diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000050_test__val_savanna__1__000039/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000050_test__val_savanna__1__000039/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..4fe05192626dff651c37d3a29a56341c791e2f92 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000050_test__val_savanna__1__000039/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:165569c1ee7a677ae94e9d38c4034a8842399b48d2f5b95b7f2af400df5ce948 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000050_test__val_savanna__1__000039/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000050_test__val_savanna__1__000039/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..b8749b3f9cb372fbb9653b3cad8057f7f845bbef --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000050_test__val_savanna__1__000039/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "50", + "global_test_index": "50", + "dataset_index": "212", + "relative_path": "test/val_savanna/1/000039.mp4", + "clip_id": "test/val_savanna/1/000039", + "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": 42000050, + "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/1/000039_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": 0.01281773205846548, + "psnr": 19.16847801208496, + "lpips": 0.2268424928188324 + }, + "frames_100_200": { + "mse": 0.013117201626300812, + "psnr": 18.941247940063477, + "lpips": 0.2753637433052063 + }, + "frames_200_300": { + "mse": 0.010047070682048798, + "psnr": 20.15143394470215, + "lpips": 0.35678279399871826 + }, + "frames_300_400": { + "mse": 0.01987767592072487, + "psnr": 17.194063186645508, + "lpips": 0.26361024379730225 + }, + "frames_400_500": { + "mse": 0.0227480698376894, + "psnr": 17.719972610473633, + "lpips": 0.12680283188819885 + }, + "overall": { + "mse": 0.015721548348665237, + "psnr": 18.635038375854492, + "lpips": 0.24988043308258057 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000051_test__val_desert__0__000001/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000051_test__val_desert__0__000001/case.json new file mode 100644 index 0000000000000000000000000000000000000000..ba83f09b16dcf665225f4f46e409843ff23b8164 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000051_test__val_desert__0__000001/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "51", + "global_test_index": "51", + "dataset_index": "4", + "relative_path": "test/val_desert/0/000001.mp4", + "clip_id": "test/val_desert/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": 42000051, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.08757628500462, + "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_seed42_job85134/cases/g000051_test__val_desert__0__000001/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000051_test__val_desert__0__000001/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..e9e8c4e8844fbe5309177273ba8d525ecbbb3186 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000051_test__val_desert__0__000001/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8cd29a823822965e745139ed261e1a3e1516fd96aac7601176aa5bec6a90dbe4 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000051_test__val_desert__0__000001/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000051_test__val_desert__0__000001/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..f323d6cec7af3cb34fee99da06db502d782e4cea --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000051_test__val_desert__0__000001/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "51", + "global_test_index": "51", + "dataset_index": "4", + "relative_path": "test/val_desert/0/000001.mp4", + "clip_id": "test/val_desert/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": 42000051, + "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/0/000001_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": 0.006593554746359587, + "psnr": 21.914018630981445, + "lpips": 0.2264053225517273 + }, + "frames_100_200": { + "mse": 0.007773218676447868, + "psnr": 21.13860321044922, + "lpips": 0.37629470229148865 + }, + "frames_200_300": { + "mse": 0.020034629851579666, + "psnr": 17.2299861907959, + "lpips": 0.4506862759590149 + }, + "frames_300_400": { + "mse": 0.03867950662970543, + "psnr": 14.204030990600586, + "lpips": 0.5283481478691101 + }, + "frames_400_500": { + "mse": 0.03607636317610741, + "psnr": 14.730402946472168, + "lpips": 0.5221948623657227 + }, + "overall": { + "mse": 0.021831456571817398, + "psnr": 17.843408584594727, + "lpips": 0.4207858145236969 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000052_test__val_desert__0__000002/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000052_test__val_desert__0__000002/case.json new file mode 100644 index 0000000000000000000000000000000000000000..9a2ba33dd3b47f68b3d4423cc4c2cc3943910009 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000052_test__val_desert__0__000002/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "52", + "global_test_index": "52", + "dataset_index": "7", + "relative_path": "test/val_desert/0/000002.mp4", + "clip_id": "test/val_desert/0/000002", + "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": 42000052, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 224.5518935790751, + "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_seed42_job85134/cases/g000052_test__val_desert__0__000002/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000052_test__val_desert__0__000002/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..02d824b74bc37df58f7b9d7212f7f5ff789e30d7 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000052_test__val_desert__0__000002/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b8de34f9a06d97bd462b58ff0fe1ee303b706ee29d87f9edb102a6ee8adcbb3a +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000052_test__val_desert__0__000002/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000052_test__val_desert__0__000002/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..975e6012e6626c3579d07f92bfc9a159ce47adac --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000052_test__val_desert__0__000002/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "52", + "global_test_index": "52", + "dataset_index": "7", + "relative_path": "test/val_desert/0/000002.mp4", + "clip_id": "test/val_desert/0/000002", + "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": 42000052, + "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/0/000002_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": 0.0064200954511761665, + "psnr": 21.98798179626465, + "lpips": 0.23029834032058716 + }, + "frames_100_200": { + "mse": 0.007390241604298353, + "psnr": 21.48937225341797, + "lpips": 0.37565118074417114 + }, + "frames_200_300": { + "mse": 0.006831190083175898, + "psnr": 21.779775619506836, + "lpips": 0.40219739079475403 + }, + "frames_300_400": { + "mse": 0.011153723113238811, + "psnr": 20.55632972717285, + "lpips": 0.371099591255188 + }, + "frames_400_500": { + "mse": 0.02551632933318615, + "psnr": 16.59532928466797, + "lpips": 0.41022568941116333 + }, + "overall": { + "mse": 0.011462315917015076, + "psnr": 20.48175811767578, + "lpips": 0.3578944206237793 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000053_test__val_iceplains__0__000026/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000053_test__val_iceplains__0__000026/case.json new file mode 100644 index 0000000000000000000000000000000000000000..c5cae10812ac2132384e080d216042059ce7f476 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000053_test__val_iceplains__0__000026/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "53", + "global_test_index": "53", + "dataset_index": "310", + "relative_path": "test/val_iceplains/0/000026.mp4", + "clip_id": "test/val_iceplains/0/000026", + "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": 42000053, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.46176380396355, + "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_seed42_job85134/cases/g000053_test__val_iceplains__0__000026/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000053_test__val_iceplains__0__000026/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..4ec6dbfef72084b3b0eb89d6248cd220e6665a61 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000053_test__val_iceplains__0__000026/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ab7c95545e97a439d1bbc457d31ec0a9572de0ac245c2a5bd10916370384cbf9 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000053_test__val_iceplains__0__000026/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000053_test__val_iceplains__0__000026/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..6aa199c10c684db749eab8467d4fa84961c20f80 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000053_test__val_iceplains__0__000026/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "53", + "global_test_index": "53", + "dataset_index": "310", + "relative_path": "test/val_iceplains/0/000026.mp4", + "clip_id": "test/val_iceplains/0/000026", + "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": 42000053, + "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/0/000026_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": 0.0018658031476661563, + "psnr": 27.841737747192383, + "lpips": 0.03534121811389923 + }, + "frames_100_200": { + "mse": 0.005697420332580805, + "psnr": 25.249099731445312, + "lpips": 0.054211169481277466 + }, + "frames_200_300": { + "mse": 0.00203794171102345, + "psnr": 28.29175567626953, + "lpips": 0.027935175225138664 + }, + "frames_300_400": { + "mse": 0.0008801512885838747, + "psnr": 30.679689407348633, + "lpips": 0.013656738214194775 + }, + "frames_400_500": { + "mse": 0.00218679616227746, + "psnr": 27.69550323486328, + "lpips": 0.03809478133916855 + }, + "overall": { + "mse": 0.002533622784540057, + "psnr": 27.951557159423828, + "lpips": 0.03384781628847122 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000054_test__val_iceplains__0__000015/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000054_test__val_iceplains__0__000015/case.json new file mode 100644 index 0000000000000000000000000000000000000000..7e1427ea06cd0d7e5a393687fdf42c5244dc85f8 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000054_test__val_iceplains__0__000015/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "54", + "global_test_index": "54", + "dataset_index": "284", + "relative_path": "test/val_iceplains/0/000015.mp4", + "clip_id": "test/val_iceplains/0/000015", + "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": 