": 0.02546302042901516 + }, + "overall": { + "mse": 0.0010430502006784081, + "psnr": 30.168216705322266, + "lpips": 0.019101988524198532 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000191_test__val_plains__0__000027/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000191_test__val_plains__0__000027/case.json new file mode 100644 index 0000000000000000000000000000000000000000..8e866054a8067b8567768b079dd07f6fc31776e8 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000191_test__val_plains__0__000027/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "191", + "global_test_index": "191", + "dataset_index": "433", + "relative_path": "test/val_plains/0/000027.mp4", + "clip_id": "test/val_plains/0/000027", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000191, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.72464248898905, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000191_test__val_plains__0__000027/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000191_test__val_plains__0__000027/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..c69991ac8e053a61e40090cb7ac99196900928db --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000191_test__val_plains__0__000027/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3bcceb7276fa63724d554cdbfac3d031dc5d2fdc8cb48ead1a155bac6e309df7 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000191_test__val_plains__0__000027/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000191_test__val_plains__0__000027/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..41631d0f99e8e5c56fdf0f0146af774c223c35d2 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000191_test__val_plains__0__000027/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "191", + "global_test_index": "191", + "dataset_index": "433", + "relative_path": "test/val_plains/0/000027.mp4", + "clip_id": "test/val_plains/0/000027", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000191, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_plains/0/000027_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.003220943734049797, + "psnr": 26.752735137939453, + "lpips": 0.06246563792228699 + }, + "frames_100_200": { + "mse": 0.00400202302262187, + "psnr": 24.545799255371094, + "lpips": 0.060568906366825104 + }, + "frames_200_300": { + "mse": 0.0022693355567753315, + "psnr": 28.474035263061523, + "lpips": 0.030814122408628464 + }, + "frames_300_400": { + "mse": 0.0010258798720315099, + "psnr": 29.937807083129883, + "lpips": 0.020224900916218758 + }, + "frames_400_500": { + "mse": 0.0009283476392738521, + "psnr": 30.330307006835938, + "lpips": 0.018852107226848602 + }, + "overall": { + "mse": 0.002289305906742811, + "psnr": 28.008134841918945, + "lpips": 0.03858513385057449 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000192_test__val_iceplains__0__000034/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000192_test__val_iceplains__0__000034/case.json new file mode 100644 index 0000000000000000000000000000000000000000..f4d388baa42aa45f901f9b6db1e38ad37bc2fd3e --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000192_test__val_iceplains__0__000034/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "192", + "global_test_index": "192", + "dataset_index": "328", + "relative_path": "test/val_iceplains/0/000034.mp4", + "clip_id": "test/val_iceplains/0/000034", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000192, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.593145093997, + "peak_generation_vram_bytes": 107565475840, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000192_test__val_iceplains__0__000034/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000192_test__val_iceplains__0__000034/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..78e6a194f84544bf5cd5ad8e8acef20bb6ec89f4 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000192_test__val_iceplains__0__000034/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4d2cc731e1ea68e34c9f36e775cf975f2a17ae56aa075c5e612ba92d0128618d +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000192_test__val_iceplains__0__000034/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000192_test__val_iceplains__0__000034/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..3acb0bc9448c91f9dc25eefa620c203a63dce864 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000192_test__val_iceplains__0__000034/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "192", + "global_test_index": "192", + "dataset_index": "328", + "relative_path": "test/val_iceplains/0/000034.mp4", + "clip_id": "test/val_iceplains/0/000034", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000192, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/0/000034_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.011516967788338661, + "psnr": 19.607519149780273, + "lpips": 0.1593947857618332 + }, + "frames_100_200": { + "mse": 0.014959121122956276, + "psnr": 18.27882194519043, + "lpips": 0.14779558777809143 + }, + "frames_200_300": { + "mse": 0.017185170203447342, + "psnr": 17.710424423217773, + "lpips": 0.19106891751289368 + }, + "frames_300_400": { + "mse": 0.017956582829356194, + "psnr": 17.643081665039062, + "lpips": 0.20077261328697205 + }, + "frames_400_500": { + "mse": 0.020363831892609596, + "psnr": 17.023014068603516, + "lpips": 0.21819168329238892 + }, + "overall": { + "mse": 0.016396336257457733, + "psnr": 18.052570343017578, + "lpips": 0.18344472348690033 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000193_test__val_savanna__0__000039/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000193_test__val_savanna__0__000039/case.json new file mode 100644 index 0000000000000000000000000000000000000000..b6eee5209cc51f07f2e3c48d0b81fd72af440980 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000193_test__val_savanna__0__000039/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "193", + "global_test_index": "193", + "dataset_index": "213", + "relative_path": "test/val_savanna/0/000039.mp4", + "clip_id": "test/val_savanna/0/000039", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000193, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.10812123306096, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000193_test__val_savanna__0__000039/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000193_test__val_savanna__0__000039/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..362bca007bca8e220749729eac973168b2b4665e --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000193_test__val_savanna__0__000039/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:937f8776c15e341b55bedf711ea41aba551db47c2c30a71100dd5719f34f3f45 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000193_test__val_savanna__0__000039/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000193_test__val_savanna__0__000039/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..e4317bacf322bbe38403a2cc7481152c4e63d9dd --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000193_test__val_savanna__0__000039/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "193", + "global_test_index": "193", + "dataset_index": "213", + "relative_path": "test/val_savanna/0/000039.mp4", + "clip_id": "test/val_savanna/0/000039", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000193, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_savanna/0/000039_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0040368265472352505, + "psnr": 24.08379364013672, + "lpips": 0.2789703607559204 + }, + "frames_100_200": { + "mse": 0.007115305867046118, + "psnr": 21.66659164428711, + "lpips": 0.39563316106796265 + }, + "frames_200_300": { + "mse": 0.018594373017549515, + "psnr": 18.94845962524414, + "lpips": 0.45217448472976685 + }, + "frames_300_400": { + "mse": 0.046067796647548676, + "psnr": 13.72076416015625, + "lpips": 0.5459051132202148 + }, + "frames_400_500": { + "mse": 0.010017034597694874, + "psnr": 20.457626342773438, + "lpips": 0.49151894450187683 + }, + "overall": { + "mse": 0.017166266217827797, + "psnr": 19.775447845458984, + "lpips": 0.4328404366970062 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000194_test__val_iceplains__0__000037/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000194_test__val_iceplains__0__000037/case.json new file mode 100644 index 0000000000000000000000000000000000000000..fc6bbe5cfedaba2efa3307059259725a2c32052a --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000194_test__val_iceplains__0__000037/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "194", + "global_test_index": "194", + "dataset_index": "339", + "relative_path": "test/val_iceplains/0/000037.mp4", + "clip_id": "test/val_iceplains/0/000037", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000194, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.3236740740249, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000194_test__val_iceplains__0__000037/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000194_test__val_iceplains__0__000037/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..75aa9c9727678833b7cfc58979b46652466698b8 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000194_test__val_iceplains__0__000037/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4803de4c2ea2d211a83d29732974bae2bc67f8533837ff9294bf531867c75f16 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000194_test__val_iceplains__0__000037/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000194_test__val_iceplains__0__000037/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..cd0757a2e47f87ad6344c9bf5d29e47717615e57 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000194_test__val_iceplains__0__000037/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "194", + "global_test_index": "194", + "dataset_index": "339", + "relative_path": "test/val_iceplains/0/000037.mp4", + "clip_id": "test/val_iceplains/0/000037", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000194, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/0/000037_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.015296014957129955, + "psnr": 18.5218448638916, + "lpips": 0.18312609195709229 + }, + "frames_100_200": { + "mse": 0.013090684078633785, + "psnr": 19.679019927978516, + "lpips": 0.18592146039009094 + }, + "frames_200_300": { + "mse": 0.08436568081378937, + "psnr": 12.304203987121582, + "lpips": 0.3707493245601654 + }, + "frames_300_400": { + "mse": 0.05411187559366226, + "psnr": 13.978283882141113, + "lpips": 0.32597777247428894 + }, + "frames_400_500": { + "mse": 0.04968024417757988, + "psnr": 17.791776657104492, + "lpips": 0.4104113280773163 + }, + "overall": { + "mse": 0.04330890253186226, + "psnr": 16.45502471923828, + "lpips": 0.2952371835708618 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000195_test__val_plains__1__000046/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000195_test__val_plains__1__000046/case.json new file mode 100644 index 0000000000000000000000000000000000000000..8fe64076ab91f6c937ae29e0283a8700e8ff2d4e --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000195_test__val_plains__1__000046/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "195", + "global_test_index": "195", + "dataset_index": "466", + "relative_path": "test/val_plains/1/000046.mp4", + "clip_id": "test/val_plains/1/000046", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000195, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.2344749569893, + "peak_generation_vram_bytes": 107565475840, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000195_test__val_plains__1__000046/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000195_test__val_plains__1__000046/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..95c0eef6ad3bc02db5edc251dfb4d638cba758ee --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000195_test__val_plains__1__000046/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f0dfea76d77e01317ff48a06a1cb9549b763457e31d7565e56d4dcae10806528 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000195_test__val_plains__1__000046/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000195_test__val_plains__1__000046/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..505968de1a47d23ed91b7f316ab7e3a4ac8aac92 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000195_test__val_plains__1__000046/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "195", + "global_test_index": "195", + "dataset_index": "466", + "relative_path": "test/val_plains/1/000046.mp4", + "clip_id": "test/val_plains/1/000046", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000195, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_plains/1/000046_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.012168174609541893, + "psnr": 19.313140869140625, + "lpips": 0.27231574058532715 + }, + "frames_100_200": { + "mse": 0.009969277307391167, + "psnr": 20.36556053161621, + "lpips": 0.35678330063819885 + }, + "frames_200_300": { + "mse": 0.023465633392333984, + "psnr": 17.102731704711914, + "lpips": 0.31830716133117676 + }, + "frames_300_400": { + "mse": 0.03326743468642235, + "psnr": 14.819114685058594, + "lpips": 0.3457588851451874 + }, + "frames_400_500": { + "mse": 0.02922452986240387, + "psnr": 15.626818656921387, + "lpips": 0.45584434270858765 + }, + "overall": { + "mse": 0.021619010716676712, + "psnr": 17.44547462463379, + "lpips": 0.3498018682003021 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000196_test__val_iceplains__3__000042/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000196_test__val_iceplains__3__000042/case.json new file mode 100644 index 0000000000000000000000000000000000000000..4244017f3c311cd2b24672e8ded6fbbca3acb63f --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000196_test__val_iceplains__3__000042/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "196", + "global_test_index": "196", + "dataset_index": "355", + "relative_path": "test/val_iceplains/3/000042.mp4", + "clip_id": "test/val_iceplains/3/000042", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000196, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.18491152103525, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000196_test__val_iceplains__3__000042/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000196_test__val_iceplains__3__000042/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..646c53e8dfe66acecc9fbacac40bea5c12b9fb19 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000196_test__val_iceplains__3__000042/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bffca582681f72c88bb9fb2fdb5a8bd7c03d128750bd20e45c3b63da8929d70d +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000196_test__val_iceplains__3__000042/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000196_test__val_iceplains__3__000042/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..19baa4ba3f00a7b9d3644eebd42fa7bde9e92562 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000196_test__val_iceplains__3__000042/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "196", + "global_test_index": "196", + "dataset_index": "355", + "relative_path": "test/val_iceplains/3/000042.mp4", + "clip_id": "test/val_iceplains/3/000042", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000196, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/3/000042_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0334123931825161, + "psnr": 17.902795791625977, + "lpips": 0.18587221205234528 + }, + "frames_100_200": { + "mse": 0.06218600273132324, + "psnr": 12.673473358154297, + "lpips": 0.3281601667404175 + }, + "frames_200_300": { + "mse": 0.03142104670405388, + "psnr": 16.45438575744629, + "lpips": 0.16733436286449432 + }, + "frames_300_400": { + "mse": 0.053473569452762604, + "psnr": 13.059839248657227, + "lpips": 0.30373460054397583 + }, + "frames_400_500": { + "mse": 0.06323057413101196, + "psnr": 13.19672966003418, + "lpips": 0.3274418115615845 + }, + "overall": { + "mse": 0.04874471202492714, + "psnr": 14.657443046569824, + "lpips": 0.2625086307525635 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000197_test__val_iceplains__3__000036/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000197_test__val_iceplains__3__000036/case.json new file mode 100644 index 0000000000000000000000000000000000000000..29dda1a96df0118b61e2b853eac3a4ea8383e824 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000197_test__val_iceplains__3__000036/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "197", + "global_test_index": "197", + "dataset_index": "336", + "relative_path": "test/val_iceplains/3/000036.mp4", + "clip_id": "test/val_iceplains/3/000036", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000197, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.1694725700654, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000197_test__val_iceplains__3__000036/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000197_test__val_iceplains__3__000036/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..99a11fc8fa870380c8dda1708eb343b9e0c614d3 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000197_test__val_iceplains__3__000036/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bea8b6748b91e08b9d6b1c8b063a7cec6d827a170568fa2fd4f14df205463308 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000197_test__val_iceplains__3__000036/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000197_test__val_iceplains__3__000036/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..31a63b1666a615f384c7144e929ae08e203e81c7 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000197_test__val_iceplains__3__000036/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "197", + "global_test_index": "197", + "dataset_index": "336", + "relative_path": "test/val_iceplains/3/000036.mp4", + "clip_id": "test/val_iceplains/3/000036", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000197, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/3/000036_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.004084817599505186, + "psnr": 24.766620635986328, + "lpips": 0.04843214154243469 + }, + "frames_100_200": { + "mse": 0.010874840430915356, + "psnr": 20.295621871948242, + "lpips": 0.10983239859342575 + }, + "frames_200_300": { + "mse": 0.015814591199159622, + "psnr": 18.193286895751953, + "lpips": 0.1812829226255417 + }, + "frames_300_400": { + "mse": 0.023578789085149765, + "psnr": 16.871885299682617, + "lpips": 0.2258908897638321 + }, + "frames_400_500": { + "mse": 0.021639054641127586, + "psnr": 17.68769073486328, + "lpips": 0.17630228400230408 + }, + "overall": { + "mse": 0.0151984179392457, + "psnr": 19.563020706176758, + "lpips": 0.1483481228351593 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000198_test__val_plains__0__000020/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000198_test__val_plains__0__000020/case.json new file mode 100644 index 0000000000000000000000000000000000000000..e0cdbcd34d5bd0c263b2916b8cab21d60049a91c --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000198_test__val_plains__0__000020/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "198", + "global_test_index": "198", + "dataset_index": "420", + "relative_path": "test/val_plains/0/000020.mp4", + "clip_id": "test/val_plains/0/000020", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000198, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.27578241296578, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000198_test__val_plains__0__000020/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000198_test__val_plains__0__000020/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..0da73642de7ad11024a83ca7f9b4d85c6f0fb1e4 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000198_test__val_plains__0__000020/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:176ed45efb7701d8901465abb5ee66ebb13032431b7c412781d5e18ed15e9cfa +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000198_test__val_plains__0__000020/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000198_test__val_plains__0__000020/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..bfe43a37fb76cf615f2887adc960b88ab8f9e5f4 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000198_test__val_plains__0__000020/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "198", + "global_test_index": "198", + "dataset_index": "420", + "relative_path": "test/val_plains/0/000020.mp4", + "clip_id": "test/val_plains/0/000020", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000198, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_plains/0/000020_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.00907343439757824, + "psnr": 20.61187744140625, + "lpips": 0.32998400926589966 + }, + "frames_100_200": { + "mse": 0.004796893801540136, + "psnr": 23.421567916870117, + "lpips": 0.37262386083602905 + }, + "frames_200_300": { + "mse": 0.010713261552155018, + "psnr": 20.328441619873047, + "lpips": 0.4261733293533325 + }, + "frames_300_400": { + "mse": 0.00529928645119071, + "psnr": 22.795488357543945, + "lpips": 0.4262222349643707 + }, + "frames_400_500": { + "mse": 0.0059500765055418015, + "psnr": 22.304899215698242, + "lpips": 0.431120365858078 + }, + "overall": { + "mse": 0.007166590541601181, + "psnr": 21.892454147338867, + "lpips": 0.3972247838973999 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000199_test__val_plains__1__000015/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000199_test__val_plains__1__000015/case.json new file mode 100644 index 0000000000000000000000000000000000000000..dec892c1dbdbc0599390dd23a1b4e88b134f5cf8 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000199_test__val_plains__1__000015/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "199", + "global_test_index": "199", + "dataset_index": "410", + "relative_path": "test/val_plains/1/000015.mp4", + "clip_id": "test/val_plains/1/000015", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000199, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 223.53065106202848, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000199_test__val_plains__1__000015/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000199_test__val_plains__1__000015/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..3a4c537bae94535186ffb346a896ff99c3038721 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000199_test__val_plains__1__000015/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6339da6899a9ab5982472332befb7f5106746c09783369e5af87e0d3d1472236 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000199_test__val_plains__1__000015/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000199_test__val_plains__1__000015/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..552badd5715d94638ae90f2a504c75387abbe752 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000199_test__val_plains__1__000015/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "199", + "global_test_index": "199", + "dataset_index": "410", + "relative_path": "test/val_plains/1/000015.mp4", + "clip_id": "test/val_plains/1/000015", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000199, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_plains/1/000015_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0019218851812183857, + "psnr": 29.299156188964844, + "lpips": 0.019482281059026718 + }, + "frames_100_200": { + "mse": 0.003867452498525381, + "psnr": 27.674177169799805, + "lpips": 0.036929816007614136 + }, + "frames_200_300": { + "mse": 0.015434023924171925, + "psnr": 22.311260223388672, + "lpips": 0.08753383904695511 + }, + "frames_300_400": { + "mse": 0.0025834408588707447, + "psnr": 27.753381729125977, + "lpips": 0.026487020775675774 + }, + "frames_400_500": { + "mse": 0.008781240321695805, + "psnr": 22.9293270111084, + "lpips": 0.06019844487309456 + }, + "overall": { + "mse": 