42000054, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.05279156798497, + "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_seed42_job85134/cases/g000054_test__val_iceplains__0__000015/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000054_test__val_iceplains__0__000015/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..0d0da8dd573d7018b8f0a31f1828f009ee24db29 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000054_test__val_iceplains__0__000015/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:32707022364498cf56ab7dcce845d43e04906e56f9091e9e5710b58b7d84263a +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000054_test__val_iceplains__0__000015/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000054_test__val_iceplains__0__000015/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..6543d73bbd1d6d1b5e4a65d934ecafcbd52855e8 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000054_test__val_iceplains__0__000015/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "54", + "global_test_index": "54", + "dataset_index": "284", + "relative_path": "test/val_iceplains/0/000015.mp4", + "clip_id": "test/val_iceplains/0/000015", + "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": 42000054, + "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/0/000015_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": 0.008303030394017696, + "psnr": 22.139076232910156, + "lpips": 0.1755104511976242 + }, + "frames_100_200": { + "mse": 0.0045371041633188725, + "psnr": 24.06930160522461, + "lpips": 0.16189704835414886 + }, + "frames_200_300": { + "mse": 0.0305424015969038, + "psnr": 17.58099937438965, + "lpips": 0.24416469037532806 + }, + "frames_300_400": { + "mse": 0.1119973361492157, + "psnr": 9.946927070617676, + "lpips": 0.4834718406200409 + }, + "frames_400_500": { + "mse": 0.06628458946943283, + "psnr": 12.845129013061523, + "lpips": 0.39761078357696533 + }, + "overall": { + "mse": 0.044332895427942276, + "psnr": 17.316287994384766, + "lpips": 0.29253098368644714 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000055_test__val_desert__3__000041/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000055_test__val_desert__3__000041/case.json new file mode 100644 index 0000000000000000000000000000000000000000..d56e7f63a5f67562b9029b6cb1cd8104b5c6f83e --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000055_test__val_desert__3__000041/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "55", + "global_test_index": "55", + "dataset_index": "109", + "relative_path": "test/val_desert/3/000041.mp4", + "clip_id": "test/val_desert/3/000041", + "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": 42000055, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 223.87361758307088, + "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_seed42_job85134/cases/g000055_test__val_desert__3__000041/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000055_test__val_desert__3__000041/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..57e188bb58d2d7a95f3fc241f615e717de032050 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000055_test__val_desert__3__000041/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:91b98b5d516f01e8b44a7b901fa8b3ce930b3fa6111198553910795eacaca64e +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000055_test__val_desert__3__000041/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000055_test__val_desert__3__000041/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..f5ab3a82b25323d632519fbb63124b6c0e07d480 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000055_test__val_desert__3__000041/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "55", + "global_test_index": "55", + "dataset_index": "109", + "relative_path": "test/val_desert/3/000041.mp4", + "clip_id": "test/val_desert/3/000041", + "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": 42000055, + "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/3/000041_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": 0.0007929328712634742, + "psnr": 31.014480590820312, + "lpips": 0.010584792122244835 + }, + "frames_100_200": { + "mse": 0.005065763369202614, + "psnr": 24.539955139160156, + "lpips": 0.11547103524208069 + }, + "frames_200_300": { + "mse": 0.007749306503683329, + "psnr": 21.16834259033203, + "lpips": 0.29419413208961487 + }, + "frames_300_400": { + "mse": 0.007736444007605314, + "psnr": 21.937335968017578, + "lpips": 0.12103287875652313 + }, + "frames_400_500": { + "mse": 0.002124630846083164, + "psnr": 29.129228591918945, + "lpips": 0.022363809868693352 + }, + "overall": { + "mse": 0.004693815018981695, + "psnr": 25.5578670501709, + "lpips": 0.11272934079170227 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000056_test__val_desert__0__000018/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000056_test__val_desert__0__000018/case.json new file mode 100644 index 0000000000000000000000000000000000000000..704cfb18299b160bc1ab599fe460dd00c34473af --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000056_test__val_desert__0__000018/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "56", + "global_test_index": "56", + "dataset_index": "45", + "relative_path": "test/val_desert/0/000018.mp4", + "clip_id": "test/val_desert/0/000018", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_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": 42000056, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 223.6321197010111, + "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_seed42_job85134/cases/g000056_test__val_desert__0__000018/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000056_test__val_desert__0__000018/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..a0db0f5e7b858eb5338e8c2b2b690aefadba413b --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000056_test__val_desert__0__000018/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b833678829369a2702d0f4aa57db792c943179df1b68213ae1e1a0fcbff78c78 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000056_test__val_desert__0__000018/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000056_test__val_desert__0__000018/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..7614831dec6a842e00d9cba150dff7c7657e7cee --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000056_test__val_desert__0__000018/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "56", + "global_test_index": "56", + "dataset_index": "45", + "relative_path": "test/val_desert/0/000018.mp4", + "clip_id": "test/val_desert/0/000018", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_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": 42000056, + "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/0/000018_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": 0.006088231690227985, + "psnr": 22.45157241821289, + "lpips": 0.19701290130615234 + }, + "frames_100_200": { + "mse": 0.009286098182201385, + "psnr": 20.43706703186035, + "lpips": 0.33266693353652954 + }, + "frames_200_300": { + "mse": 0.012429223395884037, + "psnr": 19.222253799438477, + "lpips": 0.33672207593917847 + }, + "frames_300_400": { + "mse": 0.009608939290046692, + "psnr": 20.20525360107422, + "lpips": 0.3575161099433899 + }, + "frames_400_500": { + "mse": 0.00916953943669796, + "psnr": 20.41895294189453, + "lpips": 0.34661537408828735 + }, + "overall": { + "mse": 0.009316406212747097, + "psnr": 20.547019958496094, + "lpips": 0.31410667300224304 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000057_test__val_desert__0__000035/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000057_test__val_desert__0__000035/case.json new file mode 100644 index 0000000000000000000000000000000000000000..0acad5ac8dcf4e4aa45e3086caa010c4004eec94 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000057_test__val_desert__0__000035/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "57", + "global_test_index": "57", + "dataset_index": "93", + "relative_path": "test/val_desert/0/000035.mp4", + "clip_id": "test/val_desert/0/000035", + "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": 42000057, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.7706102420343, + "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_seed42_job85134/cases/g000057_test__val_desert__0__000035/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000057_test__val_desert__0__000035/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..4f7371c3c79a43ef86eb6d540d1698d997b14edd --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000057_test__val_desert__0__000035/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:506ca8c28760f9a64cfb876e78073881d35013c5a2c913f4d66b9c22824711b9 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000057_test__val_desert__0__000035/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000057_test__val_desert__0__000035/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..b2d47671acb09622b0581894a6b52269bbff4331 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000057_test__val_desert__0__000035/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "57", + "global_test_index": "57", + "dataset_index": "93", + "relative_path": "test/val_desert/0/000035.mp4", + "clip_id": "test/val_desert/0/000035", + "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": 42000057, + "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/0/000035_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": 0.004532197955995798, + "psnr": 23.77544593811035, + "lpips": 0.07937413454055786 + }, + "frames_100_200": { + "mse": 0.0028935109730809927, + "psnr": 27.076318740844727, + "lpips": 0.043587204068899155 + }, + "frames_200_300": { + "mse": 0.0022081900388002396, + "psnr": 27.867591857910156, + "lpips": 0.039256080985069275 + }, + "frames_300_400": { + "mse": 0.001822782913222909, + "psnr": 28.343305587768555, + "lpips": 0.03512802720069885 + }, + "frames_400_500": { + "mse": 0.001926264725625515, + "psnr": 29.109539031982422, + "lpips": 0.024003958329558372 + }, + "overall": { + "mse": 0.0026765894144773483, + "psnr": 27.23444175720215, + "lpips": 0.044269878417253494 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000058_test__val_plains__3__000011/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000058_test__val_plains__3__000011/case.json new file mode 100644 index 0000000000000000000000000000000000000000..a3d7de1184b5686c36427cb4e5e5cce18634edb5 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000058_test__val_plains__3__000011/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "58", + "global_test_index": "58", + "dataset_index": "400", + "relative_path": "test/val_plains/3/000011.mp4", + "clip_id": "test/val_plains/3/000011", + "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": 42000058, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.29261356103234, + "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_seed42_job85134/cases/g000058_test__val_plains__3__000011/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000058_test__val_plains__3__000011/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..9127231e2fe0f8ef043116d4917f74270c74179b --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000058_test__val_plains__3__000011/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:308617c02b18e658cbb68b2892c9dc7e1dba3c18187bfe7432dbc64e2adf1d11 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000058_test__val_plains__3__000011/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000058_test__val_plains__3__000011/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..498947a4bda6c9fe707a933183e08ce6cbffb64b --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000058_test__val_plains__3__000011/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "58", + "global_test_index": "58", + "dataset_index": "400", + "relative_path": "test/val_plains/3/000011.mp4", + "clip_id": "test/val_plains/3/000011", + "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": 42000058, + "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/3/000011_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": 0.00603787787258625, + "psnr": 22.69359588623047, + "lpips": 0.3797861933708191 + }, + "frames_100_200": { + "mse": 0.01774163544178009, + "psnr": 18.300983428955078, + "lpips": 0.48931434750556946 + }, + "frames_200_300": { + "mse": 0.027612991631031036, + "psnr": 16.206443786621094, + "lpips": 0.49500373005867004 + }, + "frames_300_400": { + "mse": 0.02530408464372158, + "psnr": 16.11359977722168, + "lpips": 0.5465338230133057 + }, + "frames_400_500": { + "mse": 0.026118190959095955, + "psnr": 16.89780616760254, + "lpips": 0.5202139616012573 + }, + "overall": { + "mse": 0.020562954246997833, + "psnr": 18.042484283447266, + "lpips": 0.4861704111099243 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000059_test__val_iceplains__2__000021/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000059_test__val_iceplains__2__000021/case.json new file mode 100644 index 0000000000000000000000000000000000000000..8d9d8c4aed901ccff1d5125c6ad3be7b2b7247a3 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000059_test__val_iceplains__2__000021/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "59", + "global_test_index": "59", + "dataset_index": "300", + "relative_path": "test/val_iceplains/2/000021.mp4", + "clip_id": "test/val_iceplains/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": 42000059, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.35614393197466, + "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_seed42_job85134/cases/g000059_test__val_iceplains__2__000021/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000059_test__val_iceplains__2__000021/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..3e3aeb7b2503a712aa5508c2f17d94029e3db2af --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000059_test__val_iceplains__2__000021/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9b3c613f2404dadd4137622c7e664b1a3b81cdab6448d85cc23a7b77ddbb2d91 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000059_test__val_iceplains__2__000021/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000059_test__val_iceplains__2__000021/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..47a94a5d434e4fff3daae09d4b79c202aa949b44 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000059_test__val_iceplains__2__000021/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "59", + "global_test_index": "59", + "dataset_index": "300", + "relative_path": "test/val_iceplains/2/000021.mp4", + "clip_id": "test/val_iceplains/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": 42000059, + "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/2/000021_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": 0.018934069201350212, + "psnr": 17.89548110961914, + "lpips": 0.17162969708442688 + }, + "frames_100_200": { + "mse": 0.014202394522726536, + "psnr": 21.451112747192383, + "lpips": 0.11373543739318848 + }, + "frames_200_300": { + "mse": 0.027053222060203552, + "psnr": 15.804211616516113, + "lpips": 0.2272443175315857 + }, + "frames_300_400": { + "mse": 0.03401561826467514, + "psnr": 14.837555885314941, + "lpips": 0.21404288709163666 + }, + "frames_400_500": { + "mse": 0.02656516246497631, + "psnr": 15.911725044250488, + "lpips": 0.16110843420028687 + }, + "overall": { + "mse": 0.024154091253876686, + "psnr": 17.180015563964844, + "lpips": 0.17755214869976044 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000060_test__val_savanna__2__000043/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000060_test__val_savanna__2__000043/case.json new file mode 100644 index 0000000000000000000000000000000000000000..cfb02ae7a419d109cb8ccccb6d4aaa8266650452 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000060_test__val_savanna__2__000043/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "60", + "global_test_index": "60", + "dataset_index": "226", + "relative_path": "test/val_savanna/2/000043.mp4", + "clip_id": "test/val_savanna/2/000043", + "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": 42000060, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 224.27725944609847, + "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_seed42_job85134/cases/g000060_test__val_savanna__2__000043/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000060_test__val_savanna__2__000043/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..b7b7faae91702d01b0e143ce6e17fca3280d4039 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000060_test__val_savanna__2__000043/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2e3f13df0886b218408e2c11b4f01b0836cd0821c28c6a0e4225223d781fef42 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000060_test__val_savanna__2__000043/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000060_test__val_savanna__2__000043/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..98def7643c0643bae421ea3f704a5d053e59a731 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000060_test__val_savanna__2__000043/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "60", + "global_test_index": "60", + "dataset_index": "226", + "relative_path": "test/val_savanna/2/000043.mp4", + "clip_id": "test/val_savanna/2/000043", + "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": 42000060, + "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/2/000043_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": 0.01737293228507042, + "psnr": 18.086406707763672, + "lpips": 0.3246663808822632 + }, + "frames_100_200": { + "mse": 0.03770436719059944, + "psnr": 14.984142303466797, + "lpips": 0.437561571598053 + }, + "frames_200_300": { + "mse": 0.018243083730340004, + "psnr": 17.667831420898438, + "lpips": 0.4050629436969757 + }, + "frames_300_400": { + "mse": 0.017339542508125305, + "psnr": 17.925874710083008, + "lpips": 0.3813725709915161 + }, + "frames_400_500": { + "mse": 0.011507952585816383, + "psnr": 19.753137588500977, + "lpips": 0.4251561462879181 + }, + "overall": { + "mse": 0.0204335767775774, + "psnr": 17.68347930908203, + "lpips": 0.39476391673088074 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000061_test__val_plains__2__000018/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000061_test__val_plains__2__000018/case.json new file mode 100644 index 0000000000000000000000000000000000000000..7375f56ad3f617c04c2dde5cc495061f16cc7c81 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000061_test__val_plains__2__000018/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "61", + "global_test_index": "61", + "dataset_index": "417", + "relative_path": "test/val_plains/2/000018.mp4", + "clip_id": "test/val_plains/2/000018", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_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": 42000061, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.05573471600655, + "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_seed42_job85134/cases/g000061_test__val_plains__2__000018/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000061_test__val_plains__2__000018/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..08ec8976c0283ec49e8569699abdd8fd5e43fdbf --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000061_test__val_plains__2__000018/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7ad3aba95babde3567d1e22c38b8479cfe55d27775cec23ed5b5b6105e850b85 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000061_test__val_plains__2__000018/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000061_test__val_plains__2__000018/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..2da3f1bc937ad990b6b68370f3ecb8cfefc1f9e3 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000061_test__val_plains__2__000018/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "61", + "global_test_index": "61", + "dataset_index": "417", + "relative_path": "test/val_plains/2/000018.mp4", + "clip_id": "test/val_plains/2/000018", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_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": 42000061, + "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/2/000018_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": 0.010624246671795845, + "psnr": 21.020343780517578, + "lpips": 0.24182870984077454 + }, + "frames_100_200": { + "mse": 0.02053832821547985, + "psnr": 18.039464950561523, + "lpips": 0.3289952576160431 + }, + "frames_200_300": { + "mse": 0.007536681368947029, + "psnr": 21.603212356567383, + "lpips": 0.30272597074508667 + }, + "frames_300_400": { + "mse": 0.004895126912742853, + "psnr": 23.151559829711914, + "lpips": 0.3585084080696106 + }, + "frames_400_500": { + "mse": 0.004367949441075325, + "psnr": 23.63399887084961, + "lpips": 0.40432724356651306 + }, + "overall": { + "mse": 0.009592466056346893, + "psnr": 21.489717483520508, + "lpips": 0.3272770941257477 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000062_test__val_desert__3__000046/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000062_test__val_desert__3__000046/case.json new file mode 100644 index 0000000000000000000000000000000000000000..555d8fa510a6bd65c1599270a8e9017074ce7999 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000062_test__val_desert__3__000046/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "62", + "global_test_index": "62", + "dataset_index": "120", + "relative_path": "test/val_desert/3/000046.mp4", + "clip_id": "test/val_desert/3/000046", + "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": 42000062, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 