0.006517609115689993, + "psnr": 25.99346160888672, + "lpips": 0.04612627625465393 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000200_test__val_desert__1__000011/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000200_test__val_desert__1__000011/case.json new file mode 100644 index 0000000000000000000000000000000000000000..ba7ea55eecb9959541df340f0220d433d1ec8c30 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000200_test__val_desert__1__000011/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "200", + "global_test_index": "200", + "dataset_index": "27", + "relative_path": "test/val_desert/1/000011.mp4", + "clip_id": "test/val_desert/1/000011", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000200, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.5548032920342, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000200_test__val_desert__1__000011/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000200_test__val_desert__1__000011/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..aec89d62180ef2ca22f723a209710b87bf1d2839 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000200_test__val_desert__1__000011/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1ddaad9cd444e2039b8189a6f904702107f26f18c7475ab054eb66ee24270470 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000200_test__val_desert__1__000011/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000200_test__val_desert__1__000011/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..413cb1fc8e11a85c4d465d8f64ea942189f2ae41 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000200_test__val_desert__1__000011/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "200", + "global_test_index": "200", + "dataset_index": "27", + "relative_path": "test/val_desert/1/000011.mp4", + "clip_id": "test/val_desert/1/000011", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000200, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_desert/1/000011_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.00362486788071692, + "psnr": 25.391002655029297, + "lpips": 0.09392304718494415 + }, + "frames_100_200": { + "mse": 0.012368638068437576, + "psnr": 19.30221939086914, + "lpips": 0.24656955897808075 + }, + "frames_200_300": { + "mse": 0.010800915770232677, + "psnr": 20.225322723388672, + "lpips": 0.18555137515068054 + }, + "frames_300_400": { + "mse": 0.0072687482461333275, + "psnr": 21.93059539794922, + "lpips": 0.14408521354198456 + }, + "frames_400_500": { + "mse": 0.008595293387770653, + "psnr": 22.148836135864258, + "lpips": 0.10513468831777573 + }, + "overall": { + "mse": 0.008531692437827587, + "psnr": 21.799596786499023, + "lpips": 0.1550527811050415 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000201_test__val_desert__1__000033/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000201_test__val_desert__1__000033/case.json new file mode 100644 index 0000000000000000000000000000000000000000..9aa8cc2f594c8beb5f4e3213ae1e6ef47d372680 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000201_test__val_desert__1__000033/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "201", + "global_test_index": "201", + "dataset_index": "86", + "relative_path": "test/val_desert/1/000033.mp4", + "clip_id": "test/val_desert/1/000033", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000201, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.63187725190073, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000201_test__val_desert__1__000033/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000201_test__val_desert__1__000033/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..6b160f2e0c597293051a8ecd2c82ef03757d82ef --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000201_test__val_desert__1__000033/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:58d5d625c7b5e15f31788484c940200c01fa2aca146d4fb0f14db715bd22268a +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000201_test__val_desert__1__000033/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000201_test__val_desert__1__000033/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..01f7985511d37537b2ff9dd73c919b5cd3f691ec --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000201_test__val_desert__1__000033/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "201", + "global_test_index": "201", + "dataset_index": "86", + "relative_path": "test/val_desert/1/000033.mp4", + "clip_id": "test/val_desert/1/000033", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000201, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_desert/1/000033_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0011489070020616055, + "psnr": 30.037368774414062, + "lpips": 0.013051777146756649 + }, + "frames_100_200": { + "mse": 0.0007827995577827096, + "psnr": 31.069875717163086, + "lpips": 0.010831968858838081 + }, + "frames_200_300": { + "mse": 0.0008554576197639108, + "psnr": 30.703258514404297, + "lpips": 0.012771803885698318 + }, + "frames_300_400": { + "mse": 0.0060013229958713055, + "psnr": 23.67142105102539, + "lpips": 0.08193662017583847 + }, + "frames_400_500": { + "mse": 0.008755329065024853, + "psnr": 21.033493041992188, + "lpips": 0.15296438336372375 + }, + "overall": { + "mse": 0.003508763387799263, + "psnr": 27.303081512451172, + "lpips": 0.05431130900979042 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000202_test__val_savanna__3__000038/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000202_test__val_savanna__3__000038/case.json new file mode 100644 index 0000000000000000000000000000000000000000..7b891f0f3a0076d4a298317bca59da7cac77c933 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000202_test__val_savanna__3__000038/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "202", + "global_test_index": "202", + "dataset_index": "210", + "relative_path": "test/val_savanna/3/000038.mp4", + "clip_id": "test/val_savanna/3/000038", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000202, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.25867512600962, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000202_test__val_savanna__3__000038/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000202_test__val_savanna__3__000038/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..77d2865f865f7a4a4f938aece8fc2ab17eb2130d --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000202_test__val_savanna__3__000038/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d7ade6f8ef44f91e4f351b084a433fe6842ea13cf5031b05f8999aa8ea61be6b +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000202_test__val_savanna__3__000038/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000202_test__val_savanna__3__000038/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..d0e2d5b88fca2488f560add91a760c76a5ac7262 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000202_test__val_savanna__3__000038/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "202", + "global_test_index": "202", + "dataset_index": "210", + "relative_path": "test/val_savanna/3/000038.mp4", + "clip_id": "test/val_savanna/3/000038", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000202, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_savanna/3/000038_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.004524491261690855, + "psnr": 23.59777069091797, + "lpips": 0.36287960410118103 + }, + "frames_100_200": { + "mse": 0.005037670023739338, + "psnr": 23.015853881835938, + "lpips": 0.39514410495758057 + }, + "frames_200_300": { + "mse": 0.011023546569049358, + "psnr": 20.114669799804688, + "lpips": 0.43252670764923096 + }, + "frames_300_400": { + "mse": 0.011861315928399563, + "psnr": 20.503843307495117, + "lpips": 0.4467502236366272 + }, + "frames_400_500": { + "mse": 0.007937407121062279, + "psnr": 21.113178253173828, + "lpips": 0.45269694924354553 + }, + "overall": { + "mse": 0.00807688757777214, + "psnr": 21.669063568115234, + "lpips": 0.4179995059967041 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000203_test__val_desert__3__000039/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000203_test__val_desert__3__000039/case.json new file mode 100644 index 0000000000000000000000000000000000000000..1671ced3d7ea125b29268747d3188d43c0b67b33 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000203_test__val_desert__3__000039/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "203", + "global_test_index": "203", + "dataset_index": "103", + "relative_path": "test/val_desert/3/000039.mp4", + "clip_id": "test/val_desert/3/000039", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000203, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 226.44227901590057, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000203_test__val_desert__3__000039/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000203_test__val_desert__3__000039/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..f651f7b1f6f2db0167f0fe5dcdb99b361c671e00 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000203_test__val_desert__3__000039/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:817d3a07d437c3946bb2202a85f64c85844c1fc2e7be770930a1e120e7a30070 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000203_test__val_desert__3__000039/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000203_test__val_desert__3__000039/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..eda7f6af8432a17dabbe295ae4c854d9178da13e --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000203_test__val_desert__3__000039/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "203", + "global_test_index": "203", + "dataset_index": "103", + "relative_path": "test/val_desert/3/000039.mp4", + "clip_id": "test/val_desert/3/000039", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000203, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_desert/3/000039_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.007648169994354248, + "psnr": 21.386049270629883, + "lpips": 0.19229121506214142 + }, + "frames_100_200": { + "mse": 0.014335393905639648, + "psnr": 18.64691925048828, + "lpips": 0.24679747223854065 + }, + "frames_200_300": { + "mse": 0.015434670262038708, + "psnr": 21.231834411621094, + "lpips": 0.17364324629306793 + }, + "frames_300_400": { + "mse": 0.00895875133574009, + "psnr": 22.315946578979492, + "lpips": 0.13760308921337128 + }, + "frames_400_500": { + "mse": 0.01399846188724041, + "psnr": 19.976242065429688, + "lpips": 0.2212902009487152 + }, + "overall": { + "mse": 0.01207508984953165, + "psnr": 20.71139907836914, + "lpips": 0.1943250447511673 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000204_test__val_iceplains__3__000000/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000204_test__val_iceplains__3__000000/case.json new file mode 100644 index 0000000000000000000000000000000000000000..63caf32504694de82efe5f8735f8a4631075e023 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000204_test__val_iceplains__3__000000/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "204", + "global_test_index": "204", + "dataset_index": "245", + "relative_path": "test/val_iceplains/3/000000.mp4", + "clip_id": "test/val_iceplains/3/000000", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000204, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 223.98518726008479, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000204_test__val_iceplains__3__000000/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000204_test__val_iceplains__3__000000/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..8416fb3ac126210802551fd24c3619c0e84cdb1a --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000204_test__val_iceplains__3__000000/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6337a55c11b9ebae4c75d14a7b39de75eeb1bc0e4bf06c6612d0428ed45cc1cd +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000204_test__val_iceplains__3__000000/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000204_test__val_iceplains__3__000000/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..d0756d9b022d5be7972be9db5df274bf634f8d4b --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000204_test__val_iceplains__3__000000/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "204", + "global_test_index": "204", + "dataset_index": "245", + "relative_path": "test/val_iceplains/3/000000.mp4", + "clip_id": "test/val_iceplains/3/000000", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000204, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/3/000000_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0011150840437039733, + "psnr": 30.852479934692383, + "lpips": 0.013374615460634232 + }, + "frames_100_200": { + "mse": 0.0007544684340246022, + "psnr": 31.239885330200195, + "lpips": 0.010767140425741673 + }, + "frames_200_300": { + "mse": 0.000785920477937907, + "psnr": 31.073209762573242, + "lpips": 0.011145710945129395 + }, + "frames_300_400": { + "mse": 0.0007488331175409257, + "psnr": 31.270095825195312, + "lpips": 0.010845008306205273 + }, + "frames_400_500": { + "mse": 0.0007993900799192488, + "psnr": 31.035350799560547, + "lpips": 0.012134998105466366 + }, + "overall": { + "mse": 0.0008407392306253314, + "psnr": 31.094205856323242, + "lpips": 0.011653494089841843 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000205_test__val_desert__2__000009/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000205_test__val_desert__2__000009/case.json new file mode 100644 index 0000000000000000000000000000000000000000..88a8e92bef57a5c170aa8f33a1abd38e5dac044a --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000205_test__val_desert__2__000009/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "205", + "global_test_index": "205", + "dataset_index": "25", + "relative_path": "test/val_desert/2/000009.mp4", + "clip_id": "test/val_desert/2/000009", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000205, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.83104086096864, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000205_test__val_desert__2__000009/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000205_test__val_desert__2__000009/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..55b1cf5c24506e5c2e5c92c403524e3e220da0cb --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000205_test__val_desert__2__000009/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:80cd36e6e52c5e76c39077f26c4853443753cede5c7f4ced636c51930156d17f +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000205_test__val_desert__2__000009/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000205_test__val_desert__2__000009/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..7809849aeeacf5a672d59977090bd87d07bf428d --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000205_test__val_desert__2__000009/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "205", + "global_test_index": "205", + "dataset_index": "25", + "relative_path": "test/val_desert/2/000009.mp4", + "clip_id": "test/val_desert/2/000009", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000205, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_desert/2/000009_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0015696038026362658, + "psnr": 28.220458984375, + "lpips": 0.12413526326417923 + }, + "frames_100_200": { + "mse": 0.0018215924501419067, + "psnr": 27.621397018432617, + "lpips": 0.14396890997886658 + }, + "frames_200_300": { + "mse": 0.0661718100309372, + "psnr": 12.231398582458496, + "lpips": 0.477171391248703 + }, + "frames_300_400": { + "mse": 0.08482906222343445, + "psnr": 10.868263244628906, + "lpips": 0.48449334502220154 + }, + "frames_400_500": { + "mse": 0.1258002072572708, + "psnr": 9.147476196289062, + "lpips": 0.5845692157745361 + }, + "overall": { + "mse": 0.056038450449705124, + "psnr": 17.617799758911133, + "lpips": 0.3628676235675812 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000206_test__val_iceplains__2__000037/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000206_test__val_iceplains__2__000037/case.json new file mode 100644 index 0000000000000000000000000000000000000000..d90b9c7b275698406902d8e428215cb6cd772995 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000206_test__val_iceplains__2__000037/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "206", + "global_test_index": "206", + "dataset_index": "340", + "relative_path": "test/val_iceplains/2/000037.mp4", + "clip_id": "test/val_iceplains/2/000037", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000206, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.78348480002023, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000206_test__val_iceplains__2__000037/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000206_test__val_iceplains__2__000037/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..662146fde9815346f0db07e4f60d90bea4270ac9 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000206_test__val_iceplains__2__000037/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:931e223e78d2fd343a328ded7692b819b121d29d14802e06536291f53b102661 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000206_test__val_iceplains__2__000037/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000206_test__val_iceplains__2__000037/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..122e9c20f09d2e189e94fe0c6a1ebdca515df034 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000206_test__val_iceplains__2__000037/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "206", + "global_test_index": "206", + "dataset_index": "340", + "relative_path": "test/val_iceplains/2/000037.mp4", + "clip_id": "test/val_iceplains/2/000037", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000206, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/2/000037_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.05317787081003189, + "psnr": 15.050275802612305, + "lpips": 0.30287691950798035 + }, + "frames_100_200": { + "mse": 0.07052438706159592, + "psnr": 11.753364562988281, + "lpips": 0.40387600660324097 + }, + "frames_200_300": { + "mse": 0.10666000097990036, + "psnr": 9.936388969421387, + "lpips": 0.5392968058586121 + }, + "frames_300_400": { + "mse": 0.1384522020816803, + "psnr": 8.788873672485352, + "lpips": 0.5936058759689331 + }, + "frames_400_500": { + "mse": 0.14915595948696136, + "psnr": 8.362622261047363, + "lpips": 0.6046961545944214 + }, + "overall": { + "mse": 0.1035940945148468, + "psnr": 10.778304100036621, + "lpips": 0.4888703525066376 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000207_test__val_savanna__0__000016/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000207_test__val_savanna__0__000016/case.json new file mode 100644 index 0000000000000000000000000000000000000000..c147841d85f5b0890883f7c6bdd35e352c1ab93f --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000207_test__val_savanna__0__000016/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "207", + "global_test_index": "207", + "dataset_index": "167", + "relative_path": "test/val_savanna/0/000016.mp4", + "clip_id": "test/val_savanna/0/000016", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000207, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 223.78182911500335, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000207_test__val_savanna__0__000016/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000207_test__val_savanna__0__000016/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..172ef61515930cd2e24f54060b2cb8da8c270437 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000207_test__val_savanna__0__000016/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:938d647f4b7088fb0c6c081be069bbea8b0e3e0f5159e629f2827b086da91973 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000207_test__val_savanna__0__000016/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000207_test__val_savanna__0__000016/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..61de09210d4159a6216ae15ba360a989f00098fa --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000207_test__val_savanna__0__000016/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "207", + "global_test_index": "207", + "dataset_index": "167", + "relative_path": "test/val_savanna/0/000016.mp4", + "clip_id": "test/val_savanna/0/000016", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000207, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_savanna/0/000016_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.010428035631775856, + "psnr": 20.23902702331543, + "lpips": 0.19459208846092224 + }, + "frames_100_200": { + "mse": 0.0170405525714159, + "psnr": 17.921899795532227, + "lpips": 0.26109394431114197 + }, + "frames_200_300": { + "mse": 0.03811243176460266, + "psnr": 14.484735488891602, + "lpips": 0.43681156635284424 + }, + "frames_300_400": { + "mse": 0.0720912367105484, + "psnr": 11.564650535583496, + "lpips": 0.6191455721855164 + }, + "frames_400_500": { + "mse": 0.03327961638569832, + "psnr": 15.239958763122559, + "lpips": 0.39650431275367737 + }, + "overall": { + "mse": 0.03419037535786629, + "psnr": 15.890053749084473, + "lpips": 0.38162946701049805 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000208_test__val_plains__0__000022/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000208_test__val_plains__0__000022/case.json new file mode 100644 index 0000000000000000000000000000000000000000..6728108aa9a161d8afa793ef7a720031c13894ef --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000208_test__val_plains__0__000022/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "208", + "global_test_index": "208", + "dataset_index": "424", + "relative_path": "test/val_plains/0/000022.mp4", + "clip_id": "test/val_plains/0/000022", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000208, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 223.9986240719445, + "peak_generation_vram_bytes": 107565475840, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000208_test__val_plains__0__000022/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000208_test__val_plains__0__000022/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..ba51c0a15122e90409af801b60383c03e6330025 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000208_test__val_plains__0__000022/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a3130709822dc802bdcbdd1437e88150a6e560984e53ddbb034b6fd4a9b358be +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000208_test__val_plains__0__000022/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000208_test__val_plains__0__000022/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..3bc1bdbccdfdcaba7a07278ba0d3a5112bc2926b --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000208_test__val_plains__0__000022/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "208", + "global_test_index": "208", + "dataset_index": "424", + "relative_path": "test/val_plains/0/000022.mp4", + "clip_id": "test/val_plains/0/000022", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000208, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_plains/0/000022_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0008588198688812554, + "psnr": 30.673137664794922, + "lpips": 0.012842624448239803 + }, + "frames_100_200": { + "mse": 0.0009080051095224917, + "psnr": 30.432289123535156, + "lpips": 0.01701403222978115 + }, + "frames_200_300": { + "mse": 0.0010945785325020552, + "psnr": 29.81890296936035, + "lpips": 0.018163049593567848 + }, + "frames_300_400": { + "mse": 0.0009583086357451975, + "psnr": 30.189790725708008, + "lpips": 0.01808917708694935 + }, + "frames_400_500": { + "mse": 0.0016446857480332255, + "psnr": 29.176837921142578, + "lpips": 0.02501816675066948 + }, + "overall": { + "mse": 0.0010928795672953129, + "psnr": 30.058189392089844, + "lpips": 0.0182254109531641 