224.82224708097056, + "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_seed42_job85134/cases/g000062_test__val_desert__3__000046/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000062_test__val_desert__3__000046/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..9e9ae67fc3c159441d29d8d6ab7ba7005e4f3ea9 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000062_test__val_desert__3__000046/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:727b1a7d92055cec91668ffaba69750961c5af1ec2b3ddd492a9655d99a02a69 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000062_test__val_desert__3__000046/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000062_test__val_desert__3__000046/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..2649ba3fb11a7d6d5748697075cdfa59ff6f3d02 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000062_test__val_desert__3__000046/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "62", + "global_test_index": "62", + "dataset_index": "120", + "relative_path": "test/val_desert/3/000046.mp4", + "clip_id": "test/val_desert/3/000046", + "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": 42000062, + "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/3/000046_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": 0.012358702719211578, + "psnr": 19.15817642211914, + "lpips": 0.3659857213497162 + }, + "frames_100_200": { + "mse": 0.03631207346916199, + "psnr": 16.170042037963867, + "lpips": 0.357427179813385 + }, + "frames_200_300": { + "mse": 0.020444028079509735, + "psnr": 18.358335494995117, + "lpips": 0.33876219391822815 + }, + "frames_300_400": { + "mse": 0.012078933417797089, + "psnr": 19.390216827392578, + "lpips": 0.33182767033576965 + }, + "frames_400_500": { + "mse": 0.014504080638289452, + "psnr": 18.59488296508789, + "lpips": 0.32638099789619446 + }, + "overall": { + "mse": 0.01913956180214882, + "psnr": 18.33432960510254, + "lpips": 0.3440767526626587 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000063_test__val_desert__3__000040/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000063_test__val_desert__3__000040/case.json new file mode 100644 index 0000000000000000000000000000000000000000..4b7496545321b6429af99a57e267ccee681a3a63 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000063_test__val_desert__3__000040/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "63", + "global_test_index": "63", + "dataset_index": "106", + "relative_path": "test/val_desert/3/000040.mp4", + "clip_id": "test/val_desert/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": 42000063, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 223.8858439399628, + "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_seed42_job85134/cases/g000063_test__val_desert__3__000040/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000063_test__val_desert__3__000040/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..2e95945dcbe7d8886b75e7c2843a559be63ef0f2 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000063_test__val_desert__3__000040/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:74ed6d199462917bbe3be3f9bc2f6dd7653c385812148f588817eea7dfb044e4 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000063_test__val_desert__3__000040/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000063_test__val_desert__3__000040/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..7e69590d199ec62c3fa64dc75599a6c157172bcd --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000063_test__val_desert__3__000040/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "63", + "global_test_index": "63", + "dataset_index": "106", + "relative_path": "test/val_desert/3/000040.mp4", + "clip_id": "test/val_desert/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": 42000063, + "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/3/000040_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": 0.0007784578483551741, + "psnr": 31.101484298706055, + "lpips": 0.010347029194235802 + }, + "frames_100_200": { + "mse": 0.0008519005496054888, + "psnr": 30.953893661499023, + "lpips": 0.011311287060379982 + }, + "frames_200_300": { + "mse": 0.0012531774118542671, + "psnr": 29.801176071166992, + "lpips": 0.015969354659318924 + }, + "frames_300_400": { + "mse": 0.004415316507220268, + "psnr": 24.23685073852539, + "lpips": 0.07857000827789307 + }, + "frames_400_500": { + "mse": 0.007315449882298708, + "psnr": 21.456235885620117, + "lpips": 0.14613014459609985 + }, + "overall": { + "mse": 0.0029228604398667812, + "psnr": 27.509929656982422, + "lpips": 0.052465565502643585 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000064_test__val_iceplains__1__000039/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000064_test__val_iceplains__1__000039/case.json new file mode 100644 index 0000000000000000000000000000000000000000..a838dfaf5a2428399b3da15de34234e867cad30c --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000064_test__val_iceplains__1__000039/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "64", + "global_test_index": "64", + "dataset_index": "345", + "relative_path": "test/val_iceplains/1/000039.mp4", + "clip_id": "test/val_iceplains/1/000039", + "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": 42000064, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 223.93702536204364, + "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_seed42_job85134/cases/g000064_test__val_iceplains__1__000039/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000064_test__val_iceplains__1__000039/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..c6810c086bd2cb763d233a9df76da8d44b328e9b --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000064_test__val_iceplains__1__000039/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f0b72f9c44095e9c1367f8b498d0865a6d3dbeceaa7b858c16ac7b7cbf5fae94 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000064_test__val_iceplains__1__000039/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000064_test__val_iceplains__1__000039/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..946e26246282fceb8787277217a6c78857fa18b7 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000064_test__val_iceplains__1__000039/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "64", + "global_test_index": "64", + "dataset_index": "345", + "relative_path": "test/val_iceplains/1/000039.mp4", + "clip_id": "test/val_iceplains/1/000039", + "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": 42000064, + "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/1/000039_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": 0.01983753591775894, + "psnr": 17.218830108642578, + "lpips": 0.249090313911438 + }, + "frames_100_200": { + "mse": 0.032760947942733765, + "psnr": 15.878928184509277, + "lpips": 0.3520737588405609 + }, + "frames_200_300": { + "mse": 0.0912536084651947, + "psnr": 10.58728313446045, + "lpips": 0.5581620335578918 + }, + "frames_300_400": { + "mse": 0.04389288276433945, + "psnr": 14.705743789672852, + "lpips": 0.440093457698822 + }, + "frames_400_500": { + "mse": 0.0257883220911026, + "psnr": 16.127197265625, + "lpips": 0.4100409746170044 + }, + "overall": { + "mse": 0.04270665720105171, + "psnr": 14.903595924377441, + "lpips": 0.4018920958042145 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000065_test__val_savanna__2__000016/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000065_test__val_savanna__2__000016/case.json new file mode 100644 index 0000000000000000000000000000000000000000..7389c8b8605c5b71d34190cf2756f5cd8b4f69f6 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000065_test__val_savanna__2__000016/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "65", + "global_test_index": "65", + "dataset_index": "169", + "relative_path": "test/val_savanna/2/000016.mp4", + "clip_id": "test/val_savanna/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": 42000065, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 224.74953948694747, + "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_seed42_job85134/cases/g000065_test__val_savanna__2__000016/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000065_test__val_savanna__2__000016/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..c98ff1f161cbe6b1ea3fd50e33eee444a6dacf6c --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000065_test__val_savanna__2__000016/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:06ad8c9ae7680e220efd59f0caf111256dca8027ed0d91eae6aad5b8f92d483c +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000065_test__val_savanna__2__000016/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000065_test__val_savanna__2__000016/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..ec4c5431271f7259c12f88675d6da9fb363741f6 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000065_test__val_savanna__2__000016/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "65", + "global_test_index": "65", + "dataset_index": "169", + "relative_path": "test/val_savanna/2/000016.mp4", + "clip_id": "test/val_savanna/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": 42000065, + "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/2/000016_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": 0.00432266341522336, + "psnr": 24.156435012817383, + "lpips": 0.23762348294258118 + }, + "frames_100_200": { + "mse": 0.011512440629303455, + "psnr": 19.731117248535156, + "lpips": 0.373709112405777 + }, + "frames_200_300": { + "mse": 0.04785365238785744, + "psnr": 14.448519706726074, + "lpips": 0.5341333150863647 + }, + "frames_300_400": { + "mse": 0.04272455349564552, + "psnr": 14.2890043258667, + "lpips": 0.5257330536842346 + }, + "frames_400_500": { + "mse": 0.03408496081829071, + "psnr": 15.440713882446289, + "lpips": 0.5538658499717712 + }, + "overall": { + "mse": 0.028099650517106056, + "psnr": 17.613155364990234, + "lpips": 0.4450129270553589 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000066_test__val_iceplains__1__000027/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000066_test__val_iceplains__1__000027/case.json new file mode 100644 index 0000000000000000000000000000000000000000..0a6a837fa8456ee0e2a5a613d346ad7d3d627c60 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000066_test__val_iceplains__1__000027/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "66", + "global_test_index": "66", + "dataset_index": "312", + "relative_path": "test/val_iceplains/1/000027.mp4", + "clip_id": "test/val_iceplains/1/000027", + "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": 42000066, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.3585003289627, + "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_seed42_job85134/cases/g000066_test__val_iceplains__1__000027/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000066_test__val_iceplains__1__000027/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..0baac229d535cc426620ca9343c2cd65c4fcace3 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000066_test__val_iceplains__1__000027/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d4ae3fed24feeeafa51bcdb9b2e08cfb80fbd129469922971681740541754441 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000066_test__val_iceplains__1__000027/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000066_test__val_iceplains__1__000027/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..d47b91478bdad9e2c05f20e329bb3a9e3661fe9e --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000066_test__val_iceplains__1__000027/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "66", + "global_test_index": "66", + "dataset_index": "312", + "relative_path": "test/val_iceplains/1/000027.mp4", + "clip_id": "test/val_iceplains/1/000027", + "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": 42000066, + "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/1/000027_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": 0.034344762563705444, + "psnr": 15.63531494140625, + "lpips": 0.2211812138557434 + }, + "frames_100_200": { + "mse": 0.0478346049785614, + "psnr": 13.427999496459961, + "lpips": 0.32975757122039795 + }, + "frames_200_300": { + "mse": 0.05932416394352913, + "psnr": 12.483528137207031, + "lpips": 0.3255031108856201 + }, + "frames_300_400": { + "mse": 0.04627980664372444, + "psnr": 13.973345756530762, + "lpips": 0.28107750415802 + }, + "frames_400_500": { + "mse": 0.022097134962677956, + "psnr": 16.705860137939453, + "lpips": 0.1758783757686615 + }, + "overall": { + "mse": 0.04197609797120094, + "psnr": 14.445208549499512, + "lpips": 0.2666795551776886 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000067_test__val_iceplains__0__000048/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000067_test__val_iceplains__0__000048/case.json new file mode 100644 index 0000000000000000000000000000000000000000..394874d1715e5ddd7401d4a9ac04100ae08666ce --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000067_test__val_iceplains__0__000048/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "67", + "global_test_index": "67", + "dataset_index": "368", + "relative_path": "test/val_iceplains/0/000048.mp4", + "clip_id": "test/val_iceplains/0/000048", + "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": 42000067, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.6357566670049, + "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_seed42_job85134/cases/g000067_test__val_iceplains__0__000048/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000067_test__val_iceplains__0__000048/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..f36e82b01bfef76c74fceb2e6304c1dcb89b1da1 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000067_test__val_iceplains__0__000048/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ae406ccb72c005b6df0b651bad2e080bdfaec7dc3e98d817459725b1e8f7ded6 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000067_test__val_iceplains__0__000048/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000067_test__val_iceplains__0__000048/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..18cba8976839f57bbe1504bb47129333ff9475be --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000067_test__val_iceplains__0__000048/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "67", + "global_test_index": "67", + "dataset_index": "368", + "relative_path": "test/val_iceplains/0/000048.mp4", + "clip_id": "test/val_iceplains/0/000048", + "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": 42000067, + "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/0/000048_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": 0.01891854964196682, + "psnr": 17.52164649963379, + "lpips": 0.1965583860874176 + }, + "frames_100_200": { + "mse": 0.018786415457725525, + "psnr": 17.4252872467041, + "lpips": 0.2274828404188156 + }, + "frames_200_300": { + "mse": 0.018185630440711975, + "psnr": 17.523998260498047, + "lpips": 0.1842530220746994 + }, + "frames_300_400": { + "mse": 0.0222228541970253, + "psnr": 16.65692138671875, + "lpips": 0.2092321366071701 + }, + "frames_400_500": { + "mse": 0.02721560001373291, + "psnr": 15.847143173217773, + "lpips": 0.24623443186283112 + }, + "overall": { + "mse": 0.021065808832645416, + "psnr": 16.9950008392334, + "lpips": 0.21275216341018677 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000068_test__val_plains__1__000034/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000068_test__val_plains__1__000034/case.json new file mode 100644 index 0000000000000000000000000000000000000000..d3b28e6d6d8b033824f2fc1ccc88edf9227b6833 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000068_test__val_plains__1__000034/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "68", + "global_test_index": "68", + "dataset_index": "449", + "relative_path": "test/val_plains/1/000034.mp4", + "clip_id": "test/val_plains/1/000034", + "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": 42000068, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 224.33758163300809, + "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_seed42_job85134/cases/g000068_test__val_plains__1__000034/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000068_test__val_plains__1__000034/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..947d3739660ad0c20a854c519694db1137c56d82 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000068_test__val_plains__1__000034/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:25870138d2ead69b75d61b936c2c7080d8638c3755af2748c5227c692dc2e81e +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000068_test__val_plains__1__000034/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000068_test__val_plains__1__000034/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..8940fe4345f2ed78c0ffce1a6aaf872c183c0db6 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000068_test__val_plains__1__000034/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "68", + "global_test_index": "68", + "dataset_index": "449", + "relative_path": "test/val_plains/1/000034.mp4", + "clip_id": "test/val_plains/1/000034", + "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": 42000068, + "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/1/000034_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": 0.03099524974822998, + "psnr": 19.124900817871094, + "lpips": 0.1926562339067459 + }, + "frames_100_200": { + "mse": 0.058294638991355896, + "psnr": 12.684697151184082, + "lpips": 0.5118173360824585 + }, + "frames_200_300": { + "mse": 0.0408584326505661, + "psnr": 14.397507667541504, + "lpips": 0.5659259557723999 + }, + "frames_300_400": { + "mse": 0.02882986329495907, + "psnr": 16.215368270874023, + "lpips": 0.545640766620636 + }, + "frames_400_500": { + "mse": 0.01810409687459469, + "psnr": 17.638633728027344, + "lpips": 0.5223133563995361 + }, + "overall": { + "mse": 0.03541645407676697, + "psnr": 16.012222290039062, + "lpips": 0.4676707088947296 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000069_test__val_desert__1__000004/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000069_test__val_desert__1__000004/case.json new file mode 100644 index 0000000000000000000000000000000000000000..22f8517a7cc8ba9473278797915b21f795edf4c2 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000069_test__val_desert__1__000004/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "69", + "global_test_index": "69", + "dataset_index": "11", + "relative_path": "test/val_desert/1/000004.mp4", + "clip_id": "test/val_desert/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": 42000069, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.00808653607965, + "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_seed42_job85134/cases/g000069_test__val_desert__1__000004/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000069_test__val_desert__1__000004/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..015d0e7472504cba2249f8752f5f45f1c86fcbbb --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000069_test__val_desert__1__000004/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:df1b5440ce2c1c8e7f8e6d4fe64252b9b35cbf5606460a467b0499d15283b5b1 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000069_test__val_desert__1__000004/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000069_test__val_desert__1__000004/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..a3e7029637b0b6756d9b7cebf09b9a6f1f0556e0 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000069_test__val_desert__1__000004/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "69", + "global_test_index": "69", + "dataset_index": "11", + "relative_path": "test/val_desert/1/000004.mp4", + "clip_id": "test/val_desert/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": 42000069, + "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/1/000004_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": 0.006474233698099852, + "psnr": 21.994840621948242, + "lpips": 0.24653521180152893 + }, + "frames_100_200": { + "mse": 0.008515804074704647, + "psnr": 21.195144653320312, + "lpips": 0.3559550940990448 + }, + "frames_200_300": { + "mse": 0.019080186262726784, + "psnr": 17.81348419189453, + "lpips": 0.4240451455116272 + }, + "frames_300_400": { + "mse": 0.040147315710783005, + "psnr": 14.523170471191406, + "lpips": 0.45699799060821533 + }, + "frames_400_500": { + "mse": 0.029666276648640633, + "psnr": 15.593818664550781, + "lpips": 0.4567188322544098 + }, + "overall": { + "mse": 0.020776759833097458, + "psnr": 18.224090576171875, + "lpips": 0.38805046677589417 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000070_test__val_desert__2__000001/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000070_test__val_desert__2__000001/case.json new file mode 100644 index 0000000000000000000000000000000000000000..efe172e07f1160ea7c49fab007b9303ff7dd7e8e --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000070_test__val_desert__2__000001/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "70", + "global_test_index": "70", + "dataset_index": "6", + "relative_path": "test/val_desert/2/000001.mp4", + "clip_id": "test/val_desert/2/000001", + "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": 42000070, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.23269413108937, + "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_seed42_job85134/cases/g000070_test__val_desert__2__000001/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000070_test__val_desert__2__000001/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..7841e53bb7f93c5d01288b35eb30b42e1a6429d2 