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000209_test__val_plains__2__000033/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000209_test__val_plains__2__000033/case.json new file mode 100644 index 0000000000000000000000000000000000000000..63033731416000924723208f764e032a8cef7929 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000209_test__val_plains__2__000033/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "209", + "global_test_index": "209", + "dataset_index": "448", + "relative_path": "test/val_plains/2/000033.mp4", + "clip_id": "test/val_plains/2/000033", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000209, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.03433804993983, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000209_test__val_plains__2__000033/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000209_test__val_plains__2__000033/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..dbc717e01855604e5724091739955f3e682a3d2f --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000209_test__val_plains__2__000033/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4c37ef4af600673b8e097bf38900f0b09614f9c61132e6a85be8413cc413c5eb +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000209_test__val_plains__2__000033/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000209_test__val_plains__2__000033/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..fdc1b10ab6253b89b1a87be4516f938e3882b196 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000209_test__val_plains__2__000033/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "209", + "global_test_index": "209", + "dataset_index": "448", + "relative_path": "test/val_plains/2/000033.mp4", + "clip_id": "test/val_plains/2/000033", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000209, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_plains/2/000033_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.007034074515104294, + "psnr": 22.37718963623047, + "lpips": 0.11747369915246964 + }, + "frames_100_200": { + "mse": 0.006710109766572714, + "psnr": 22.8861026763916, + "lpips": 0.08881363272666931 + }, + "frames_200_300": { + "mse": 0.014096688479185104, + "psnr": 19.47840690612793, + "lpips": 0.18888786435127258 + }, + "frames_300_400": { + "mse": 0.0099519407376647, + "psnr": 22.89112091064453, + "lpips": 0.13021889328956604 + }, + "frames_400_500": { + "mse": 0.0009298513177782297, + "psnr": 30.322717666625977, + "lpips": 0.01819494180381298 + }, + "overall": { + "mse": 0.007744533009827137, + "psnr": 23.591108322143555, + "lpips": 0.10871779173612595 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000210_test__val_plains__0__000006/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000210_test__val_plains__0__000006/case.json new file mode 100644 index 0000000000000000000000000000000000000000..f3f22464b496a8dcafdb7801adec33ee00e74960 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000210_test__val_plains__0__000006/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "210", + "global_test_index": "210", + "dataset_index": "386", + "relative_path": "test/val_plains/0/000006.mp4", + "clip_id": "test/val_plains/0/000006", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000210, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.08248647500295, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000210_test__val_plains__0__000006/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000210_test__val_plains__0__000006/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..ab956c978e988df94dcae66de5deac75d240e521 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000210_test__val_plains__0__000006/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d95660084612f87b9af8a967eda4475fd351145834e14bced61ade5efdd0ccfe +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000210_test__val_plains__0__000006/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000210_test__val_plains__0__000006/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..7245a6e9b5ab2d72c7ca2cf9811d67b00955cb63 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000210_test__val_plains__0__000006/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "210", + "global_test_index": "210", + "dataset_index": "386", + "relative_path": "test/val_plains/0/000006.mp4", + "clip_id": "test/val_plains/0/000006", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000210, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_plains/0/000006_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.007474067155271769, + "psnr": 21.602956771850586, + "lpips": 0.2250880002975464 + }, + "frames_100_200": { + "mse": 0.013674180954694748, + "psnr": 18.995695114135742, + "lpips": 0.3354555368423462 + }, + "frames_200_300": { + "mse": 0.01414920762181282, + "psnr": 18.768869400024414, + "lpips": 0.3322499990463257 + }, + "frames_300_400": { + "mse": 0.007583660073578358, + "psnr": 21.374073028564453, + "lpips": 0.35985156893730164 + }, + "frames_400_500": { + "mse": 0.006205554120242596, + "psnr": 22.204322814941406, + "lpips": 0.346213161945343 + }, + "overall": { + "mse": 0.009817332960665226, + "psnr": 20.589183807373047, + "lpips": 0.3197717070579529 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000211_test__val_plains__3__000008/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000211_test__val_plains__3__000008/case.json new file mode 100644 index 0000000000000000000000000000000000000000..41edcef97003e546c2a24d53740508e0f15b8a1e --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000211_test__val_plains__3__000008/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "211", + "global_test_index": "211", + "dataset_index": "392", + "relative_path": "test/val_plains/3/000008.mp4", + "clip_id": "test/val_plains/3/000008", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000211, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.59536224405747, + "peak_generation_vram_bytes": 107565475840, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000211_test__val_plains__3__000008/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000211_test__val_plains__3__000008/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..55e0498e9fca58815aab7836ea0ad95f393cb3b2 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000211_test__val_plains__3__000008/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2d02766dd75efe454a5607500b0ec101fb1f7d10ad78fa34efc8cf6eb00a3fc1 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000211_test__val_plains__3__000008/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000211_test__val_plains__3__000008/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..9b798fb0e7b2d7202621b95faee1c8fd31a46569 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000211_test__val_plains__3__000008/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "211", + "global_test_index": "211", + "dataset_index": "392", + "relative_path": "test/val_plains/3/000008.mp4", + "clip_id": "test/val_plains/3/000008", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000211, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_plains/3/000008_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.01725086010992527, + "psnr": 18.09272575378418, + "lpips": 0.33931347727775574 + }, + "frames_100_200": { + "mse": 0.015820028260350227, + "psnr": 18.902639389038086, + "lpips": 0.43926993012428284 + }, + "frames_200_300": { + "mse": 0.008666392415761948, + "psnr": 20.804054260253906, + "lpips": 0.47942110896110535 + }, + "frames_300_400": { + "mse": 0.007453298196196556, + "psnr": 22.16885757446289, + "lpips": 0.4156564772129059 + }, + "frames_400_500": { + "mse": 0.01185040082782507, + "psnr": 19.716144561767578, + "lpips": 0.4827936291694641 + }, + "overall": { + "mse": 0.01220819540321827, + "psnr": 19.936885833740234, + "lpips": 0.4312909245491028 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000212_test__val_desert__1__000040/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000212_test__val_desert__1__000040/case.json new file mode 100644 index 0000000000000000000000000000000000000000..2ae187550bd876394632d3e0d1591ceb91497349 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000212_test__val_desert__1__000040/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "212", + "global_test_index": "212", + "dataset_index": "105", + "relative_path": "test/val_desert/1/000040.mp4", + "clip_id": "test/val_desert/1/000040", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000212, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 223.3563361520646, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000212_test__val_desert__1__000040/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000212_test__val_desert__1__000040/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..36266cabacfc1b8935f76e429c2a5f81d4b80fe2 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000212_test__val_desert__1__000040/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:caf8b98235d454e7badb4e8bccfc10ee3b8951bdc0bfa033858f6131596ab6d1 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000212_test__val_desert__1__000040/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000212_test__val_desert__1__000040/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..f7124d3bc6e5ada476aca77f2ad40ef173c4a810 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000212_test__val_desert__1__000040/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "212", + "global_test_index": "212", + "dataset_index": "105", + "relative_path": "test/val_desert/1/000040.mp4", + "clip_id": "test/val_desert/1/000040", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000212, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_desert/1/000040_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.007907472550868988, + "psnr": 21.131078720092773, + "lpips": 0.2386961132287979 + }, + "frames_100_200": { + "mse": 0.010370508767664433, + "psnr": 20.06398582458496, + "lpips": 0.29631754755973816 + }, + "frames_200_300": { + "mse": 0.010622474364936352, + "psnr": 19.98723030090332, + "lpips": 0.36939293146133423 + }, + "frames_300_400": { + "mse": 0.010182800702750683, + "psnr": 20.16885757446289, + "lpips": 0.3304908871650696 + }, + "frames_400_500": { + "mse": 0.02493419125676155, + "psnr": 16.94945526123047, + "lpips": 0.25662362575531006 + }, + "overall": { + "mse": 0.012803489342331886, + "psnr": 19.66012191772461, + "lpips": 0.2983042001724243 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000213_test__val_iceplains__2__000008/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000213_test__val_iceplains__2__000008/case.json new file mode 100644 index 0000000000000000000000000000000000000000..b7e05dc8434c15d2508ece6d9ffdcc5e03cfdaa7 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000213_test__val_iceplains__2__000008/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "213", + "global_test_index": "213", + "dataset_index": "267", + "relative_path": "test/val_iceplains/2/000008.mp4", + "clip_id": "test/val_iceplains/2/000008", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000213, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.74430554604623, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000213_test__val_iceplains__2__000008/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000213_test__val_iceplains__2__000008/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..a864a0990dc8f008742d02aac4eea495c244e2d9 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000213_test__val_iceplains__2__000008/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:16ca461a7bb9a3bd965988e61cc3dc2d8fa70b74ceff8ce9fb025507d91e34f0 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000213_test__val_iceplains__2__000008/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000213_test__val_iceplains__2__000008/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..cf74ab098997b0893b0fe3e8cb8b770524d8302d --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000213_test__val_iceplains__2__000008/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "213", + "global_test_index": "213", + "dataset_index": "267", + "relative_path": "test/val_iceplains/2/000008.mp4", + "clip_id": "test/val_iceplains/2/000008", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000213, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/2/000008_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.011368852108716965, + "psnr": 19.586759567260742, + "lpips": 0.12805984914302826 + }, + "frames_100_200": { + "mse": 0.00978352501988411, + "psnr": 20.861434936523438, + "lpips": 0.08861030638217926 + }, + "frames_200_300": { + "mse": 0.00234915129840374, + "psnr": 28.032678604125977, + "lpips": 0.020007295534014702 + }, + "frames_300_400": { + "mse": 0.019218238070607185, + "psnr": 18.826366424560547, + "lpips": 0.12574553489685059 + }, + "frames_400_500": { + "mse": 0.022967228665947914, + "psnr": 16.736881256103516, + "lpips": 0.1693171262741089 + }, + "overall": { + "mse": 0.013137399218976498, + "psnr": 20.80882453918457, + "lpips": 0.10634802281856537 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000214_test__val_desert__0__000034/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000214_test__val_desert__0__000034/case.json new file mode 100644 index 0000000000000000000000000000000000000000..df983782446a3ac678388fa03f531d4a259775ac --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000214_test__val_desert__0__000034/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "214", + "global_test_index": "214", + "dataset_index": "90", + "relative_path": "test/val_desert/0/000034.mp4", + "clip_id": "test/val_desert/0/000034", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000214, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.94803450093605, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000214_test__val_desert__0__000034/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000214_test__val_desert__0__000034/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..fbc93e9a9cd0cfdd5832cbdc9f1d5f5b57ebfb14 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000214_test__val_desert__0__000034/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c53e87caaf3973915d61b21beaa4cf7a3bd5ed76c1b3c1cd1eed117fb28455e8 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000214_test__val_desert__0__000034/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000214_test__val_desert__0__000034/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..e6259cab9756600328e9aa08a23264169f6c67bd --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000214_test__val_desert__0__000034/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "214", + "global_test_index": "214", + "dataset_index": "90", + "relative_path": "test/val_desert/0/000034.mp4", + "clip_id": "test/val_desert/0/000034", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000214, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_desert/0/000034_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0007848013192415237, + "psnr": 31.087034225463867, + "lpips": 0.010595793835818768 + }, + "frames_100_200": { + "mse": 0.0007784524350427091, + "psnr": 31.107187271118164, + "lpips": 0.011414812877774239 + }, + "frames_200_300": { + "mse": 0.0007631819462403655, + "psnr": 31.18096923828125, + "lpips": 0.010925225913524628 + }, + "frames_300_400": { + "mse": 0.0007979315705597401, + "psnr": 31.021486282348633, + "lpips": 0.011674851179122925 + }, + "frames_400_500": { + "mse": 0.005059526767581701, + "psnr": 24.089675903320312, + "lpips": 0.09869822859764099 + }, + "overall": { + "mse": 0.0016367788193747401, + "psnr": 29.6972713470459, + "lpips": 0.028661785647273064 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000215_test__val_plains__0__000023/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000215_test__val_plains__0__000023/case.json new file mode 100644 index 0000000000000000000000000000000000000000..412c9b70ead765751eed68fff6c5a501dcdc0c62 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000215_test__val_plains__0__000023/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "215", + "global_test_index": "215", + "dataset_index": "426", + "relative_path": "test/val_plains/0/000023.mp4", + "clip_id": "test/val_plains/0/000023", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000215, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.10591115790885, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000215_test__val_plains__0__000023/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000215_test__val_plains__0__000023/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..57fb361ccd6bab42784fbe88394fecb0dade6f4f --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000215_test__val_plains__0__000023/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fd07be9265d1e5a01545a11b67c08d77c29197815f67c66c968103abd59d8b3d +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000215_test__val_plains__0__000023/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000215_test__val_plains__0__000023/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..295164c9a7c20e50092ba68451fd083f80fa858b --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000215_test__val_plains__0__000023/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "215", + "global_test_index": "215", + "dataset_index": "426", + "relative_path": "test/val_plains/0/000023.mp4", + "clip_id": "test/val_plains/0/000023", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000215, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_plains/0/000023_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.005009747110307217, + "psnr": 23.066930770874023, + "lpips": 0.32579460740089417 + }, + "frames_100_200": { + "mse": 0.00803834106773138, + "psnr": 21.067224502563477, + "lpips": 0.4423395097255707 + }, + "frames_200_300": { + "mse": 0.00698363920673728, + "psnr": 21.639070510864258, + "lpips": 0.4839877784252167 + }, + "frames_300_400": { + "mse": 0.006516131106764078, + "psnr": 22.057466506958008, + "lpips": 0.4287278354167938 + }, + "frames_400_500": { + "mse": 0.007474583573639393, + "psnr": 21.407428741455078, + "lpips": 0.45720764994621277 + }, + "overall": { + "mse": 0.00680448766797781, + "psnr": 21.847623825073242, + "lpips": 0.42761150002479553 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000216_test__val_savanna__3__000016/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000216_test__val_savanna__3__000016/case.json new file mode 100644 index 0000000000000000000000000000000000000000..676f7f06a5e667d7980ba45a1bc6196f27ed4fd8 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000216_test__val_savanna__3__000016/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "216", + "global_test_index": "216", + "dataset_index": "168", + "relative_path": "test/val_savanna/3/000016.mp4", + "clip_id": "test/val_savanna/3/000016", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000216, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 223.50121226604097, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000216_test__val_savanna__3__000016/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000216_test__val_savanna__3__000016/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..a438aca9b1956a8ae39812839d01131459cca0f5 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000216_test__val_savanna__3__000016/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8a87db6ba4ccfe5277c1803f5632ae5b492554c0807e55c088c528a622aed778 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000216_test__val_savanna__3__000016/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000216_test__val_savanna__3__000016/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..0aa2a029e539d3947ad98b70beafe508d51396ac --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000216_test__val_savanna__3__000016/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "216", + "global_test_index": "216", + "dataset_index": "168", + "relative_path": "test/val_savanna/3/000016.mp4", + "clip_id": "test/val_savanna/3/000016", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000216, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_savanna/3/000016_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.026863619685173035, + "psnr": 16.149721145629883, + "lpips": 0.18918904662132263 + }, + "frames_100_200": { + "mse": 0.05099629983305931, + "psnr": 13.415950775146484, + "lpips": 0.31994760036468506 + }, + "frames_200_300": { + "mse": 0.07149861007928848, + "psnr": 11.676440238952637, + "lpips": 0.4478011727333069 + }, + "frames_300_400": { + "mse": 0.09781863540410995, + "psnr": 10.40023422241211, + "lpips": 0.4426145851612091 + }, + "frames_400_500": { + "mse": 0.09060297906398773, + "psnr": 10.643257141113281, + "lpips": 0.5939880609512329 + }, + "overall": { + "mse": 0.06755602359771729, + "psnr": 12.457119941711426, + "lpips": 0.3987080752849579 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000217_test__val_savanna__3__000040/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000217_test__val_savanna__3__000040/case.json new file mode 100644 index 0000000000000000000000000000000000000000..5ab09dc26d37bde44b237e8e30530ab9a6ec6ea6 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000217_test__val_savanna__3__000040/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "217", + "global_test_index": "217", + "dataset_index": "218", + "relative_path": "test/val_savanna/3/000040.mp4", + "clip_id": "test/val_savanna/3/000040", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000217, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.3934717969969, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000217_test__val_savanna__3__000040/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000217_test__val_savanna__3__000040/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..29510a881e16059176a28b2c920c2224019660ad --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000217_test__val_savanna__3__000040/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f1360c92afdbab9d84b992f9e244af9619c4f0faf1d68a8618411e2845466cd6 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000217_test__val_savanna__3__000040/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000217_test__val_savanna__3__000040/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..71b66bf89cff122b0d3fb712abe49bca2ccfe99a --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000217_test__val_savanna__3__000040/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "217", + "global_test_index": "217", + "dataset_index": "218", + "relative_path": "test/val_savanna/3/000040.mp4", + "clip_id": "test/val_savanna/3/000040", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000217, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_savanna/3/000040_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0037043390329927206, + "psnr": 24.98428726196289, + "lpips": 0.038296688348054886 + }, + "frames_100_200": { + "mse": 0.013170207850635052, + "psnr": 21.55585289001465, + "lpips": 0.0646873340010643 + }, + "frames_200_300": { + "mse": 0.021473271772265434, + "psnr": 18.07326889038086, + "lpips": 0.12488950043916702 + }, + "frames_300_400": { + "mse": 0.030079009011387825, + "psnr": 15.497206687927246, + "lpips": 0.3077736496925354 + }, + "frames_400_500": { + "mse": 0.01462413091212511, + "psnr": 18.914770126342773, + "lpips": 0.3587990999221802 + }, + "overall": { + "mse": 0.016610193997621536, + "psnr": 19.805076599121094, + "lpips": 0.17888924479484558 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000218_test__val_plains__0__000032/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000218_test__val_plains__0__000032/case.json new file mode 100644 index 