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000070_test__val_desert__2__000001/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b84c57e6998229c18a96cca0b335149d8a5ce2508d196a92a0268dddcc4e17f2 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000070_test__val_desert__2__000001/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000070_test__val_desert__2__000001/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..ac0523bfa49909108e17702d74f088443e52c121 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000070_test__val_desert__2__000001/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "70", + "global_test_index": "70", + "dataset_index": "6", + "relative_path": "test/val_desert/2/000001.mp4", + "clip_id": "test/val_desert/2/000001", + "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": 42000070, + "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/2/000001_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": 0.003631430910900235, + "psnr": 24.635896682739258, + "lpips": 0.07331319153308868 + }, + "frames_100_200": { + "mse": 0.0038447075057774782, + "psnr": 24.637723922729492, + "lpips": 0.05779415741562843 + }, + "frames_200_300": { + "mse": 0.002402223413810134, + "psnr": 27.474475860595703, + "lpips": 0.042747482657432556 + }, + "frames_300_400": { + "mse": 0.0009212736622430384, + "psnr": 30.435815811157227, + "lpips": 0.013788094744086266 + }, + "frames_400_500": { + "mse": 0.001625376520678401, + "psnr": 30.02309799194336, + "lpips": 0.01575874350965023 + }, + "overall": { + "mse": 0.002485002623870969, + "psnr": 27.441402435302734, + "lpips": 0.04068033769726753 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000071_test__val_iceplains__0__000005/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000071_test__val_iceplains__0__000005/case.json new file mode 100644 index 0000000000000000000000000000000000000000..ed1f2ef5d1d32f5a1e594fd098579287d31c8aec --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000071_test__val_iceplains__0__000005/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "71", + "global_test_index": "71", + "dataset_index": "257", + "relative_path": "test/val_iceplains/0/000005.mp4", + "clip_id": "test/val_iceplains/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": 42000071, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 223.4902678070357, + "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_seed42_job85134/cases/g000071_test__val_iceplains__0__000005/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000071_test__val_iceplains__0__000005/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..a5e7cf0674e8ab16d52df5069ac830f9a289b5a5 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000071_test__val_iceplains__0__000005/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:be0777fc8a47f124c3f4159d8c3bea15bfc60db88658ac6a836d35b54c6d2c60 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000071_test__val_iceplains__0__000005/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000071_test__val_iceplains__0__000005/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..a4c228079afd8ee24c920c8a2691734093f1982b --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000071_test__val_iceplains__0__000005/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "71", + "global_test_index": "71", + "dataset_index": "257", + "relative_path": "test/val_iceplains/0/000005.mp4", + "clip_id": "test/val_iceplains/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": 42000071, + "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/0/000005_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": 0.017541376873850822, + "psnr": 17.827558517456055, + "lpips": 0.24842417240142822 + }, + "frames_100_200": { + "mse": 0.0058481162413954735, + "psnr": 22.796560287475586, + "lpips": 0.23929955065250397 + }, + "frames_200_300": { + "mse": 0.0979788601398468, + "psnr": 12.665699005126953, + "lpips": 0.5378389954566956 + }, + "frames_300_400": { + "mse": 0.012478145770728588, + "psnr": 21.3806095123291, + "lpips": 0.28695356845855713 + }, + "frames_400_500": { + "mse": 0.020038656890392303, + "psnr": 17.37555694580078, + "lpips": 0.37045273184776306 + }, + "overall": { + "mse": 0.030777031555771828, + "psnr": 18.409196853637695, + "lpips": 0.3365938067436218 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000072_test__val_iceplains__3__000006/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000072_test__val_iceplains__3__000006/case.json new file mode 100644 index 0000000000000000000000000000000000000000..abe7be7c04f6a3073ab5bbf3bdd0356cd80ad182 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000072_test__val_iceplains__3__000006/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "72", + "global_test_index": "72", + "dataset_index": "262", + "relative_path": "test/val_iceplains/3/000006.mp4", + "clip_id": "test/val_iceplains/3/000006", + "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": 42000072, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 223.89354079193436, + "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_seed42_job85134/cases/g000072_test__val_iceplains__3__000006/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000072_test__val_iceplains__3__000006/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..582380bf28f64f84f4557c6f70308d603b16e446 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000072_test__val_iceplains__3__000006/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:597dd2ec0c8fe06e174ebb5998c3f77f87e9213041da7726b6c02e9a7ac8858d +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000072_test__val_iceplains__3__000006/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000072_test__val_iceplains__3__000006/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..d43b73c57cec7c2e3e163d900ee13990fcd14dc5 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000072_test__val_iceplains__3__000006/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "72", + "global_test_index": "72", + "dataset_index": "262", + "relative_path": "test/val_iceplains/3/000006.mp4", + "clip_id": "test/val_iceplains/3/000006", + "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": 42000072, + "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/3/000006_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": 0.004975282121449709, + "psnr": 24.22063446044922, + "lpips": 0.14950962364673615 + }, + "frames_100_200": { + "mse": 0.018487509340047836, + "psnr": 18.7320613861084, + "lpips": 0.20004354417324066 + }, + "frames_200_300": { + "mse": 0.032595884054899216, + "psnr": 15.222149848937988, + "lpips": 0.20405934751033783 + }, + "frames_300_400": { + "mse": 0.023359036073088646, + "psnr": 16.9741153717041, + "lpips": 0.14787456393241882 + }, + "frames_400_500": { + "mse": 0.028329534456133842, + "psnr": 16.116252899169922, + "lpips": 0.24443906545639038 + }, + "overall": { + "mse": 0.021549450233578682, + "psnr": 18.253042221069336, + "lpips": 0.189185231924057 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000073_test__val_savanna__0__000015/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000073_test__val_savanna__0__000015/case.json new file mode 100644 index 0000000000000000000000000000000000000000..3f1e68171b222becc170973fd1ff7d893f535ca2 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000073_test__val_savanna__0__000015/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "73", + "global_test_index": "73", + "dataset_index": "164", + "relative_path": "test/val_savanna/0/000015.mp4", + "clip_id": "test/val_savanna/0/000015", + "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": 42000073, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.1403517529834, + "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_seed42_job85134/cases/g000073_test__val_savanna__0__000015/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000073_test__val_savanna__0__000015/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..37f8c1e70f86f979d1a214ce35b197dcfcab6b9c --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000073_test__val_savanna__0__000015/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b3a669afacf1710cc23eaa74479b2fe23064103420cf4d9bb62ee86a519e99c2 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000073_test__val_savanna__0__000015/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000073_test__val_savanna__0__000015/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..a4d009858f05727f463568f2b4e3a18714a1b61c --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000073_test__val_savanna__0__000015/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "73", + "global_test_index": "73", + "dataset_index": "164", + "relative_path": "test/val_savanna/0/000015.mp4", + "clip_id": "test/val_savanna/0/000015", + "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": 42000073, + "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/0/000015_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": 0.01695999689400196, + "psnr": 19.479995727539062, + "lpips": 0.11950689554214478 + }, + "frames_100_200": { + "mse": 0.03955571725964546, + "psnr": 14.696781158447266, + "lpips": 0.2820087671279907 + }, + "frames_200_300": { + "mse": 0.0713566318154335, + "psnr": 13.500188827514648, + "lpips": 0.3838798403739929 + }, + "frames_300_400": { + "mse": 0.15040099620819092, + "psnr": 8.939238548278809, + "lpips": 0.531254231929779 + }, + "frames_400_500": { + "mse": 0.15533095598220825, + "psnr": 8.191690444946289, + "lpips": 0.5027943253517151 + }, + "overall": { + "mse": 0.08672086149454117, + "psnr": 12.961579322814941, + "lpips": 0.36388880014419556 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000074_test__val_iceplains__2__000036/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000074_test__val_iceplains__2__000036/case.json new file mode 100644 index 0000000000000000000000000000000000000000..fd91e963a89517818b1d3a2f92dbad6b6ed27d11 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000074_test__val_iceplains__2__000036/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "74", + "global_test_index": "74", + "dataset_index": "337", + "relative_path": "test/val_iceplains/2/000036.mp4", + "clip_id": "test/val_iceplains/2/000036", + "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": 42000074, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.87042201298755, + "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_seed42_job85134/cases/g000074_test__val_iceplains__2__000036/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000074_test__val_iceplains__2__000036/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..880fda60826a245fd9c838cbfafe1d66fad1a6c0 