0000000000000000000000000000000000000000..bf0d37b1a8ac8eda9d34d0e96cc8f95b0fa5228c --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000218_test__val_plains__0__000032/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "218", + "global_test_index": "218", + "dataset_index": "445", + "relative_path": "test/val_plains/0/000032.mp4", + "clip_id": "test/val_plains/0/000032", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000218, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.38628776394762, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000218_test__val_plains__0__000032/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000218_test__val_plains__0__000032/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..18e84f7b2168fa890c73341be3408ddfcbc4e96f --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000218_test__val_plains__0__000032/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5149bf957511b66cbb697a7aca314fbfcab60432b342a8c8da3c0dd887e9610e +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000218_test__val_plains__0__000032/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000218_test__val_plains__0__000032/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..2c49b1f872df8aa0e580d4217cb1962eddfeed32 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000218_test__val_plains__0__000032/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "218", + "global_test_index": "218", + "dataset_index": "445", + "relative_path": "test/val_plains/0/000032.mp4", + "clip_id": "test/val_plains/0/000032", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000218, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_plains/0/000032_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0025487288367003202, + "psnr": 28.386049270629883, + "lpips": 0.030475491657853127 + }, + "frames_100_200": { + "mse": 0.0013406354701146483, + "psnr": 29.680116653442383, + "lpips": 0.01946215331554413 + }, + "frames_200_300": { + "mse": 0.0012692579766735435, + "psnr": 30.075414657592773, + "lpips": 0.018258564174175262 + }, + "frames_300_400": { + "mse": 0.001018040580675006, + "psnr": 30.123506546020508, + "lpips": 0.016694944351911545 + }, + "frames_400_500": { + "mse": 0.005271767266094685, + "psnr": 25.204795837402344, + "lpips": 0.08140264451503754 + }, + "overall": { + "mse": 0.0022896858863532543, + "psnr": 28.69397735595703, + "lpips": 0.03325876221060753 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000219_test__val_desert__3__000019/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000219_test__val_desert__3__000019/case.json new file mode 100644 index 0000000000000000000000000000000000000000..daa26941a2dd20d0505240cf926aa9203ed967d2 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000219_test__val_desert__3__000019/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "219", + "global_test_index": "219", + "dataset_index": "48", + "relative_path": "test/val_desert/3/000019.mp4", + "clip_id": "test/val_desert/3/000019", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000219, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.91231207107194, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000219_test__val_desert__3__000019/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000219_test__val_desert__3__000019/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..b06b2efd1a81fe0fc32ed748c8280d9f8a2dd7c1 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000219_test__val_desert__3__000019/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5dff3be26ac9c5571f5ca8b764e1622a6379b6552a7ae901e504eb68c9af45bb +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000219_test__val_desert__3__000019/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000219_test__val_desert__3__000019/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..6cbf45408772b0728999ed175a7412d07928cc45 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000219_test__val_desert__3__000019/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "219", + "global_test_index": "219", + "dataset_index": "48", + "relative_path": "test/val_desert/3/000019.mp4", + "clip_id": "test/val_desert/3/000019", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000219, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_desert/3/000019_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0032459525391459465, + "psnr": 26.680139541625977, + "lpips": 0.046321090310811996 + }, + "frames_100_200": { + "mse": 0.006856989115476608, + "psnr": 23.111083984375, + "lpips": 0.10847728699445724 + }, + "frames_200_300": { + "mse": 0.0044417851604521275, + "psnr": 26.177892684936523, + "lpips": 0.07233352959156036 + }, + "frames_300_400": { + "mse": 0.006881745997816324, + "psnr": 25.32361602783203, + "lpips": 0.08460855484008789 + }, + "frames_400_500": { + "mse": 0.020516840741038322, + "psnr": 17.926252365112305, + "lpips": 0.30637049674987793 + }, + "overall": { + "mse": 0.008388662710785866, + "psnr": 23.843795776367188, + "lpips": 0.12362217903137207 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000220_test__val_desert__2__000039/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000220_test__val_desert__2__000039/case.json new file mode 100644 index 0000000000000000000000000000000000000000..16b65a6a0adbe677830ed3df3cce0100ccfe7a6d --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000220_test__val_desert__2__000039/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "220", + "global_test_index": "220", + "dataset_index": "104", + "relative_path": "test/val_desert/2/000039.mp4", + "clip_id": "test/val_desert/2/000039", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000220, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 223.62837327702437, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000220_test__val_desert__2__000039/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000220_test__val_desert__2__000039/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..c16bc1da7e9f689dc0fbde459179816a2486b5f6 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000220_test__val_desert__2__000039/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3deed2d6e445587c7d34c824577e6123ccd6f1d79eef2fa415a5485cec5a78c8 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000220_test__val_desert__2__000039/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000220_test__val_desert__2__000039/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..e326c70157d427c317e04f412325761f5a89b021 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000220_test__val_desert__2__000039/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "220", + "global_test_index": "220", + "dataset_index": "104", + "relative_path": "test/val_desert/2/000039.mp4", + "clip_id": "test/val_desert/2/000039", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000220, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_desert/2/000039_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0039498452097177505, + "psnr": 25.733016967773438, + "lpips": 0.1068490594625473 + }, + "frames_100_200": { + "mse": 0.012621057219803333, + "psnr": 19.27391815185547, + "lpips": 0.32654082775115967 + }, + "frames_200_300": { + "mse": 0.016743892803788185, + "psnr": 18.181137084960938, + "lpips": 0.46695148944854736 + }, + "frames_300_400": { + "mse": 0.014495103619992733, + "psnr": 18.902873992919922, + "lpips": 0.4378800690174103 + }, + "frames_400_500": { + "mse": 0.013673527166247368, + "psnr": 19.438392639160156, + "lpips": 0.4087081253528595 + }, + "overall": { + "mse": 0.012296684086322784, + "psnr": 20.305870056152344, + "lpips": 0.34938591718673706 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000221_test__val_savanna__2__000035/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000221_test__val_savanna__2__000035/case.json new file mode 100644 index 0000000000000000000000000000000000000000..cf3917e06cf95af164b413bcbf581ed22db27e8b --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000221_test__val_savanna__2__000035/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "221", + "global_test_index": "221", + "dataset_index": "206", + "relative_path": "test/val_savanna/2/000035.mp4", + "clip_id": "test/val_savanna/2/000035", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000221, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.03134497103747, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000221_test__val_savanna__2__000035/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000221_test__val_savanna__2__000035/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..ce08365ce8518a1001c157dff32db92186509fac --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000221_test__val_savanna__2__000035/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:70cb2536cc8445b6df13ae5cd86e94f19243657b2345a7ef603d5838ca04a8fc +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000221_test__val_savanna__2__000035/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000221_test__val_savanna__2__000035/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..80b8823728c7521ed2b7920c746dadcdc0979c94 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000221_test__val_savanna__2__000035/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "221", + "global_test_index": "221", + "dataset_index": "206", + "relative_path": "test/val_savanna/2/000035.mp4", + "clip_id": "test/val_savanna/2/000035", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000221, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_savanna/2/000035_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.012610059231519699, + "psnr": 19.189489364624023, + "lpips": 0.27420884370803833 + }, + "frames_100_200": { + "mse": 0.015337416902184486, + "psnr": 18.22705841064453, + "lpips": 0.2749364376068115 + }, + "frames_200_300": { + "mse": 0.00928487814962864, + "psnr": 21.057947158813477, + "lpips": 0.3538711965084076 + }, + "frames_300_400": { + "mse": 0.007784114684909582, + "psnr": 21.44344711303711, + "lpips": 0.3885710537433624 + }, + "frames_400_500": { + "mse": 0.01808890327811241, + "psnr": 17.823453903198242, + "lpips": 0.4521767795085907 + }, + "overall": { + "mse": 0.012621073983609676, + "psnr": 19.548280715942383, + "lpips": 0.34875285625457764 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000222_test__val_savanna__1__000045/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000222_test__val_savanna__1__000045/case.json new file mode 100644 index 0000000000000000000000000000000000000000..366d3b496052d6e095d87998f4adca3eedd5d81b --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000222_test__val_savanna__1__000045/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "222", + "global_test_index": "222", + "dataset_index": "228", + "relative_path": "test/val_savanna/1/000045.mp4", + "clip_id": "test/val_savanna/1/000045", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000222, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.51808174396865, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000222_test__val_savanna__1__000045/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000222_test__val_savanna__1__000045/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..233f4bb140be88acc8e5c71419c3ab5b794a0295 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000222_test__val_savanna__1__000045/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:68f5fa862a30cca32c77262500513f89b497d7ba98874f4a9f1878636e34aecd +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000222_test__val_savanna__1__000045/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000222_test__val_savanna__1__000045/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..d075197949cc3097e968f18831238a84762501b4 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000222_test__val_savanna__1__000045/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "222", + "global_test_index": "222", + "dataset_index": "228", + "relative_path": "test/val_savanna/1/000045.mp4", + "clip_id": "test/val_savanna/1/000045", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000222, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_savanna/1/000045_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.013780869543552399, + "psnr": 18.910694122314453, + "lpips": 0.17347244918346405 + }, + "frames_100_200": { + "mse": 0.016987591981887817, + "psnr": 18.003807067871094, + "lpips": 0.1594371646642685 + }, + "frames_200_300": { + "mse": 0.02174699492752552, + "psnr": 16.972164154052734, + "lpips": 0.21326705813407898 + }, + "frames_300_400": { + "mse": 0.013865390792489052, + "psnr": 18.9025936126709, + "lpips": 0.2938325107097626 + }, + "frames_400_500": { + "mse": 0.016370519995689392, + "psnr": 18.19989013671875, + "lpips": 0.3556690514087677 + }, + "overall": { + "mse": 0.016550274565815926, + "psnr": 18.197830200195312, + "lpips": 0.23913566768169403 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000223_test__val_savanna__2__000033/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000223_test__val_savanna__2__000033/case.json new file mode 100644 index 0000000000000000000000000000000000000000..a039644111649ebaa061ece37cdcc2ff45447950 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000223_test__val_savanna__2__000033/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "223", + "global_test_index": "223", + "dataset_index": "200", + "relative_path": "test/val_savanna/2/000033.mp4", + "clip_id": "test/val_savanna/2/000033", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000223, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.8623388078995, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000223_test__val_savanna__2__000033/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000223_test__val_savanna__2__000033/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..25941e260a81d0b88e333c10cd4d169a7cb4d939 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000223_test__val_savanna__2__000033/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d81b2dc53e819888b433fbe1b6ab34d7e7fc42ac0a8b2f7e41523022b13a7fe9 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000223_test__val_savanna__2__000033/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000223_test__val_savanna__2__000033/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..541b2178aca8d2df2e6a2258579101ba9638fb0d --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000223_test__val_savanna__2__000033/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "223", + "global_test_index": "223", + "dataset_index": "200", + "relative_path": "test/val_savanna/2/000033.mp4", + "clip_id": "test/val_savanna/2/000033", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000223, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_savanna/2/000033_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0032833879813551903, + "psnr": 24.981342315673828, + "lpips": 0.21568593382835388 + }, + "frames_100_200": { + "mse": 0.017852377146482468, + "psnr": 18.84058380126953, + "lpips": 0.41115668416023254 + }, + "frames_200_300": { + "mse": 0.004905843175947666, + "psnr": 23.15727996826172, + "lpips": 0.34452760219573975 + }, + "frames_300_400": { + "mse": 0.03265206515789032, + "psnr": 15.291728973388672, + "lpips": 0.4253012537956238 + }, + "frames_400_500": { + "mse": 0.05408202484250069, + "psnr": 13.213866233825684, + "lpips": 0.36859428882598877 + }, + "overall": { + "mse": 0.022555140778422356, + "psnr": 19.096960067749023, + "lpips": 0.35305312275886536 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000224_test__val_desert__2__000024/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000224_test__val_desert__2__000024/case.json new file mode 100644 index 0000000000000000000000000000000000000000..a8943fce434611e3bd3be9b9ab57c384d1594f27 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000224_test__val_desert__2__000024/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "224", + "global_test_index": "224", + "dataset_index": "62", + "relative_path": "test/val_desert/2/000024.mp4", + "clip_id": "test/val_desert/2/000024", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000224, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.25552737689577, + "peak_generation_vram_bytes": 107565475840, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000224_test__val_desert__2__000024/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000224_test__val_desert__2__000024/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..a54d5c4e181bc4ff562a63f456818a5f034254c3 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000224_test__val_desert__2__000024/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c2d65cddeceb436a4103aed16e2ed7fb4784d7ab13920725fcc00d9097b685af +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000224_test__val_desert__2__000024/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000224_test__val_desert__2__000024/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..98c4cd48e08d47214693d9457a5a5f9e6e232ccd --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000224_test__val_desert__2__000024/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "224", + "global_test_index": "224", + "dataset_index": "62", + "relative_path": "test/val_desert/2/000024.mp4", + "clip_id": "test/val_desert/2/000024", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000224, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_desert/2/000024_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.006686240900307894, + "psnr": 21.818944931030273, + "lpips": 0.25288447737693787 + }, + "frames_100_200": { + "mse": 0.012995000928640366, + "psnr": 19.077226638793945, + "lpips": 0.3439461588859558 + }, + "frames_200_300": { + "mse": 0.013408638536930084, + "psnr": 19.1705322265625, + "lpips": 0.36335909366607666 + }, + "frames_300_400": { + "mse": 0.023025697097182274, + "psnr": 16.746015548706055, + "lpips": 0.44068634510040283 + }, + "frames_400_500": { + "mse": 0.020927133038640022, + "psnr": 17.327285766601562, + "lpips": 0.45361077785491943 + }, + "overall": { + "mse": 0.015408542938530445, + "psnr": 18.828001022338867, + "lpips": 0.3708973824977875 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000225_test__val_desert__1__000047/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000225_test__val_desert__1__000047/case.json new file mode 100644 index 0000000000000000000000000000000000000000..068684f43d45a8e7d3b1b9d5868ffae42729c903 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000225_test__val_desert__1__000047/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "225", + "global_test_index": "225", + "dataset_index": "121", + "relative_path": "test/val_desert/1/000047.mp4", + "clip_id": "test/val_desert/1/000047", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000225, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.778245427995, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000225_test__val_desert__1__000047/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000225_test__val_desert__1__000047/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..a2a892eb946c0b2c2db90379b86f2df5d6fb3201 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000225_test__val_desert__1__000047/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e68bbb6571c006b8c83d9c3284fbcddf4120521445417adbd14211e586845e4f +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000225_test__val_desert__1__000047/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000225_test__val_desert__1__000047/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..8bba701431042eba5ca79d5e03d3217d49e05377 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000225_test__val_desert__1__000047/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "225", + "global_test_index": "225", + "dataset_index": "121", + "relative_path": "test/val_desert/1/000047.mp4", + "clip_id": "test/val_desert/1/000047", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000225, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_desert/1/000047_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0059392149560153484, + "psnr": 22.351194381713867, + "lpips": 0.25051647424697876 + }, + "frames_100_200": { + "mse": 0.03076557070016861, + "psnr": 16.585145950317383, + "lpips": 0.5084598064422607 + }, + "frames_200_300": { + "mse": 0.030330417677760124, + "psnr": 15.486842155456543, + "lpips": 0.4551944434642792 + }, + "frames_300_400": { + "mse": 0.03492484241724014, + "psnr": 15.237181663513184, + "lpips": 0.4953426420688629 + }, + "frames_400_500": { + "mse": 0.05202120915055275, + "psnr": 13.643117904663086, + "lpips": 0.49255695939064026 + }, + "overall": { + "mse": 0.030796248465776443, + "psnr": 16.660696029663086, + "lpips": 0.44041404128074646 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000226_test__val_savanna__0__000007/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000226_test__val_savanna__0__000007/case.json new file mode 100644 index 0000000000000000000000000000000000000000..e18838bc84cb9bedcd22eb7ad88d406119c5e800 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000226_test__val_savanna__0__000007/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "226", + "global_test_index": "226", + "dataset_index": "147", + "relative_path": "test/val_savanna/0/000007.mp4", + "clip_id": "test/val_savanna/0/000007", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000226, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.72644292400219, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000226_test__val_savanna__0__000007/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000226_test__val_savanna__0__000007/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..b1783a921cd0702955ff802e5ac4b6f46e9fe48f --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000226_test__val_savanna__0__000007/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d5216c662c55fafcfde670cd4e44ca346cf5e5af740708a012559b649d4b3b04 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000226_test__val_savanna__0__000007/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000226_test__val_savanna__0__000007/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..3ad2580d25dd6be9912d5754ca9486c5c002fd7a --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000226_test__val_savanna__0__000007/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "226", + "global_test_index": "226", + "dataset_index": "147", + "relative_path": "test/val_savanna/0/000007.mp4", + "clip_id": "test/val_savanna/0/000007", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000226, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_savanna/0/000007_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.004082139115780592, + "psnr": 23.976728439331055, + "lpips": 0.24713577330112457 + }, + "frames_100_200": { + "mse": 0.012760141864418983, + "psnr": 20.300064086914062, + "lpips": 0.343179851770401 + }, + "frames_200_300": { + "mse": 0.03231477737426758, + "psnr": 15.344454765319824, + "lpips": 0.35483038425445557 + }, + "frames_300_400": { + "mse": 0.021045098081231117, + "psnr": 16.944293975830078, + "lpips": 0.3302120566368103 + }, + "frames_400_500": { + "mse": 0.04597921669483185, + "psnr": 14.309688568115234, + "lpips": 0.3812545835971832 + }, + "overall": { + "mse": 0.02323627471923828, + "psnr": 18.175045013427734, + "lpips": 0.3313225209712982 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000227_test__val_savanna__2__000006/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000227_test__val_savanna__2__000006/case.json new file mode 100644 index 0000000000000000000000000000000000000000..40df8726fdd7748879f27cfafa73d22783078853 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000227_test__val_savanna__2__000006/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "227", + "global_test_index": "227", + "dataset_index": "146", + "relative_path": "test/val_savanna/2/000006.mp4", + "clip_id": "test/val_savanna/2/000006", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000227, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.90007369604427, + "peak_generation_vram_bytes": 107565475840, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000227_test__val_savanna__2__000006/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000227_test__val_savanna__2__000006/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..358869e95797d43a23c61df11b3a6b90f18a0023 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000227_test__val_savanna__2__000006/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:95c6aaf89fb17752fe6c2f71d4110fbfe8950b4f2e4a7a59bb3c5a9745b384aa +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000227_test__val_savanna__2__000006/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000227_test__val_savanna__2__000006/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..306a812091090ef3fae88e6b1ea29827bb042c10 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000227_test__val_savanna__2__000006/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "227", + "global_test_index": "227", + "dataset_index": "146", + "relative_path": "test/val_savanna/2/000006.mp4", + "clip_id": "test/val_savanna/2/000006", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000227, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_savanna/2/000006_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.01786040887236595, + "psnr": 19.670726776123047, + "lpips": 0.40614935755729675 + }, + "frames_100_200": { + "mse": 0.029671411961317062, + "psnr": 15.489340782165527, + "lpips": 0.4162171185016632 + }, + "frames_200_300": { + "mse": 0.007630201056599617, + "psnr": 22.319368362426758, + "lpips": 0.36865779757499695 + }, + "frames_300_400": { + "mse": 0.01284707710146904, + "psnr": 20.870018005371094, + "lpips": 0.4014800190925598 + }, + "frames_400_500": { + "mse": 0.06561490893363953, + "psnr": 13.204656600952148, + "lpips": 0.49279916286468506 + }, + "overall": { + "mse": 0.0267248023301363, + "psnr": 18.310821533203125, + "lpips": 0.41706064343452454 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000228_test__val_savanna__1__000043/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000228_test__val_savanna__1__000043/case.json new file mode 100644 index 0000000000000000000000000000000000000000..dd2a8d6a6a51a67726d81d67b822b64b08b05e17 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000228_test__val_savanna__1__000043/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "228", + "global_test_index": "228", + "dataset_index": "225", + "relative_path": "test/val_savanna/1/000043.mp4", + "clip_id": "test/val_savanna/1/000043", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000228, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.03342578304, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000228_test__val_savanna__1__000043/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000228_test__val_savanna__1__000043/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..82582fc2441fd79df9ac0d761f016e1109a2db28 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000228_test__val_savanna__1__000043/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b48eb8fa7c7f395dcbb61b51adaab5604a5fa44408ca4fabc48952711cfc00cb +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000228_test__val_savanna__1__000043/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000228_test__val_savanna__1__000043/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..e6697c848483abfd2e818577176641b64b29381c --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000228_test__val_savanna__1__000043/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "228", + "global_test_index": "228", + "dataset_index": "225", + "relative_path": "test/val_savanna/1/000043.mp4", + "clip_id": "test/val_savanna/1/000043", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000228, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_savanna/1/000043_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.008341227658092976, + "psnr": 21.557336807250977, + "lpips": 0.32760941982269287 + }, + "frames_100_200": { + "mse": 0.009272603318095207, + "psnr": 20.76740264892578, + "lpips": 0.3620641827583313 + }, + "frames_200_300": { + "mse": 0.014598377048969269, + "psnr": 18.692276000976562, + "lpips": 0.46308985352516174 + }, + "frames_300_400": { + "mse": 0.00761282816529274, + "psnr": 21.321395874023438, + "lpips": 0.4128775894641876 + }, + "frames_400_500": { + "mse": 0.006337843369692564, + "psnr": 22.142227172851562, + "lpips": 0.4123630225658417 + }, + "overall": { + "mse": 0.009232576005160809, + "psnr": 20.89612579345703, + "lpips": 0.395600825548172 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000229_test__val_iceplains__3__000027/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000229_test__val_iceplains__3__000027/case.json new file mode 100644 index 0000000000000000000000000000000000000000..6d74a41d5a7277bdb02eb14eea71c9b8f4044a39 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000229_test__val_iceplains__3__000027/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "229", + "global_test_index": "229", + "dataset_index": "313", + "relative_path": "test/val_iceplains/3/000027.mp4", + "clip_id": "test/val_iceplains/3/000027", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000229, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 226.11218560300767, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000229_test__val_iceplains__3__000027/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000229_test__val_iceplains__3__000027/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..c7c878aa68d11a29873cd64c2e5e190764206af3 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000229_test__val_iceplains__3__000027/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:80b36dd44191b6c16f712af5e2f53afc07aaab8c88debf4920d4bf8436eab3d7 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000229_test__val_iceplains__3__000027/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000229_test__val_iceplains__3__000027/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..b0ab9ce8c018658a2eeab37984018e95745bd5af --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000229_test__val_iceplains__3__000027/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "229", + "global_test_index": "229", + "dataset_index": "313", + "relative_path": "test/val_iceplains/3/000027.mp4", + "clip_id": "test/val_iceplains/3/000027", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000229, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/3/000027_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0007577044307254255, + "psnr": 31.216196060180664, + "lpips": 0.010337477549910545 + }, + "frames_100_200": { + "mse": 0.001993660582229495, + "psnr": 29.226713180541992, + "lpips": 0.01854749582707882 + }, + "frames_200_300": { + "mse": 0.0009766280418261886, + "psnr": 30.132640838623047, + "lpips": 0.014864000491797924 + }, + "frames_300_400": { + "mse": 0.0012513601686805487, + "psnr": 29.696739196777344, + "lpips": 0.0171724371612072 + }, + "frames_400_500": { + "mse": 0.0033378503285348415, + "psnr": 27.193225860595703, + "lpips": 0.030334828421473503 + }, + "overall": { + "mse": 0.001663440722040832, + "psnr": 29.493101119995117, + "lpips": 0.018251247704029083 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000230_test__val_iceplains__0__000042/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000230_test__val_iceplains__0__000042/case.json new file mode 100644 index 0000000000000000000000000000000000000000..d864f3c743da1a0f9ba45a1a682c94d555c868c3 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000230_test__val_iceplains__0__000042/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "230", + "global_test_index": "230", + "dataset_index": "354", + "relative_path": "test/val_iceplains/0/000042.mp4", + "clip_id": "test/val_iceplains/0/000042", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000230, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.5463069069665, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000230_test__val_iceplains__0__000042/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000230_test__val_iceplains__0__000042/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..31cbfedda523e53640a6292edf17a8cc295036a6 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000230_test__val_iceplains__0__000042/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3b5e678a0004b932303fefccee40a836d5904542ee590206ca606ac4d5b6af4a +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000230_test__val_iceplains__0__000042/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000230_test__val_iceplains__0__000042/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..28dcf4fad55c98d6455243223b03c9e572a96d7f --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000230_test__val_iceplains__0__000042/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "230", + "global_test_index": "230", + "dataset_index": "354", + "relative_path": "test/val_iceplains/0/000042.mp4", + "clip_id": "test/val_iceplains/0/000042", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000230, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/0/000042_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.020596470683813095, + "psnr": 17.501020431518555, + "lpips": 0.19218532741069794 + }, + "frames_100_200": { + "mse": 0.03185220807790756, + "psnr": 15.228421211242676, + "lpips": 0.2781713604927063 + }, + "frames_200_300": { + "mse": 0.049326859414577484, + "psnr": 13.217507362365723, + "lpips": 0.3094511330127716 + }, + "frames_300_400": { + "mse": 0.04289847984910011, + "psnr": 14.705426216125488, + "lpips": 0.23844078183174133 + }, + "frames_400_500": { + "mse": 0.012408161535859108, + "psnr": 20.263425827026367, + "lpips": 0.09051792323589325 + }, + "overall": { + "mse": 0.03141643479466438, + "psnr": 16.18316078186035, + "lpips": 0.22175332903862 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000231_test__val_iceplains__1__000048/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000231_test__val_iceplains__1__000048/case.json new file mode 100644 index 0000000000000000000000000000000000000000..aec2fc8c82a99d06f273596af8aaf22ff5582a7b --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000231_test__val_iceplains__1__000048/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "231", + "global_test_index": "231", + "dataset_index": "367", + "relative_path": "test/val_iceplains/1/000048.mp4", + "clip_id": "test/val_iceplains/1/000048", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000231, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.34285371901933, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000231_test__val_iceplains__1__000048/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000231_test__val_iceplains__1__000048/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..fe196501f8632819974cbed5e2f7308e911206a9 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000231_test__val_iceplains__1__000048/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:859031083837cb2cab47e8144d5f2e1acb627e9be9e3c7948144c97adbfbe861 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000231_test__val_iceplains__1__000048/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000231_test__val_iceplains__1__000048/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..695a12a33214871e95ac81f5e20de0d4f0d5a0ad --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000231_test__val_iceplains__1__000048/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "231", + "global_test_index": "231", + "dataset_index": "367", + "relative_path": "test/val_iceplains/1/000048.mp4", + "clip_id": "test/val_iceplains/1/000048", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000231, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/1/000048_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.007373289670795202, + "psnr": 21.600696563720703, + "lpips": 0.12740756571292877 + }, + "frames_100_200": { + "mse": 0.005658411420881748, + "psnr": 23.392776489257812, + "lpips": 0.2025112509727478 + }, + "frames_200_300": { + "mse": 0.05570266768336296, + "psnr": 15.223464012145996, + "lpips": 0.4187871515750885 + }, + "frames_300_400": { + "mse": 0.004357617348432541, + "psnr": 23.622314453125, + "lpips": 0.3306454122066498 + }, + "frames_400_500": { + "mse": 0.005496360827237368, + "psnr": 23.145605087280273, + "lpips": 0.34170857071876526 + }, + "overall": { + "mse": 0.01571767032146454, + "psnr": 21.396968841552734, + "lpips": 0.28421199321746826 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000232_test__val_iceplains__3__000020/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000232_test__val_iceplains__3__000020/case.json new file mode 100644 index 0000000000000000000000000000000000000000..d9a14e5e090a8c0ffb6e14df5217ea701baa1b40 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000232_test__val_iceplains__3__000020/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "232", + "global_test_index": "232", + "dataset_index": "297", + "relative_path": "test/val_iceplains/3/000020.mp4", + "clip_id": "test/val_iceplains/3/000020", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000232, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.35978866601363, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000232_test__val_iceplains__3__000020/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000232_test__val_iceplains__3__000020/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..40e15de519a17be29be0dcc66b78f93fbeda3b18 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000232_test__val_iceplains__3__000020/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2b1e2b739261df12b1a9362478325e777a8d05aa0e930490c089e0f73ccaec68 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000232_test__val_iceplains__3__000020/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000232_test__val_iceplains__3__000020/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..d8d44ec7d066aff4a983be8ca99d6cd22c1b0c3a --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000232_test__val_iceplains__3__000020/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "232", + "global_test_index": "232", + "dataset_index": "297", + "relative_path": "test/val_iceplains/3/000020.mp4", + "clip_id": "test/val_iceplains/3/000020", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000232, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/3/000020_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.009031794033944607, + "psnr": 20.74407196044922, + "lpips": 0.17275716364383698 + }, + "frames_100_200": { + "mse": 0.010991567745804787, + "psnr": 19.791431427001953, + "lpips": 0.1957763433456421 + }, + "frames_200_300": { + "mse": 0.015356646850705147, + "psnr": 18.27318000793457, + "lpips": 0.22954607009887695 + }, + "frames_300_400": { + "mse": 0.018259327858686447, + "psnr": 17.549409866333008, + "lpips": 0.33621925115585327 + }, + "frames_400_500": { + "mse": 0.03224991634488106, + "psnr": 15.054513931274414, + "lpips": 0.3602338135242462 + }, + "overall": { + "mse": 0.017177850008010864, + "psnr": 18.282520294189453, + "lpips": 0.2589065134525299 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000233_test__val_desert__1__000048/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000233_test__val_desert__1__000048/case.json new file mode 100644 index 0000000000000000000000000000000000000000..cd3734d9e00d3d28b91cd08d7c4d9d357372ab21 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000233_test__val_desert__1__000048/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "233", + "global_test_index": "233", + "dataset_index": "124", + "relative_path": "test/val_desert/1/000048.mp4", + "clip_id": "test/val_desert/1/000048", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000233, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.18072486191522, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000233_test__val_desert__1__000048/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000233_test__val_desert__1__000048/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..52723b2ae7a0ec0bc3fe000744193dbe83d59afe --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000233_test__val_desert__1__000048/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1d150916787c145a363bc904045dfe9cb767a40cd16093bd36990be2b2a9dfdf +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000233_test__val_desert__1__000048/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000233_test__val_desert__1__000048/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..fa1a0332246eab4db58a798553c3b30c449611b2 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000233_test__val_desert__1__000048/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "233", + "global_test_index": "233", + "dataset_index": "124", + "relative_path": "test/val_desert/1/000048.mp4", + "clip_id": "test/val_desert/1/000048", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000233, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_desert/1/000048_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.006900029722601175, + "psnr": 21.752634048461914, + "lpips": 0.2074078917503357 + }, + "frames_100_200": { + "mse": 0.007397327572107315, + "psnr": 21.39401626586914, + "lpips": 0.21922491490840912 + }, + "frames_200_300": { + "mse": 0.004365419037640095, + "psnr": 24.363842010498047, + "lpips": 0.09880770742893219 + }, + "frames_300_400": { + "mse": 0.009004533290863037, + "psnr": 20.782737731933594, + "lpips": 0.20528241991996765 + }, + "frames_400_500": { + "mse": 0.01670871116220951, + "psnr": 17.872821807861328, + "lpips": 0.3117143511772156 + }, + "overall": { + "mse": 0.008875204250216484, + "psnr": 21.233211517333984, + "lpips": 0.20848745107650757 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000234_test__val_savanna__2__000017/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000234_test__val_savanna__2__000017/case.json new file mode 100644 index 0000000000000000000000000000000000000000..1de8151d7e2c9f8b83eea0cc4ab6b78e62e8d8d1 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000234_test__val_savanna__2__000017/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "234", + "global_test_index": "234", + "dataset_index": "171", + "relative_path": "test/val_savanna/2/000017.mp4", + "clip_id": "test/val_savanna/2/000017", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000234, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.84217477508355, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000234_test__val_savanna__2__000017/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000234_test__val_savanna__2__000017/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..0a7923934620b35a6af56b9191c14e507336b944 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000234_test__val_savanna__2__000017/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:695d0a4d96e2d7052688bf206fed91a2d23e90889b07fae269f5c46b512e2f8b +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000234_test__val_savanna__2__000017/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000234_test__val_savanna__2__000017/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..eacddfb85c9eae1085b26d2f0e6041fa958e37ed --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000234_test__val_savanna__2__000017/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "234", + "global_test_index": "234", + "dataset_index": "171", + "relative_path": "test/val_savanna/2/000017.mp4", + "clip_id": "test/val_savanna/2/000017", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000234, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_savanna/2/000017_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.03750782832503319, + "psnr": 15.657822608947754, + "lpips": 0.3379806578159332 + }, + "frames_100_200": { + "mse": 0.02688366360962391, + "psnr": 16.992361068725586, + "lpips": 0.4324433207511902 + }, + "frames_200_300": { + "mse": 0.005897705443203449, + "psnr": 22.438282012939453, + "lpips": 0.3725220561027527 + }, + "frames_300_400": { + "mse": 0.00526485126465559, + "psnr": 22.89144515991211, + "lpips": 0.3786814510822296 + }, + "frames_400_500": { + "mse": 0.006554132793098688, + "psnr": 22.12783432006836, + "lpips": 0.3606663942337036 + }, + "overall": { + "mse": 0.016421634703874588, + "psnr": 20.021549224853516, + "lpips": 0.3764587938785553 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000235_test__val_iceplains__0__000038/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000235_test__val_iceplains__0__000038/case.json new file mode 100644 index 0000000000000000000000000000000000000000..c36b5a859c03db214f5cb2cb9079dac609efec95 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000235_test__val_iceplains__0__000038/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "235", + "global_test_index": "235", + "dataset_index": "342", + "relative_path": "test/val_iceplains/0/000038.mp4", + "clip_id": "test/val_iceplains/0/000038", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000235, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.88533862889744, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000235_test__val_iceplains__0__000038/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000235_test__val_iceplains__0__000038/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..7705a1f5f1d77f4b587741223c440a22573d7e50 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000235_test__val_iceplains__0__000038/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1288155088f625994d64624427fa12cffea55ad0023600c5e7b154ab3a1edc22 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000235_test__val_iceplains__0__000038/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000235_test__val_iceplains__0__000038/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..f5dfad1bbf762ce5f10d9a7416e8ed9d8847b548 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000235_test__val_iceplains__0__000038/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "235", + "global_test_index": "235", + "dataset_index": "342", + "relative_path": "test/val_iceplains/0/000038.mp4", + "clip_id": "test/val_iceplains/0/000038", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000235, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/0/000038_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.010207943618297577, + "psnr": 20.69342803955078, + "lpips": 0.1524982899427414 + }, + "frames_100_200": { + "mse": 0.020455462858080864, + "psnr": 17.50543212890625, + "lpips": 0.21372665464878082 + }, + "frames_200_300": { + "mse": 0.011762366630136967, + "psnr": 19.433698654174805, + "lpips": 0.15064632892608643 + }, + "frames_300_400": { + "mse": 0.005864990875124931, + "psnr": 22.890235900878906, + "lpips": 0.15911762416362762 + }, + "frames_400_500": { + "mse": 0.005110207013785839, + "psnr": 23.287487030029297, + "lpips": 0.13776414096355438 + }, + "overall": { + "mse": 0.010680194944143295, + "psnr": 20.762056350708008, + "lpips": 0.16275063157081604 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000236_test__val_desert__3__000023/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000236_test__val_desert__3__000023/case.json new file mode 100644 index 0000000000000000000000000000000000000000..d0b0e960c51f76f496cf572b8f8298f8b2dcd546 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000236_test__val_desert__3__000023/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "236", + "global_test_index": "236", + "dataset_index": "60", + "relative_path": "test/val_desert/3/000023.mp4", + "clip_id": "test/val_desert/3/000023", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000236, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.98552276205737, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000236_test__val_desert__3__000023/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000236_test__val_desert__3__000023/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..1e8e3c93bce2901a33baffc4af658bbfb38ff334 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000236_test__val_desert__3__000023/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2011425fdbda8f72b2b59c19eabf0f6c61ca0b44ce7a169ead6e150808d5a7ef +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000236_test__val_desert__3__000023/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000236_test__val_desert__3__000023/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..7cab46e5994a2ffe77f92e2661bc96e68176347d --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000236_test__val_desert__3__000023/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "236", + "global_test_index": "236", + "dataset_index": "60", + "relative_path": "test/val_desert/3/000023.mp4", + "clip_id": "test/val_desert/3/000023", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000236, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_desert/3/000023_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0069823917001485825, + "psnr": 21.732860565185547, + "lpips": 0.2726597189903259 + }, + "frames_100_200": { + "mse": 0.010048793628811836, + "psnr": 20.067750930786133, + "lpips": 0.3943184018135071 + }, + "frames_200_300": { + "mse": 0.010908453725278378, + "psnr": 19.671178817749023, + "lpips": 0.39016208052635193 + }, + "frames_300_400": { + "mse": 0.009776645340025425, + "psnr": 20.213882446289062, + "lpips": 0.4045707583427429 + }, + "frames_400_500": { + "mse": 0.008896681480109692, + "psnr": 20.623655319213867, + "lpips": 0.39108750224113464 + }, + "overall": { + "mse": 0.009322592057287693, + "psnr": 20.46186637878418, + "lpips": 0.3705596923828125 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000237_test__val_savanna__1__000001/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000237_test__val_savanna__1__000001/case.json new file mode 100644 index 0000000000000000000000000000000000000000..3e6b363e07c5b61da439a4006e8ff43a4977ec4c --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000237_test__val_savanna__1__000001/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "237", + "global_test_index": "237", + "dataset_index": "131", + "relative_path": "test/val_savanna/1/000001.mp4", + "clip_id": "test/val_savanna/1/000001", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000237, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 223.6650989400223, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000237_test__val_savanna__1__000001/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000237_test__val_savanna__1__000001/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..f64664ad054902654816e6c9270be000788e18ec --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000237_test__val_savanna__1__000001/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d4cb0981a9dc5e59ce60dc0ac984a44fd8381748f8eeff5bac28dc356a86daee +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000237_test__val_savanna__1__000001/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000237_test__val_savanna__1__000001/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..accee019eebf7e46bb143697a632ca41c82767ac --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000237_test__val_savanna__1__000001/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "237", + "global_test_index": "237", + "dataset_index": "131", + "relative_path": "test/val_savanna/1/000001.mp4", + "clip_id": "test/val_savanna/1/000001", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000237, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_savanna/1/000001_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.004560120869427919, + "psnr": 23.51433563232422, + "lpips": 0.29827213287353516 + }, + "frames_100_200": { + "mse": 0.004695790354162455, + "psnr": 23.342748641967773, + "lpips": 0.3613608181476593 + }, + "frames_200_300": { + "mse": 0.005040014162659645, + "psnr": 23.039203643798828, + "lpips": 0.3627241551876068 + }, + "frames_300_400": { + "mse": 0.006568997632712126, + "psnr": 21.866357803344727, + "lpips": 0.3911687135696411 + }, + "frames_400_500": { + "mse": 0.005389518104493618, + "psnr": 22.74230194091797, + "lpips": 0.3583947718143463 + }, + "overall": { + "mse": 0.0052508884109556675, + "psnr": 22.90098762512207, + "lpips": 0.3543841242790222 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000238_test__val_iceplains__1__000047/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000238_test__val_iceplains__1__000047/case.json new file mode 100644 index 0000000000000000000000000000000000000000..8b84203de1057b48900648c6ccdd915f64f8416e --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000238_test__val_iceplains__1__000047/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "238", + "global_test_index": "238", + "dataset_index": "365", + "relative_path": "test/val_iceplains/1/000047.mp4", + "clip_id": "test/val_iceplains/1/000047", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000238, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.82757347996812, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000238_test__val_iceplains__1__000047/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000238_test__val_iceplains__1__000047/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..869c1201bd834804c6fa1f634271e87ec102086e --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000238_test__val_iceplains__1__000047/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:51e30119275ac5e7ee840c5b3962555b881dc50813885ab212db032e72971335 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000238_test__val_iceplains__1__000047/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000238_test__val_iceplains__1__000047/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..198df8470551a1b1e8d47ebdd99e2065188d6d4b --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000238_test__val_iceplains__1__000047/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "238", + "global_test_index": "238", + "dataset_index": "365", + "relative_path": "test/val_iceplains/1/000047.mp4", + "clip_id": "test/val_iceplains/1/000047", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000238, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/1/000047_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.005772189702838659, + "psnr": 23.116079330444336, + "lpips": 0.12548977136611938 + }, + "frames_100_200": { + "mse": 0.017451563850045204, + "psnr": 21.260841369628906, + "lpips": 0.19912715256214142 + }, + "frames_200_300": { + "mse": 0.04849817603826523, + "psnr": 14.583291053771973, + "lpips": 0.4188375174999237 + }, + "frames_300_400": { + "mse": 0.10068679600954056, + "psnr": 10.369680404663086, + "lpips": 0.5381397008895874 + }, + "frames_400_500": { + "mse": 0.1499849557876587, + "psnr": 8.356998443603516, + "lpips": 0.5838033556938171 + }, + "overall": { + "mse": 0.06447874009609222, + "psnr": 15.537378311157227, + "lpips": 0.3730795383453369 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000239_test__val_iceplains__2__000028/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000239_test__val_iceplains__2__000028/case.json new file mode 100644 index 0000000000000000000000000000000000000000..b57993b7e93f4e9332e1706de3b89c2cd2fdaf66 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000239_test__val_iceplains__2__000028/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "239", + "global_test_index": "239", + "dataset_index": "315", + "relative_path": "test/val_iceplains/2/000028.mp4", + "clip_id": "test/val_iceplains/2/000028", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000239, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.815907630953, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000239_test__val_iceplains__2__000028/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000239_test__val_iceplains__2__000028/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..e23b04c6da2c04e39ec8df07a0143527b4e1942b --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000239_test__val_iceplains__2__000028/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6e1e9ef07c239d76e15c56de4fafde217a87ae52fdefc48123e2afe8a64188cc +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000239_test__val_iceplains__2__000028/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000239_test__val_iceplains__2__000028/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..9f9ed51cd3ade55c01a99ff9c440fb80cf215696 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000239_test__val_iceplains__2__000028/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "239", + "global_test_index": "239", + "dataset_index": "315", + "relative_path": "test/val_iceplains/2/000028.mp4", + "clip_id": "test/val_iceplains/2/000028", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000239, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/2/000028_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.012226438149809837, + "psnr": 19.662353515625, + "lpips": 0.10657597333192825 + }, + "frames_100_200": { + "mse": 0.06449150294065475, + "psnr": 12.769797325134277, + "lpips": 0.354634553194046 + }, + "frames_200_300": { + "mse": 0.07251937687397003, + "psnr": 12.075216293334961, + "lpips": 0.34529945254325867 + }, + "frames_300_400": { + "mse": 0.09166176617145538, + "psnr": 10.518149375915527, + "lpips": 0.4712623655796051 + }, + "frames_400_500": { + "mse": 0.1057274341583252, + "psnr": 9.91873550415039, + "lpips": 0.49596598744392395 + }, + "overall": { + "mse": 0.06932530552148819, + "psnr": 12.988850593566895, + "lpips": 0.3547476530075073 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000240_test__val_savanna__1__000041/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000240_test__val_savanna__1__000041/case.json new file mode 100644 index 0000000000000000000000000000000000000000..827ab031f75608d12b1a3d056d4b2393d2124164 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000240_test__val_savanna__1__000041/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "240", + "global_test_index": "240", + "dataset_index": "220", + "relative_path": "test/val_savanna/1/000041.mp4", + "clip_id": "test/val_savanna/1/000041", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000240, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.92473953496665, + "peak_generation_vram_bytes": 107565475840, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000240_test__val_savanna__1__000041/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000240_test__val_savanna__1__000041/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..ed5fd29c1db2ca1741b17535628f2b32e8786209 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000240_test__val_savanna__1__000041/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f1702926b7017946bb919b121220f5024e43f0159bf912de3deec3cab5c77ddc +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000240_test__val_savanna__1__000041/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000240_test__val_savanna__1__000041/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..d6e8124ac6bb88fb02edaaacd2a8fbb52da11125 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000240_test__val_savanna__1__000041/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "240", + "global_test_index": "240", + "dataset_index": "220", + "relative_path": "test/val_savanna/1/000041.mp4", + "clip_id": "test/val_savanna/1/000041", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000240, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_savanna/1/000041_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.012239744886755943, + "psnr": 19.492149353027344, + "lpips": 0.16505493223667145 + }, + "frames_100_200": { + "mse": 0.016145870089530945, + "psnr": 18.320619583129883, + "lpips": 0.20324811339378357 + }, + "frames_200_300": { + "mse": 0.01854459010064602, + "psnr": 17.667085647583008, + "lpips": 0.3378596603870392 + }, + "frames_300_400": { + "mse": 0.01204095408320427, + "psnr": 19.318601608276367, + "lpips": 0.35395461320877075 + }, + "frames_400_500": { + "mse": 0.013324621133506298, + "psnr": 19.22606658935547, + "lpips": 0.358144074678421 + }, + "overall": { + "mse": 0.01445915549993515, + "psnr": 18.804906845092773, + "lpips": 0.28365227580070496 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000241_test__val_desert__1__000003/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000241_test__val_desert__1__000003/case.json new file mode 100644 index 0000000000000000000000000000000000000000..93f5a99a1e99214cc8676e53b034ec3ec5988d36 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000241_test__val_desert__1__000003/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "241", + "global_test_index": "241", + "dataset_index": "8", + "relative_path": "test/val_desert/1/000003.mp4", + "clip_id": "test/val_desert/1/000003", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000241, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.77035407396033, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000241_test__val_desert__1__000003/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000241_test__val_desert__1__000003/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..d5f77219513daf3f40b2317611fc1013a5c232a5 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000241_test__val_desert__1__000003/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:586c54c95edfd9ea2cc1e53f06b9c063f1972ac1b6f8312234f688301993686c +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000241_test__val_desert__1__000003/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000241_test__val_desert__1__000003/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..b0abb7047b0e9d5a0499b02e9f97de4c9d98ea2f --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000241_test__val_desert__1__000003/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "241", + "global_test_index": "241", + "dataset_index": "8", + "relative_path": "test/val_desert/1/000003.mp4", + "clip_id": "test/val_desert/1/000003", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000241, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_desert/1/000003_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.001165994442999363, + "psnr": 30.467090606689453, + "lpips": 0.014940854161977768 + }, + "frames_100_200": { + "mse": 0.0008118937839753926, + "psnr": 30.927871704101562, + "lpips": 0.011644316837191582 + }, + "frames_200_300": { + "mse": 0.001190223847515881, + "psnr": 30.507368087768555, + "lpips": 0.014914054423570633 + }, + "frames_300_400": { + "mse": 0.0007692479412071407, + "psnr": 31.149839401245117, + "lpips": 0.010760212317109108 + }, + "frames_400_500": { + "mse": 0.0008386506233364344, + "psnr": 30.813493728637695, + "lpips": 0.013595359399914742 + }, + "overall": { + "mse": 0.0009552020928822458, + "psnr": 30.773128509521484, + "lpips": 0.013170959427952766 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000242_test__val_iceplains__0__000006/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000242_test__val_iceplains__0__000006/case.json new file mode 100644 index 0000000000000000000000000000000000000000..534e59593e1f76d075d6ffd5cda8fbc659ab2f19 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000242_test__val_iceplains__0__000006/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "242", + "global_test_index": "242", + "dataset_index": "261", + "relative_path": "test/val_iceplains/0/000006.mp4", + "clip_id": "test/val_iceplains/0/000006", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000242, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.50504507601727, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000242_test__val_iceplains__0__000006/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000242_test__val_iceplains__0__000006/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..624bd0a676601e82a9332eb5c725c3e5e9ecc3da --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000242_test__val_iceplains__0__000006/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a89f09a6d6748878b091bc47c17b927adb61c61d56446a0411f921ac5977f85d +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000242_test__val_iceplains__0__000006/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000242_test__val_iceplains__0__000006/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..65cfac7c3974c7c3fa301186c7f90cf75e0ec2fd --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000242_test__val_iceplains__0__000006/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "242", + "global_test_index": "242", + "dataset_index": "261", + "relative_path": "test/val_iceplains/0/000006.mp4", + "clip_id": "test/val_iceplains/0/000006", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000242, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/0/000006_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.008432742208242416, + "psnr": 21.880844116210938, + "lpips": 0.21070365607738495 + }, + "frames_100_200": { + "mse": 0.1353977620601654, + "psnr": 11.908424377441406, + "lpips": 0.4426135718822479 + }, + "frames_200_300": { + "mse": 0.1122446060180664, + "psnr": 9.633312225341797, + "lpips": 0.5291843414306641 + }, + "frames_300_400": { + "mse": 0.1390187293291092, + "psnr": 8.797527313232422, + "lpips": 0.5340901017189026 + }, + "frames_400_500": { + "mse": 0.17507387697696686, + "psnr": 9.83187484741211, + "lpips": 0.4902043044567108 + }, + "overall": { + "mse": 0.1140335351228714, + "psnr": 12.41039752960205, + "lpips": 0.4413592219352722 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000243_test__val_desert__1__000007/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000243_test__val_desert__1__000007/case.json new file mode 100644 index 0000000000000000000000000000000000000000..725927f437f1cb7d1695e2cb951fe5f689fbd072 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000243_test__val_desert__1__000007/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "243", + "global_test_index": "243", + "dataset_index": "17", + "relative_path": "test/val_desert/1/000007.mp4", + "clip_id": "test/val_desert/1/000007", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000243, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.41893462103326, + "peak_generation_vram_bytes": 107565475840, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000243_test__val_desert__1__000007/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000243_test__val_desert__1__000007/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..813a28f4e65d7daf74aafc2b6ca71b5a6498af9e --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000243_test__val_desert__1__000007/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7087cfadf88f80a9354cb505a363975e710f412ea4b9614d129fe2a5c0e114f4 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000243_test__val_desert__1__000007/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000243_test__val_desert__1__000007/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..6f4e9c8ffc30f9064705330012b319a0d6ae6b02 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000243_test__val_desert__1__000007/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "243", + "global_test_index": "243", + "dataset_index": "17", + "relative_path": "test/val_desert/1/000007.mp4", + "clip_id": "test/val_desert/1/000007", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000243, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_desert/1/000007_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.004771084059029818, + "psnr": 25.366788864135742, + "lpips": 0.05312604084610939 + }, + "frames_100_200": { + "mse": 0.012178000062704086, + "psnr": 22.070005416870117, + "lpips": 0.08429396897554398 + }, + "frames_200_300": { + "mse": 0.012659284286201, + "psnr": 20.87729263305664, + "lpips": 0.09162354469299316 + }, + "frames_300_400": { + "mse": 0.007961191236972809, + "psnr": 23.73114013671875, + "lpips": 0.1059633269906044 + }, + "frames_400_500": { + "mse": 0.00799665879458189, + "psnr": 23.50628662109375, + "lpips": 0.07592077553272247 + }, + "overall": { + "mse": 0.009113242849707603, + "psnr": 23.11030387878418, + "lpips": 0.0821855291724205 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000244_test__val_plains__1__000043/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000244_test__val_plains__1__000043/case.json new file mode 100644 index 0000000000000000000000000000000000000000..ae5f5fe4d5d43d936ba40ef480c56ccc91e15bb9 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000244_test__val_plains__1__000043/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "244", + "global_test_index": "244", + "dataset_index": "462", + "relative_path": "test/val_plains/1/000043.mp4", + "clip_id": "test/val_plains/1/000043", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000244, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 223.7572308209492, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000244_test__val_plains__1__000043/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000244_test__val_plains__1__000043/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..50dc6b5da9a6576575f8292e62da2ef8d7af9d7c --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000244_test__val_plains__1__000043/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d9a857dfe0e356453835dd069d42fa786bbbec916c3e37c625c59a6b94458d5b +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000244_test__val_plains__1__000043/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000244_test__val_plains__1__000043/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..707c4f9fd2a96d0c46a6a9fb5665f034194125e3 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000244_test__val_plains__1__000043/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "244", + "global_test_index": "244", + "dataset_index": "462", + "relative_path": "test/val_plains/1/000043.mp4", + "clip_id": "test/val_plains/1/000043", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000244, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_plains/1/000043_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.004509656224399805, + "psnr": 23.8902587890625, + "lpips": 0.14481540024280548 + }, + "frames_100_200": { + "mse": 0.007482993416488171, + "psnr": 21.952322006225586, + "lpips": 0.20787911117076874 + }, + "frames_200_300": { + "mse": 0.005241048522293568, + "psnr": 23.091079711914062, + "lpips": 0.08721651136875153 + }, + "frames_300_400": { + "mse": 0.0030697211623191833, + "psnr": 25.543519973754883, + "lpips": 0.07090502977371216 + }, + "frames_400_500": { + "mse": 0.0009680018411017954, + "psnr": 30.23395538330078, + "lpips": 0.018152626231312752 + }, + "overall": { + "mse": 0.004254283849149942, + "psnr": 24.94222640991211, + "lpips": 0.10579372197389603 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000245_test__val_plains__3__000038/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000245_test__val_plains__3__000038/case.json new file mode 100644 index 0000000000000000000000000000000000000000..9812992b8d1b210661986ac9084a66639738075f --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000245_test__val_plains__3__000038/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "245", + "global_test_index": "245", + "dataset_index": "454", + "relative_path": "test/val_plains/3/000038.mp4", + "clip_id": "test/val_plains/3/000038", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000245, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.1556386379525, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000245_test__val_plains__3__000038/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000245_test__val_plains__3__000038/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..1f0733498bb6392391fb9205392cd44abef44eeb --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000245_test__val_plains__3__000038/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0a162e2a4af5a14e13801a8374f71a56faf1979092e20c51e91977e17682dbfc +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000245_test__val_plains__3__000038/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000245_test__val_plains__3__000038/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..519667dfb367ed40dd7f7fae19177d41125d3f0a --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000245_test__val_plains__3__000038/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "245", + "global_test_index": "245", + "dataset_index": "454", + "relative_path": "test/val_plains/3/000038.mp4", + "clip_id": "test/val_plains/3/000038", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000245, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_plains/3/000038_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.004936047364026308, + "psnr": 23.150278091430664, + "lpips": 0.3025384545326233 + }, + "frames_100_200": { + "mse": 0.0061860536225140095, + "psnr": 22.198932647705078, + "lpips": 0.39797964692115784 + }, + "frames_200_300": { + "mse": 0.00699949124827981, + "psnr": 21.625791549682617, + "lpips": 0.4388505518436432 + }, + "frames_300_400": { + "mse": 0.009992673061788082, + "psnr": 20.67193603515625, + "lpips": 0.5019224286079407 + }, + "frames_400_500": { + "mse": 0.006608156021684408, + "psnr": 21.877195358276367, + "lpips": 0.4316340684890747 + }, + "overall": { + "mse": 0.006944484077394009, + "psnr": 21.904827117919922, + "lpips": 0.41458502411842346 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000246_test__val_iceplains__1__000041/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000246_test__val_iceplains__1__000041/case.json new file mode 100644 index 0000000000000000000000000000000000000000..6652e1b9ff4e8df952ab4ccc796746d6605acbfe --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000246_test__val_iceplains__1__000041/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "246", + "global_test_index": "246", + "dataset_index": "351", + "relative_path": "test/val_iceplains/1/000041.mp4", + "clip_id": "test/val_iceplains/1/000041", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000246, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.9901513151126, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000246_test__val_iceplains__1__000041/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000246_test__val_iceplains__1__000041/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..c2d0d3156d369ab827dde2a5a9cd719ea85a3913 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000246_test__val_iceplains__1__000041/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0404589d93bb7da70e8eecfb3d0f76946afd2f8e4b87e0470337fa15412733a1 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000246_test__val_iceplains__1__000041/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000246_test__val_iceplains__1__000041/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..fd3f9ccce1cf3889a04093bc5fb6a283bdce2739 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000246_test__val_iceplains__1__000041/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "246", + "global_test_index": "246", + "dataset_index": "351", + "relative_path": "test/val_iceplains/1/000041.mp4", + "clip_id": "test/val_iceplains/1/000041", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000246, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/1/000041_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0007677110843360424, + "psnr": 31.163257598876953, + "lpips": 0.010581429116427898 + }, + "frames_100_200": { + "mse": 0.0013028017710894346, + "psnr": 30.382606506347656, + "lpips": 0.013181978836655617 + }, + "frames_200_300": { + "mse": 0.0008317634928971529, + "psnr": 30.838623046875, + "lpips": 0.012192890979349613 + }, + "frames_300_400": { + "mse": 0.01659473590552807, + "psnr": 28.817371368408203, + "lpips": 0.05279314145445824 + }, + "frames_400_500": { + "mse": 0.043333958834409714, + "psnr": 13.991600036621094, + "lpips": 0.0850866287946701 + }, + "overall": { + "mse": 0.01256619393825531, + "psnr": 27.0386905670166, + "lpips": 0.03476721793413162 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000247_test__val_plains__3__000005/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000247_test__val_plains__3__000005/case.json new file mode 100644 index 0000000000000000000000000000000000000000..829c5f806725bd850e4ae2aa4a164c75413e40d7 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000247_test__val_plains__3__000005/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "247", + "global_test_index": "247", + "dataset_index": "383", + "relative_path": "test/val_plains/3/000005.mp4", + "clip_id": "test/val_plains/3/000005", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000247, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.27994706796017, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000247_test__val_plains__3__000005/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000247_test__val_plains__3__000005/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..850137dd18a15226222d55cc43c4c5be4c8c7f28 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000247_test__val_plains__3__000005/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0c3b902abb55ec63addf81be07549d1bad0f0f99251b30ee5581f1b10df7f818 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000247_test__val_plains__3__000005/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000247_test__val_plains__3__000005/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..362e985095bf81b69426be9d8bf782193e265793 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000247_test__val_plains__3__000005/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "247", + "global_test_index": "247", + "dataset_index": "383", + "relative_path": "test/val_plains/3/000005.mp4", + "clip_id": "test/val_plains/3/000005", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000247, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_plains/3/000005_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0007955450564622879, + "psnr": 31.018413543701172, + "lpips": 0.010833727195858955 + }, + "frames_100_200": { + "mse": 0.001049949787557125, + "psnr": 29.95941162109375, + "lpips": 0.016556911170482635 + }, + "frames_200_300": { + "mse": 0.0010739094577729702, + "psnr": 29.728586196899414, + "lpips": 0.021251235157251358 + }, + "frames_300_400": { + "mse": 0.009070217609405518, + "psnr": 25.544511795043945, + "lpips": 0.06964361667633057 + }, + "frames_400_500": { + "mse": 0.005313382018357515, + "psnr": 27.010292053222656, + "lpips": 0.04762210324406624 + }, + "overall": { + "mse": 0.003460601205006242, + "psnr": 28.652244567871094, + "lpips": 0.03318151831626892 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000248_test__val_desert__2__000007/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000248_test__val_desert__2__000007/case.json new file mode 100644 index 0000000000000000000000000000000000000000..ff17060512e9f2f804e7e8ab162294c41a4850b8 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000248_test__val_desert__2__000007/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "248", + "global_test_index": "248", + "dataset_index": "18", + "relative_path": "test/val_desert/2/000007.mp4", + "clip_id": "test/val_desert/2/000007", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000248, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.72851635003462, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000248_test__val_desert__2__000007/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000248_test__val_desert__2__000007/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..5ae2f3b8acc68a0ac7c289594b63300ea4acc965 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000248_test__val_desert__2__000007/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:054e6bb726c3f46c490918194869ee6491b7783a4b8fe8af20bb6229a428af0d +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000248_test__val_desert__2__000007/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000248_test__val_desert__2__000007/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..2e4405911e8268576fd946509937073d40b31611 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000248_test__val_desert__2__000007/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "248", + "global_test_index": "248", + "dataset_index": "18", + "relative_path": "test/val_desert/2/000007.mp4", + "clip_id": "test/val_desert/2/000007", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000248, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_desert/2/000007_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0029501088429242373, + "psnr": 25.7303524017334, + "lpips": 0.05728140473365784 + }, + "frames_100_200": { + "mse": 0.007003093138337135, + "psnr": 21.808446884155273, + "lpips": 0.2160571813583374 + }, + "frames_200_300": { + "mse": 0.010995878838002682, + "psnr": 19.792409896850586, + "lpips": 0.3601968288421631 + }, + "frames_300_400": { + "mse": 0.014262200333178043, + "psnr": 18.776058197021484, + "lpips": 0.43202513456344604 + }, + "frames_400_500": { + "mse": 0.015904322266578674, + "psnr": 18.275684356689453, + "lpips": 0.43720999360084534 + }, + "overall": { + "mse": 0.01022312045097351, + "psnr": 20.876590728759766, + "lpips": 0.3005540668964386 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000249_test__val_plains__1__000040/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000249_test__val_plains__1__000040/case.json new file mode 100644 index 0000000000000000000000000000000000000000..797e1c7fc1acbd61d5e0a9e56a19998ca1b20624 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000249_test__val_plains__1__000040/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "249", + "global_test_index": "249", + "dataset_index": "457", + "relative_path": "test/val_plains/1/000040.mp4", + "clip_id": "test/val_plains/1/000040", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000249, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.66553614498116, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000249_test__val_plains__1__000040/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000249_test__val_plains__1__000040/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..5ff594ad38e73cbcf6d40727f1473457f9496038 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000249_test__val_plains__1__000040/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:493bedf9632845136672983d7e14bcdf0a38ae8ddbee8de3e0b4a356844111ac +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000249_test__val_plains__1__000040/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000249_test__val_plains__1__000040/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..1aa4139989bbc1f5db7f8c07e61a6bc7d73b273f --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000249_test__val_plains__1__000040/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "249", + "global_test_index": "249", + "dataset_index": "457", + "relative_path": "test/val_plains/1/000040.mp4", + "clip_id": "test/val_plains/1/000040", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000249, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_plains/1/000040_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.012119513005018234, + "psnr": 20.159761428833008, + "lpips": 0.12787838280200958 + }, + "frames_100_200": { + "mse": 0.014870178885757923, + "psnr": 18.715919494628906, + "lpips": 0.2689007818698883 + }, + "frames_200_300": { + "mse": 0.022097570821642876, + "psnr": 18.18804168701172, + "lpips": 0.28538864850997925 + }, + "frames_300_400": { + "mse": 0.014262448996305466, + "psnr": 22.6163272857666, + "lpips": 0.13256292045116425 + }, + "frames_400_500": { + "mse": 0.03858881816267967, + "psnr": 14.398415565490723, + "lpips": 0.48623260855674744 + }, + "overall": { + "mse": 0.02038770727813244, + "psnr": 18.815692901611328, + "lpips": 0.2601926922798157 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000250_test__val_iceplains__3__000033/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000250_test__val_iceplains__3__000033/case.json new file mode 100644 index 0000000000000000000000000000000000000000..552cc8fb4bc911b5344be92546554524ebec33fc --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000250_test__val_iceplains__3__000033/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "250", + "global_test_index": "250", + "dataset_index": "326", + "relative_path": "test/val_iceplains/3/000033.mp4", + "clip_id": "test/val_iceplains/3/000033", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000250, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.43860249302816, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000250_test__val_iceplains__3__000033/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000250_test__val_iceplains__3__000033/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..8301e8e63f8d4845db37836315af005aac2a6101 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000250_test__val_iceplains__3__000033/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:88192125bbca674d6b64413bb0f021f6aa1512ac33069c22f19b5b2751129737 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000250_test__val_iceplains__3__000033/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000250_test__val_iceplains__3__000033/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..95c6506724f061eaaf23998166cad585264e3d12 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000250_test__val_iceplains__3__000033/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "250", + "global_test_index": "250", + "dataset_index": "326", + "relative_path": "test/val_iceplains/3/000033.mp4", + "clip_id": "test/val_iceplains/3/000033", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000250, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/3/000033_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0009517200523987412, + "psnr": 30.578601837158203, + "lpips": 0.012394152581691742 + }, + "frames_100_200": { + "mse": 0.0008872960461303592, + "psnr": 30.951791763305664, + "lpips": 0.010243546217679977 + }, + "frames_200_300": { + "mse": 0.00359646906144917, + "psnr": 26.61761474609375, + "lpips": 0.023892132565379143 + }, + "frames_300_400": { + "mse": 0.011360096745193005, + "psnr": 20.660558700561523, + "lpips": 0.09963025897741318 + }, + "frames_400_500": { + "mse": 0.0025073217693716288, + "psnr": 27.733068466186523, + "lpips": 0.022735944017767906 + }, + "overall": { + "mse": 0.003860580502077937, + "psnr": 27.308326721191406, + "lpips": 0.03377920761704445 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000251_test__val_iceplains__1__000004/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000251_test__val_iceplains__1__000004/case.json new file mode 100644 index 0000000000000000000000000000000000000000..0e9c9cb2c3e260b3709dd5daa0cdacf0ef6154b8 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000251_test__val_iceplains__1__000004/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "251", + "global_test_index": "251", + "dataset_index": "254", + "relative_path": "test/val_iceplains/1/000004.mp4", + "clip_id": "test/val_iceplains/1/000004", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000251, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.25536180997733, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000251_test__val_iceplains__1__000004/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000251_test__val_iceplains__1__000004/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..b8e309cb45038f0a9402caf786ff562d9729eb5d --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000251_test__val_iceplains__1__000004/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:84fac874ad3d2bc0c897f949fe8a5595ed5b5a527b93a0a855f494d829b04606 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000251_test__val_iceplains__1__000004/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000251_test__val_iceplains__1__000004/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..5a21cb767faab51b2f082cb8b601886cc3716a6c --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000251_test__val_iceplains__1__000004/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "251", + "global_test_index": "251", + "dataset_index": "254", + "relative_path": "test/val_iceplains/1/000004.mp4", + "clip_id": "test/val_iceplains/1/000004", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000251, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/1/000004_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.010149712674319744, + "psnr": 21.22600555419922, + "lpips": 0.15166959166526794 + }, + "frames_100_200": { + "mse": 0.02500809170305729, + "psnr": 16.797975540161133, + "lpips": 0.2006334811449051 + }, + "frames_200_300": { + "mse": 0.021904869005084038, + "psnr": 17.20272445678711, + "lpips": 0.12727466225624084 + }, + "frames_300_400": { + "mse": 0.05604596063494682, + "psnr": 12.7514066696167, + "lpips": 0.2786364257335663 + }, + "frames_400_500": { + "mse": 0.10084564983844757, + "psnr": 10.578470230102539, + "lpips": 0.4557775855064392 + }, + "overall": { + "mse": 0.04279085993766785, + "psnr": 15.71131706237793, + "lpips": 0.24279837310314178 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000252_test__val_plains__1__000002/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000252_test__val_plains__1__000002/case.json new file mode 100644 index 0000000000000000000000000000000000000000..4033e8693abc77e885b7ee1cdbcbc8a6ae3b7a3e --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000252_test__val_plains__1__000002/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "252", + "global_test_index": "252", + "dataset_index": "375", + "relative_path": "test/val_plains/1/000002.mp4", + "clip_id": "test/val_plains/1/000002", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000252, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 223.94275476201437, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000252_test__val_plains__1__000002/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000252_test__val_plains__1__000002/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..2f9f3f9138f9abfff68023fc646ca182701209e4 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000252_test__val_plains__1__000002/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:298df44b62724564b3ac823829b9e41e9a57cdfd8e73216178ee950e88598145 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000252_test__val_plains__1__000002/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000252_test__val_plains__1__000002/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..d19388c52fc1205a7512d68770b0981078f8f4de --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000252_test__val_plains__1__000002/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "252", + "global_test_index": "252", + "dataset_index": "375", + "relative_path": "test/val_plains/1/000002.mp4", + "clip_id": "test/val_plains/1/000002", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000252, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_plains/1/000002_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0034824362955987453, + "psnr": 25.169761657714844, + "lpips": 0.0627240389585495 + }, + "frames_100_200": { + "mse": 0.001575447036884725, + "psnr": 28.56790542602539, + "lpips": 0.022849077358841896 + }, + "frames_200_300": { + "mse": 0.0027098029386252165, + "psnr": 26.990633010864258, + "lpips": 0.027103366330266 + }, + "frames_300_400": { + "mse": 0.002372290473431349, + "psnr": 27.337486267089844, + "lpips": 0.029128627851605415 + }, + "frames_400_500": { + "mse": 0.002033020369708538, + "psnr": 27.71883773803711, + "lpips": 0.029189612716436386 + }, + "overall": { + "mse": 0.0024345992133021355, + "psnr": 27.15692138671875, + "lpips": 0.03419894352555275 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000253_test__val_savanna__0__000005/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000253_test__val_savanna__0__000005/case.json new file mode 100644 index 0000000000000000000000000000000000000000..d316780d6138278759bf59f43f7323311c0d165d --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000253_test__val_savanna__0__000005/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "253", + "global_test_index": "253", + "dataset_index": "141", + "relative_path": "test/val_savanna/0/000005.mp4", + "clip_id": "test/val_savanna/0/000005", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000253, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.0325505399378, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000253_test__val_savanna__0__000005/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000253_test__val_savanna__0__000005/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..d6941a6241c694b6010463ec43bb6c3dcfb6dfea --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000253_test__val_savanna__0__000005/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:532acbc7275c9939fad2aa1cb7eca544a63e644c736f811d004bd8c338d0a554 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000253_test__val_savanna__0__000005/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000253_test__val_savanna__0__000005/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..a9238adc2fdbcbb0bf3c2a6ca3b615d5b12f7496 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000253_test__val_savanna__0__000005/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "253", + "global_test_index": "253", + "dataset_index": "141", + "relative_path": "test/val_savanna/0/000005.mp4", + "clip_id": "test/val_savanna/0/000005", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000253, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_savanna/0/000005_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.009385495446622372, + "psnr": 20.704313278198242, + "lpips": 0.07857644557952881 + }, + "frames_100_200": { + "mse": 0.01552948821336031, + "psnr": 18.772998809814453, + "lpips": 0.14716878533363342 + }, + "frames_200_300": { + "mse": 0.011887295171618462, + "psnr": 20.964069366455078, + "lpips": 0.12267794460058212 + }, + "frames_300_400": { + "mse": 0.02854195050895214, + "psnr": 17.39002799987793, + "lpips": 0.24239790439605713 + }, + "frames_400_500": { + "mse": 0.0326286256313324, + "psnr": 15.500401496887207, + "lpips": 0.26917195320129395 + }, + "overall": { + "mse": 0.019594570621848106, + "psnr": 18.66636085510254, + "lpips": 0.17199859023094177 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000254_test__val_iceplains__2__000024/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000254_test__val_iceplains__2__000024/case.json new file mode 100644 index 0000000000000000000000000000000000000000..46431a5d236af3ad424a497b348bb264d81c3795 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000254_test__val_iceplains__2__000024/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "254", + "global_test_index": "254", + "dataset_index": "306", + "relative_path": "test/val_iceplains/2/000024.mp4", + "clip_id": "test/val_iceplains/2/000024", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000254, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.35526933695655, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000254_test__val_iceplains__2__000024/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000254_test__val_iceplains__2__000024/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..18878e86e4db9f1b7b6973df211bf6e3acb66d58 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000254_test__val_iceplains__2__000024/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:eab4983f2c9596484614f6482ee53421f7e09ee660632abd8fc9a88230cbb78b +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000254_test__val_iceplains__2__000024/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000254_test__val_iceplains__2__000024/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..61a03e44474ced04490d01d063a0d8124d7a8792 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000254_test__val_iceplains__2__000024/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "254", + "global_test_index": "254", + "dataset_index": "306", + "relative_path": "test/val_iceplains/2/000024.mp4", + "clip_id": "test/val_iceplains/2/000024", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000254, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/2/000024_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.013654878363013268, + "psnr": 19.019296646118164, + "lpips": 0.14473174512386322 + }, + "frames_100_200": { + "mse": 0.0232087429612875, + "psnr": 17.015901565551758, + "lpips": 0.16372103989124298 + }, + "frames_200_300": { + "mse": 0.009077183902263641, + "psnr": 21.224111557006836, + "lpips": 0.06376715004444122 + }, + "frames_300_400": { + "mse": 0.011818391270935535, + "psnr": 20.001768112182617, + "lpips": 0.1002173200249672 + }, + "frames_400_500": { + "mse": 0.007881238125264645, + "psnr": 22.7407169342041, + "lpips": 0.06114912033081055 + }, + "overall": { + "mse": 0.013128085993230343, + "psnr": 20.00035858154297, + "lpips": 0.10671728104352951 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000255_test__val_desert__2__000003/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000255_test__val_desert__2__000003/case.json new file mode 100644 index 0000000000000000000000000000000000000000..d1651ab510e0f1ec937b797f3b2043d33fe0223e --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000255_test__val_desert__2__000003/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "255", + "global_test_index": "255", + "dataset_index": "10", + "relative_path": "test/val_desert/2/000003.mp4", + "clip_id": "test/val_desert/2/000003", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000255, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 223.61922039894853, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000255_test__val_desert__2__000003/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000255_test__val_desert__2__000003/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..402a1eea6ecc843b66d7a5c5566ff4ea5be03ab9 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000255_test__val_desert__2__000003/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f3d52c2da244172cb3e59539a8c05f2219060c24201c8b6d61b5489fd9d0d160 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000255_test__val_desert__2__000003/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000255_test__val_desert__2__000003/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..c1c709b7ba88307af140c1162a8c67ec4f54cf1d --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000255_test__val_desert__2__000003/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "255", + "global_test_index": "255", + "dataset_index": "10", + "relative_path": "test/val_desert/2/000003.mp4", + "clip_id": "test/val_desert/2/000003", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000255, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_desert/2/000003_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.0008472291519865394, + "psnr": 30.728736877441406, + "lpips": 0.010708298534154892 + }, + "frames_100_200": { + "mse": 0.001100523630157113, + "psnr": 30.55615234375, + "lpips": 0.014124656096100807 + }, + "frames_200_300": { + "mse": 0.0010903336806222796, + "psnr": 30.54913330078125, + "lpips": 0.014424030669033527 + }, + "frames_300_400": { + "mse": 0.0008891052566468716, + "psnr": 30.659494400024414, + "lpips": 0.014485605992376804 + }, + "frames_400_500": { + "mse": 0.0008983163279481232, + "psnr": 30.560649871826172, + "lpips": 0.01341088768094778 + }, + "overall": { + "mse": 0.000965101586189121, + "psnr": 30.610836029052734, + "lpips": 0.013430697843432426 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000256_test__val_savanna__0__000020/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000256_test__val_savanna__0__000020/case.json new file mode 100644 index 0000000000000000000000000000000000000000..9cd481db8101a91b95ecbf87b70a274b2f843fbd --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000256_test__val_savanna__0__000020/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "256", + "global_test_index": "256", + "dataset_index": "173", + "relative_path": "test/val_savanna/0/000020.mp4", + "clip_id": "test/val_savanna/0/000020", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000256, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.3227255269885, + "peak_generation_vram_bytes": 107565475840, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000256_test__val_savanna__0__000020/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000256_test__val_savanna__0__000020/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..bf4337e66c1f39d06e750f17c17ec133264f9067 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000256_test__val_savanna__0__000020/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:754f99a58d9b2eb015ca8c0167c93b1d766628994b165edd86b7603a6f446019 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000256_test__val_savanna__0__000020/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000256_test__val_savanna__0__000020/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..9ec8c8a6dd253b6da618aa5b52de9ddfed692a74 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000256_test__val_savanna__0__000020/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "256", + "global_test_index": "256", + "dataset_index": "173", + "relative_path": "test/val_savanna/0/000020.mp4", + "clip_id": "test/val_savanna/0/000020", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000256, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_savanna/0/000020_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.025751128792762756, + "psnr": 16.340229034423828, + "lpips": 0.2349228411912918 + }, + "frames_100_200": { + "mse": 0.04496900737285614, + "psnr": 15.921581268310547, + "lpips": 0.20947428047657013 + }, + "frames_200_300": { + "mse": 0.016573606058955193, + "psnr": 19.77880096435547, + "lpips": 0.11371506750583649 + }, + "frames_300_400": { + "mse": 0.011559528298676014, + "psnr": 20.497661590576172, + "lpips": 0.07000237703323364 + }, + "frames_400_500": { + "mse": 0.0408894345164299, + "psnr": 15.573758125305176, + "lpips": 0.2961890399456024 + }, + "overall": { + "mse": 0.027948541566729546, + "psnr": 17.622406005859375, + "lpips": 0.1848607212305069 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000257_test__val_iceplains__1__000032/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000257_test__val_iceplains__1__000032/case.json new file mode 100644 index 0000000000000000000000000000000000000000..9efca8a8ec7e2e18e25928ff3113c24a0be71cac --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000257_test__val_iceplains__1__000032/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "257", + "global_test_index": "257", + "dataset_index": "320", + "relative_path": "test/val_iceplains/1/000032.mp4", + "clip_id": "test/val_iceplains/1/000032", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000257, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 225.6394427740015, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000257_test__val_iceplains__1__000032/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000257_test__val_iceplains__1__000032/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..b539cc84dc913b7aed32fcaa40fa4663140df4fe --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000257_test__val_iceplains__1__000032/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b5ea2b8c8b15daaaaf5142d6597c92e60b9313d018225837f663ba77094a7395 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000257_test__val_iceplains__1__000032/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000257_test__val_iceplains__1__000032/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..0ec6416290a13e87eea190cfec2d99edc4b3d8e9 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000257_test__val_iceplains__1__000032/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "257", + "global_test_index": "257", + "dataset_index": "320", + "relative_path": "test/val_iceplains/1/000032.mp4", + "clip_id": "test/val_iceplains/1/000032", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "sample_seed": 43000257, + "dataset": "minecraft", + "frames": 500, + "native_resolution": [ + 352, + 640 + ], + "metric_resolution": [ + 360, + 640 + ], + "resize": "bicubic align_corners=False antialias=True", + "reference_source": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features/test/val_iceplains/1/000032_vae_feature.npy", + "vae_checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/worldmem/vae_only.ckpt", + "torch": "2.8.0+cu128", + "torchmetrics": "0.11.4", + "lpips": "alex; normalize=True; mean of individual frame values", + "psnr_epsilon": 1.1920928955078125e-07, + "fid": "OpenCV decoding of libx264 quality8 yuv420p uint8 videos", + "video": { + "codec": "libx264", + "quality": 8, + "pixelformat": "yuv420p", + "macro_block_size": 1, + "fps": 15 + }, + "ranges": { + "frames_000_100": { + "mse": 0.001624181168153882, + "psnr": 30.041112899780273, + "lpips": 0.015480279922485352 + }, + "frames_100_200": { + "mse": 0.0014655754202976823, + "psnr": 30.418298721313477, + "lpips": 0.014119994826614857 + }, + "frames_200_300": { + "mse": 0.0008004309493117034, + "psnr": 31.010671615600586, + "lpips": 0.010761750862002373 + }, + "frames_300_400": { + "mse": 0.0015821738634258509, + "psnr": 30.364458084106445, + "lpips": 0.015000165440142155 + }, + "frames_400_500": { + "mse": 0.0009388901526108384, + "psnr": 30.380268096923828, + "lpips": 0.016799695789813995 + }, + "overall": { + "mse": 0.0012822503922507167, + "psnr": 30.442962646484375, + "lpips": 0.014432378113269806 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000258_test__val_desert__1__000044/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000258_test__val_desert__1__000044/case.json new file mode 100644 index 0000000000000000000000000000000000000000..d3726e1d55332dd867359ff55af324e695f9d53b --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed43_job85134/cases/g000258_test__val_desert__1__000044/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "258", + "global_test_index": "258", + "dataset_index": "115", + "relative_path": "test/val_desert/1/000044.mp4", + "clip_id": "test/val_desert/1/000044", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_valid": "1100", + "context_frames": "600", + "target_frames": "500", + "frame_skip": "1", + "raw_frame_count": "1501", + "action_count": "1501", + "pose_count": "1501", + "latent_count": "1501" + }, + "method": "decmem", + "implementation": "official public Wan2.1-T2V-1.3B reference", + "checkpoint": "/proj/cvl/users/x_fahkh2/worldmem-baseline-evals/checkpoints/decmem/decmem.pt", + "weight_source": "released generator state; no EMA substitution", + "parameter_counts": { + "generator_total": 1197639296, + "generator_trainable": 0, + "action_encoder": 1600, + "memory_projection": 70778880, + "vae_total": 126892531, + "vae_trainable": 0 + }, + "dtype": "bfloat16", + "batch_size": 1, + "cfg_batch_size": 2, + "sampling": { + "solver": "UniPC", + "steps": 20, + "guidance_scale": 5.0, + "timestep_shift": 5.0, + "source": "scripts/infer_example.sh" + }, + "memory": { + "profile": "stm6_topk27", + "use_ltm_attn": true, + "stm_sliding_window": 6, + "ltm_topk": 27, + "ltm_tile_size": [ + 1, + 8, + 8 + ], + "num_frame_per_block": 4, + "cached_past_latents": 2, + "tokens_per_latent": 880, + "dense_key_tokens_per_query": 5280, + "fine_key_tokens_per_query_upper_bound": 1728, + "scope": "dense local context and sparse fine-attention reads; not a whole-frame or storage cap", + "candidate_history": "all committed history retained; coarse ranking includes all history and current block", + "temporal_caveat": "4x temporal VAE and 4-latent joint generation; first noisy block has 3 observed latents", + "final_logical_bytes": 99850327680, + "final_storage_bytes": 101147940480 + }, + "history_frames": 600, + "history_source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699 + ], + "history_left_padding": 1, + "condition_latents": 151, + "total_latents": 276, + "generated_frames": 500, + "scored_frames": 500, + "future_padding": 0, + "action_mapping": "official WorldMem 25 channels, no scale changes; prepend one noop", + "pose_mapping": "translate xyz relative to source100; negate yaw; angles modulo360; padded source indices[::4]", + "latent_pose_source_indices": [ + 100, + 103, + 107, + 111, + 115, + 119, + 123, + 127, + 131, + 135, + 139, + 143, + 147, + 151, + 155, + 159, + 163, + 167, + 171, + 175, + 179, + 183, + 187, + 191, + 195, + 199, + 203, + 207, + 211, + 215, + 219, + 223, + 227, + 231, + 235, + 239, + 243, + 247, + 251, + 255, + 259, + 263, + 267, + 271, + 275, + 279, + 283, + 287, + 291, + 295, + 299, + 303, + 307, + 311, + 315, + 319, + 323, + 327, + 331, + 335, + 339, + 343, + 347, + 351, + 355, + 359, + 363, + 367, + 371, + 375, + 379, + 383, + 387, + 391, + 395, + 399, + 403, + 407, + 411, + 415, + 419, + 423, + 427, + 431, + 435, + 439, + 443, + 447, + 451, + 455, + 459, + 463, + 467, + 471, + 475, + 479, + 483, + 487, + 491, + 495, + 499, + 503, + 507, + 511, + 515, + 519, + 523, + 527, + 531, + 535, + 539, + 543, + 547, + 551, + 555, + 559, + 563, + 567, + 571, + 575, + 579, + 583, + 587, + 591, + 595, + 599, + 603, + 607, + 611, + 615, + 619, + 623, + 627, + 631, + 635, + 639, + 643, + 647, + 651, + 655, + 659, + 663, + 667, + 671, + 675, + 679, + 683, + 687, + 691, + 695, + 699, + 703, + 707, + 711, + 715, + 719, + 723, + 727, + 731, + 735, + 739, + 743, + 747, + 751, + 755, + 759, + 763, + 767, + 771, + 775, + 779, + 783, + 787, + 791, + 795, + 799, + 803, + 807, + 811, + 815, + 819, + 823, + 827, + 831, + 835, + 839, + 843, + 847, + 851, + 855, + 859, + 863, + 867, + 871, + 875, + 879, + 883, + 887, + 891, + 895, + 899, + 903, + 907, + 911, + 915, + 919, + 923, + 927, + 931, + 935, + 939, + 943, + 947, + 951, + 955, + 959, + 963, + 967, + 971, + 975, + 979, + 983, + 987, + 991, + 995, + 999, + 1003, + 1007, + 1011, + 1015, + 1019, + 1023, + 1027, + 1031, + 1035, + 1039, + 1043, + 1047, + 1051, + 1055, + 1059, + 1063, + 1067, + 1071, + 1075, + 1079, + 1083, + 1087, + 1091, + 1095, + 1099, + 1103, + 1107, + 1111, + 1115, + 1119, + 1123, + 1127, + 1131, + 1135, + 1139, + 1143, + 1147, + 1151, + 1155, + 1159, + 1163, + 1167, + 1171, + 1175, + 1179, + 1183, + 1187, + 1191, + 1195, + 1199 + ], + "pose_matrix_precision": "float32 construction, then bfloat16 model input", + "prope_max_time": 60, + "prope_time_semantics": "native absolute latent indices;60 is not a sequence length cap", + "native_resolution": [ + 352, + 640 + ], + "preprocessing": "full-frame bilinear resize with antialias; no crop", + "target_contract": "shared fixed Oasis VAE reconstruction; no method VAE target substitution", + "seed": 43000258, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 43, + "generation_seconds": 224.5787472790107, + "peak_generation_vram_bytes": 107563141120, + "source_indices": [ + 100, + 101, + 102, + 103, + 104, + 105, + 106, + 107, + 108, + 109, + 110, + 111, + 112, + 113, + 114, + 115, + 116, + 117, + 118, + 119, + 120, + 121, + 122, + 123, + 124, + 125, + 126, + 127, + 128, + 129, + 130, + 131, + 132, + 133, + 134, + 135, + 136, + 137, + 138, + 139, + 140, + 141, + 142, + 143, + 144, + 145, + 146, + 147, + 148, + 149, + 150, + 151, + 152, + 153, + 154, + 155, + 156, + 157, + 158, + 159, + 160, + 161, + 162, + 163, + 164, + 165, + 166, + 167, + 168, + 169, + 170, + 171, + 172, + 173, + 174, + 175, + 176, + 177, + 178, + 179, + 180, + 181, + 182, + 183, + 184, + 185, + 186, + 187, + 188, + 189, + 190, + 191, + 192, + 193, + 194, + 195, + 196, + 197, + 198, + 199, + 200, + 201, + 202, + 203, + 204, + 205, + 206, + 207, + 208, + 209, + 210, + 211, + 212, + 213, + 214, + 215, + 216, + 217, + 218, + 219, + 220, + 221, + 222, + 223, + 224, + 225, + 226, + 227, + 228, + 229, + 230, + 231, + 232, + 233, + 234, + 235, + 236, + 237, + 238, + 239, + 240, + 241, + 242, + 243, + 244, + 245, + 246, + 247, + 248, + 249, + 250, + 251, + 252, + 253, + 254, + 255, + 256, + 257, + 258, + 259, + 260, + 261, + 262, + 263, + 264, + 265, + 266, + 267, + 268, + 269, + 270, + 271, + 272, + 273, + 274, + 275, + 276, + 277, + 278, + 279, + 280, + 281, + 282, + 283, + 284, + 285, + 286, + 287, + 288, + 289, + 290, + 291, + 292, + 293, + 294, + 295, + 296, + 297, + 298, + 299, + 300, + 301, + 302, + 303, + 304, + 305, + 306, + 307, + 308, + 309, + 310, + 311, + 312, + 313, + 314, + 315, + 316, + 317, + 318, + 319, + 320, + 321, + 322, + 323, + 324, + 325, + 326, + 327, + 328, + 329, + 330, + 331, + 332, + 333, + 334, + 335, + 336, + 337, + 338, + 339, + 340, + 341, + 342, + 343, + 344, + 345, + 346, + 347, + 348, + 349, + 350, + 351, + 352, + 353, + 354, + 355, + 356, + 357, + 358, + 359, + 360, + 361, + 362, + 363, + 364, + 365, + 366, + 367, + 368, + 369, + 370, + 371, + 372, + 373, + 374, + 375, + 376, + 377, + 378, + 379, + 380, + 381, + 382, + 383, + 384, + 385, + 386, + 387, + 388, + 389, + 390, + 391, + 392, + 393, + 394, + 395, + 396, + 397, + 398, + 399, + 400, + 401, + 402, + 403, + 404, + 405, + 406, + 407, + 408, + 409, + 410, + 411, + 412, + 413, + 414, + 415, + 416, + 417, + 418, + 419, + 420, + 421, + 422, + 423, + 424, + 425, + 426, + 427, + 428, + 429, + 430, + 431, + 432, + 433, + 434, + 435, + 436, + 437, + 438, + 439, + 440, + 441, + 442, + 443, + 444, + 445, + 446, + 447, + 448, + 449, + 450, + 451, + 452, + 453, + 454, + 455, + 456, + 457, + 458, + 459, + 460, + 461, + 462, + 463, + 464, + 465, + 466, + 467, + 468, + 469, + 470, + 471, + 472, + 473, + 474, + 475, + 476, + 477, + 478, + 479, + 480, + 481, + 482, + 483, + 484, + 485, + 486, + 487, + 488, + 489, + 490, + 491, + 492, + 493, + 494, + 495, + 496, + 497, + 498, + 499, + 500, + 501, + 502, + 503, + 504, + 505, + 506, + 507, + 508, + 509, + 510, + 511, + 512, + 513, + 514, + 515, + 516, + 517, + 518, + 519, + 520, + 521, + 522, + 523, + 524, + 525, + 526, + 527, + 528, + 529, + 530, + 531, + 532, + 533, + 534, + 535, + 536, + 537, + 538, + 539, + 540, + 541, + 542, + 543, + 544, + 545, + 546, + 547, + 548, + 549, + 550, + 551, + 552, + 553, + 554, + 555, + 556, + 557, + 558, + 559, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 572, + 573, + 574, + 575, + 576, + 577, + 578, + 579, + 580, + 581, + 582, + 583, + 584, + 585, + 586, + 587, + 588, + 589, + 590, + 591, + 592, + 593, + 594, + 595, + 596, + 597, + 598, + 599, + 600, + 601, + 602, + 603, + 604, + 605, + 606, + 607, + 608, + 609, + 610, + 611, + 612, + 613, + 614, + 615, + 616, + 617, + 618, + 619, + 620, + 621, + 622, + 623, + 624, + 625, + 626, + 627, + 628, + 629, + 630, + 631, + 632, + 633, + 634, + 635, + 636, + 637, + 638, + 639, + 640, + 641, + 642, + 643, + 644, + 645, + 646, + 647, + 648, + 649, + 650, + 651, + 652, + 653, + 654, + 655, + 656, + 657, + 658, + 659, + 660, + 661, + 662, + 663, + 664, + 665, + 666, + 667, + 668, + 669, + 670, + 671, + 672, + 673, + 674, + 675, + 676, + 677, + 678, + 679, + 680, + 681, + 682, + 683, + 684, + 685, + 686, + 687, + 688, + 689, + 690, + 691, + 692, + 693, + 694, + 695, + 696, + 697, + 698, + 699, + 700, + 701, + 702, + 703, + 704, + 705, + 706, + 707, + 708, + 709, + 710, + 711, + 712, + 713, + 714, + 715, + 716, + 717, + 718, + 719, + 720, + 721, + 722, + 723, + 724, + 725, + 726, + 727, + 728, + 729, + 730, + 731, + 732, + 733, + 734, + 735, + 736, + 737, + 738, + 739, + 740, + 741, + 742, + 743, + 744, + 745, + 746, + 747, + 748, + 749, + 750, + 751, + 752, + 753, + 754, + 755, + 756, + 757, + 758, + 759, + 760, + 761, + 762, + 763, + 764, + 765, + 766, + 767, + 768, + 769, + 770, + 771, + 772, + 773, + 774, + 775, + 776, + 777, + 778, + 779, + 780, + 781, + 782, + 783, + 784, + 785, + 786, + 787, + 788, + 789, + 790, + 791, + 792, + 793, + 794, + 795, + 796, + 797, + 798, + 799, + 800, + 801, + 802, + 803, + 804, + 805, + 806, + 807, + 808, + 809, + 810, + 811, + 812, + 813, + 814, + 815, + 816, + 817, + 818, + 819, + 820, + 821, + 822, + 823, + 824, + 825, + 826, + 827, + 828, + 829, + 830, + 831, + 832, + 833, + 834, + 835, + 836, + 837, + 838, + 839, + 840, + 841, + 842, + 843, + 844, + 845, + 846, + 847, + 848, + 849, + 850, + 851, + 852, + 853, + 854, + 855, + 856, + 857, + 858, + 859, + 860, + 861, + 862, + 863, + 864, + 865, + 866, + 867, + 868, + 869, + 870, + 871, + 872, + 873, + 874, + 875, + 876, + 877, + 878, + 879, + 880, + 881, + 882, + 883, + 884, + 885, + 886, + 887, + 888, + 889, + 890, + 891, + 892, + 893, + 894, + 895, + 896, + 897, + 898, + 899, + 900, + 901, + 902, + 903, + 904, + 905, + 906, + 907, + 908, + 909, + 910, + 911, + 912, + 913, + 914, + 915, + 916, + 917, + 918, + 919, + 920, + 921, + 922, + 923, + 924, + 925, + 926, + 927, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 935, + 936, + 937, + 938, + 939, + 940, + 941, + 942, + 943, + 944, + 945, + 946, + 947, + 948, + 949, + 950, + 951, + 952, + 953, + 954, + 955, + 956, + 957, + 958, + 959, + 960, + 961, + 962, + 963, + 964, + 965, + 966, + 967, + 968, + 969, + 970, + 971, + 972, + 973, + 974, + 975, + 976, + 977, + 978, + 979, + 980, + 981, + 982, + 983, + 984, + 985, + 986, + 987, + 988, + 989, + 990, + 991, + 992, + 993, + 994, + 995, + 996, + 997, + 998, + 999, + 1000, + 1001, + 1002, + 1003, + 1004, + 1005, + 1006, + 1007, + 1008, + 1009, + 1010, + 1011, + 1012, + 1013, + 1014, + 1015, + 1016, + 1017, + 1018, + 1019, + 1020, + 1021, + 1022, + 1023, + 1024, + 1025, + 1026, + 1027, + 1028, + 1029, + 1030, + 1031, + 1032, + 1033, + 1034, + 1035, + 1036, + 1037, + 1038, + 1039, + 1040, + 1041, + 1042, + 1043, + 1044, + 1045, + 1046, + 1047, + 1048, + 1049, + 1050, + 1051, + 1052, + 1053, + 1054, + 1055, + 1056, + 1057, + 1058, + 1059, + 1060, + 1061, + 1062, + 1063, + 1064, + 1065, + 1066, + 1067, + 1068, + 1069, + 1070, + 1071, + 1072, + 1073, + 1074, + 1075, + 1076, + 1077, + 1078, + 1079, + 1080, + 1081, + 1082, + 1083, + 1084, + 1085, + 1086, + 1087, + 1088, + 1089, + 1090, + 1091, + 1092, + 1093, + 1094, + 1095, + 1096, + 1097, + 1098, + 1099, + 1100, + 1101, + 1102, + 1103, + 1104, + 1105, + 1106, + 1107, + 1108, + 1109, + 1110, + 1111, + 1112, + 1113, + 1114, + 1115, + 1116, + 1117, + 1118, + 1119, + 1120, + 1121, + 1122, + 1123, + 1124, + 1125, + 1126, + 1127, + 1128, + 1129, + 1130, + 1131, + 1132, + 1133, + 1134, + 1135, + 1136, + 1137, + 1138, + 1139, + 1140, + 1141, + 1142, + 1143, + 1144, + 1145, + 1146, + 1147, + 1148, + 1149, + 1150, + 1151, + 1152, + 1153, + 1154, + 1155, + 1156, + 1157, + 1158, + 1159, + 1160, + 1161, + 1162, + 1163, + 1164, + 1165, + 1166, + 1167, + 1168, + 1169, + 1170, + 1171, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1179, + 1180, + 1181, + 1182, + 1183, + 1184, + 1185, + 1186, + 1187, + 1188, + 1189, + 1190, + 1191, + 1192, + 1193, + 1194, + 1195, + 1196, + 1197, + 1198, + 1199 + ] +}