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000074_test__val_iceplains__2__000036/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3351cd473bedeaf13fd117e3c790358ac7908f4841fcc481b2a93b975d8c4676 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000074_test__val_iceplains__2__000036/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000074_test__val_iceplains__2__000036/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..8050ced53f5b54f677ec968fc352ccacc5d38eb9 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000074_test__val_iceplains__2__000036/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "74", + "global_test_index": "74", + "dataset_index": "337", + "relative_path": "test/val_iceplains/2/000036.mp4", + "clip_id": "test/val_iceplains/2/000036", + "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": 42000074, + "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/2/000036_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": 0.017849912866950035, + "psnr": 18.043781280517578, + "lpips": 0.13928359746932983 + }, + "frames_100_200": { + "mse": 0.01648661680519581, + "psnr": 18.70743179321289, + "lpips": 0.1086755245923996 + }, + "frames_200_300": { + "mse": 0.011084556579589844, + "psnr": 20.642316818237305, + "lpips": 0.09289691597223282 + }, + "frames_300_400": { + "mse": 0.006440398748964071, + "psnr": 22.318035125732422, + "lpips": 0.06184263154864311 + }, + "frames_400_500": { + "mse": 0.0015696571208536625, + "psnr": 29.210472106933594, + "lpips": 0.022646334022283554 + }, + "overall": { + "mse": 0.010686228051781654, + "psnr": 21.784406661987305, + "lpips": 0.08506901562213898 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000075_test__val_desert__0__000036/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000075_test__val_desert__0__000036/case.json new file mode 100644 index 0000000000000000000000000000000000000000..b8e7d23f08b1e1bb386b1581e9b2acc08f80d12a --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000075_test__val_desert__0__000036/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "75", + "global_test_index": "75", + "dataset_index": "95", + "relative_path": "test/val_desert/0/000036.mp4", + "clip_id": "test/val_desert/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": 42000075, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.4851709830109, + "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_seed42_job85134/cases/g000075_test__val_desert__0__000036/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000075_test__val_desert__0__000036/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..8f40f0695af16b6215a1c8206d88cceb1d3bf626 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000075_test__val_desert__0__000036/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e56fac4d680f77c45e71686f3f25c0dd99fcfb78df9ea5bd68a31f2b337758a2 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000075_test__val_desert__0__000036/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000075_test__val_desert__0__000036/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..79007925eac8e033283eb4b426746318aa67314b --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000075_test__val_desert__0__000036/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "75", + "global_test_index": "75", + "dataset_index": "95", + "relative_path": "test/val_desert/0/000036.mp4", + "clip_id": "test/val_desert/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": 42000075, + "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/0/000036_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": 0.002074980642646551, + "psnr": 28.464576721191406, + "lpips": 0.028036020696163177 + }, + "frames_100_200": { + "mse": 0.0007641555857844651, + "psnr": 31.178607940673828, + "lpips": 0.009907623752951622 + }, + "frames_200_300": { + "mse": 0.0017989210318773985, + "psnr": 28.466440200805664, + "lpips": 0.028146587312221527 + }, + "frames_300_400": { + "mse": 0.00750751793384552, + "psnr": 21.762958526611328, + "lpips": 0.13195376098155975 + }, + "frames_400_500": { + "mse": 0.01655874401330948, + "psnr": 18.106489181518555, + "lpips": 0.320051908493042 + }, + "overall": { + "mse": 0.005740863736718893, + "psnr": 25.595815658569336, + "lpips": 0.10361918061971664 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000076_test__val_iceplains__0__000013/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000076_test__val_iceplains__0__000013/case.json new file mode 100644 index 0000000000000000000000000000000000000000..a0b83ae7afbff4ea5c8a948db3b52d25043354bb --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000076_test__val_iceplains__0__000013/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "76", + "global_test_index": "76", + "dataset_index": "277", + "relative_path": "test/val_iceplains/0/000013.mp4", + "clip_id": "test/val_iceplains/0/000013", + "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": 42000076, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.93032986694016, + "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_seed42_job85134/cases/g000076_test__val_iceplains__0__000013/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000076_test__val_iceplains__0__000013/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..05c34dc36c8d2d7381610f2aed012f0c600235da --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000076_test__val_iceplains__0__000013/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f4340f2b63bf285846e246d911f40483a37af0cf7037d3b86484aa21153437ba +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000076_test__val_iceplains__0__000013/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000076_test__val_iceplains__0__000013/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..358103953b991eda113bcf660652e098a7b24a6f --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000076_test__val_iceplains__0__000013/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "76", + "global_test_index": "76", + "dataset_index": "277", + "relative_path": "test/val_iceplains/0/000013.mp4", + "clip_id": "test/val_iceplains/0/000013", + "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": 42000076, + "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/0/000013_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": 0.0011363907251507044, + "psnr": 30.20013427734375, + "lpips": 0.013569748029112816 + }, + "frames_100_200": { + "mse": 0.001628422411158681, + "psnr": 29.084508895874023, + "lpips": 0.015296811237931252 + }, + "frames_200_300": { + "mse": 0.002308582654222846, + "psnr": 28.134628295898438, + "lpips": 0.02107759192585945 + }, + "frames_300_400": { + "mse": 0.0011021294631063938, + "psnr": 30.591297149658203, + "lpips": 0.013963738456368446 + }, + "frames_400_500": { + "mse": 0.010899126529693604, + "psnr": 22.464397430419922, + "lpips": 0.12843382358551025 + }, + "overall": { + "mse": 0.003414930310100317, + "psnr": 28.09499168395996, + "lpips": 0.03846834599971771 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000077_test__val_desert__1__000010/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000077_test__val_desert__1__000010/case.json new file mode 100644 index 0000000000000000000000000000000000000000..ab2f9ebef7324e2bbf7899437472c795d069e196 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000077_test__val_desert__1__000010/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "77", + "global_test_index": "77", + "dataset_index": "26", + "relative_path": "test/val_desert/1/000010.mp4", + "clip_id": "test/val_desert/1/000010", + "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": 42000077, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.80574936594348, + "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_seed42_job85134/cases/g000077_test__val_desert__1__000010/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000077_test__val_desert__1__000010/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..bd4f606325aaa23e7edce18e8277f613104d7d08 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000077_test__val_desert__1__000010/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:780efabf454c0773141ad15bb7a8dfd2eebaee4071b4411a6acddffee59aaa2b +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000077_test__val_desert__1__000010/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000077_test__val_desert__1__000010/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..dccdc27028d3cee8502d3af27f844f82907ee91e --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000077_test__val_desert__1__000010/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "77", + "global_test_index": "77", + "dataset_index": "26", + "relative_path": "test/val_desert/1/000010.mp4", + "clip_id": "test/val_desert/1/000010", + "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": 42000077, + "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/1/000010_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": 0.012677962891757488, + "psnr": 19.395605087280273, + "lpips": 0.2991322875022888 + }, + "frames_100_200": { + "mse": 0.01670403778553009, + "psnr": 17.8996639251709, + "lpips": 0.375832736492157 + }, + "frames_200_300": { + "mse": 0.02063838392496109, + "psnr": 16.92059898376465, + "lpips": 0.3673126697540283 + }, + "frames_300_400": { + "mse": 0.013535800389945507, + "psnr": 19.32175064086914, + "lpips": 0.37497687339782715 + }, + "frames_400_500": { + "mse": 0.00842384435236454, + "psnr": 20.925521850585938, + "lpips": 0.2951222360134125 + }, + "overall": { + "mse": 0.014396006241440773, + "psnr": 18.892627716064453, + "lpips": 0.34247535467147827 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000078_test__val_desert__0__000029/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000078_test__val_desert__0__000029/case.json new file mode 100644 index 0000000000000000000000000000000000000000..0f90efffb6ad1c752769ae2eb4b562202177750d --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000078_test__val_desert__0__000029/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "78", + "global_test_index": "78", + "dataset_index": "77", + "relative_path": "test/val_desert/0/000029.mp4", + "clip_id": "test/val_desert/0/000029", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_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": 42000078, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 224.98950571205933, + "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_seed42_job85134/cases/g000078_test__val_desert__0__000029/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000078_test__val_desert__0__000029/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..0a018d2e2b5eb72a5f1885cb240cb3c51ab6d19b --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000078_test__val_desert__0__000029/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dcb206a9a40b9eff579f96c4b8c9a7eb427e88c8a9c0701e1ec3a8a66611ebab +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000078_test__val_desert__0__000029/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000078_test__val_desert__0__000029/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..6e70c33fd120ff3cdb8f774d4b933601f320c2b6 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000078_test__val_desert__0__000029/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "78", + "global_test_index": "78", + "dataset_index": "77", + "relative_path": "test/val_desert/0/000029.mp4", + "clip_id": "test/val_desert/0/000029", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_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": 42000078, + "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/0/000029_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": 0.006450569722801447, + "psnr": 22.039918899536133, + "lpips": 0.21885833144187927 + }, + "frames_100_200": { + "mse": 0.006781248841434717, + "psnr": 22.090166091918945, + "lpips": 0.19099944829940796 + }, + "frames_200_300": { + "mse": 0.007121823262423277, + "psnr": 21.625703811645508, + "lpips": 0.24410806596279144 + }, + "frames_300_400": { + "mse": 0.007612924557179213, + "psnr": 21.25684356689453, + "lpips": 0.2527366876602173 + }, + "frames_400_500": { + "mse": 0.0038614345248788595, + "psnr": 24.744810104370117, + "lpips": 0.08614877611398697 + }, + "overall": { + "mse": 0.006365600042045116, + "psnr": 22.35148811340332, + "lpips": 0.19857026636600494 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000079_test__val_iceplains__1__000019/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000079_test__val_iceplains__1__000019/case.json new file mode 100644 index 0000000000000000000000000000000000000000..8884fda730b0131d551ac4c77e490deeeaa3cb7f --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000079_test__val_iceplains__1__000019/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "79", + "global_test_index": "79", + "dataset_index": "294", + "relative_path": "test/val_iceplains/1/000019.mp4", + "clip_id": "test/val_iceplains/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": 42000079, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.71184895501938, + "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_seed42_job85134/cases/g000079_test__val_iceplains__1__000019/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000079_test__val_iceplains__1__000019/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..49d7862d1fbe629d7ba756d6607c41d6c2f9f81f --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000079_test__val_iceplains__1__000019/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7c37181b5105b513b867d906ee3bc5c73d00b07bc229f1c683de2dd5d51bf07b +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000079_test__val_iceplains__1__000019/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000079_test__val_iceplains__1__000019/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..476d4903c989e8930e0cff261a67ab06669e6502 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000079_test__val_iceplains__1__000019/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "79", + "global_test_index": "79", + "dataset_index": "294", + "relative_path": "test/val_iceplains/1/000019.mp4", + "clip_id": "test/val_iceplains/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": 42000079, + "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/1/000019_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": 0.06765586137771606, + "psnr": 13.243976593017578, + "lpips": 0.1256302446126938 + }, + "frames_100_200": { + "mse": 0.08643408864736557, + "psnr": 10.687822341918945, + "lpips": 0.1525007039308548 + }, + "frames_200_300": { + "mse": 0.07716819643974304, + "psnr": 11.131818771362305, + "lpips": 0.1367826908826828 + }, + "frames_300_400": { + "mse": 0.08422358334064484, + "psnr": 10.787413597106934, + "lpips": 0.14840561151504517 + }, + "frames_400_500": { + "mse": 0.07607077807188034, + "psnr": 11.197160720825195, + "lpips": 0.13318023085594177 + }, + "overall": { + "mse": 0.07831050455570221, + "psnr": 11.409638404846191, + "lpips": 0.1392998993396759 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000080_test__val_desert__1__000014/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000080_test__val_desert__1__000014/case.json new file mode 100644 index 0000000000000000000000000000000000000000..8e7c85de6bdc94373ae8739ffb46b98e0910de2b --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000080_test__val_desert__1__000014/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "80", + "global_test_index": "80", + "dataset_index": "34", + "relative_path": "test/val_desert/1/000014.mp4", + "clip_id": "test/val_desert/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": 42000080, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 224.43756491399836, + "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_seed42_job85134/cases/g000080_test__val_desert__1__000014/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000080_test__val_desert__1__000014/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..38d33429aaecfef23293d2bb50e32566b18ce8be --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000080_test__val_desert__1__000014/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b6e0e8ba1c1bbddcfdf630fbc7e3d558b07e2790743a001a17f53436bdd9ba64 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000080_test__val_desert__1__000014/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000080_test__val_desert__1__000014/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..f08a1c69ee9e325d244a89cc39c7f03353fc55de --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000080_test__val_desert__1__000014/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "80", + "global_test_index": "80", + "dataset_index": "34", + "relative_path": "test/val_desert/1/000014.mp4", + "clip_id": "test/val_desert/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": 42000080, + "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/1/000014_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": 0.0016344758914783597, + "psnr": 30.338045120239258, + "lpips": 0.014737879857420921 + }, + "frames_100_200": { + "mse": 0.0007928002742119133, + "psnr": 31.018526077270508, + "lpips": 0.010827239602804184 + }, + "frames_200_300": { + "mse": 0.0016426655929535627, + "psnr": 30.25250816345215, + "lpips": 0.015416990965604782 + }, + "frames_300_400": { + "mse": 0.002514862921088934, + "psnr": 29.235700607299805, + "lpips": 0.021813778206706047 + }, + "frames_400_500": { + "mse": 0.001912656589411199, + "psnr": 28.140121459960938, + "lpips": 0.029486406594514847 + }, + "overall": { + "mse": 0.001699492335319519, + "psnr": 29.796977996826172, + "lpips": 0.018456459045410156 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000081_test__val_savanna__1__000029/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000081_test__val_savanna__1__000029/case.json new file mode 100644 index 0000000000000000000000000000000000000000..b3ec4276fe8a471989817d4110cb834221cadadb --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000081_test__val_savanna__1__000029/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "81", + "global_test_index": "81", + "dataset_index": "188", + "relative_path": "test/val_savanna/1/000029.mp4", + "clip_id": "test/val_savanna/1/000029", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_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": 42000081, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.35637757205404, + "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_seed42_job85134/cases/g000081_test__val_savanna__1__000029/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000081_test__val_savanna__1__000029/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..a41cf96c39c11ba6f4272e6b6440629e73e0ae7b --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000081_test__val_savanna__1__000029/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:79b81e709685bffa0548d861c300b39d11f6b9c00af9693c5566cce28985ce05 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000081_test__val_savanna__1__000029/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000081_test__val_savanna__1__000029/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..9a450fa2436caa78af21ebb9fc881f6770568c2c --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000081_test__val_savanna__1__000029/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "81", + "global_test_index": "81", + "dataset_index": "188", + "relative_path": "test/val_savanna/1/000029.mp4", + "clip_id": "test/val_savanna/1/000029", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_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": 42000081, + "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/1/000029_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": 0.008519408293068409, + "psnr": 22.640642166137695, + "lpips": 0.05814105644822121 + }, + "frames_100_200": { + "mse": 0.021794486790895462, + "psnr": 17.175334930419922, + "lpips": 0.16780391335487366 + }, + "frames_200_300": { + "mse": 0.04050106927752495, + "psnr": 14.182506561279297, + "lpips": 0.33123061060905457 + }, + "frames_300_400": { + "mse": 0.024534311145544052, + "psnr": 16.23138427734375, + "lpips": 0.3358722925186157 + }, + "frames_400_500": { + "mse": 0.04255126416683197, + "psnr": 14.607583045959473, + "lpips": 0.4656073749065399 + }, + "overall": { + "mse": 0.027580106630921364, + "psnr": 16.967487335205078, + "lpips": 0.27173104882240295 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000082_test__val_savanna__0__000017/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000082_test__val_savanna__0__000017/case.json new file mode 100644 index 0000000000000000000000000000000000000000..8f8a9924aba6854e82f4845aa926b5bec9bb5ff6 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000082_test__val_savanna__0__000017/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "82", + "global_test_index": "82", + "dataset_index": "170", + "relative_path": "test/val_savanna/0/000017.mp4", + "clip_id": "test/val_savanna/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": 42000082, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.3789222909836, + "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_seed42_job85134/cases/g000082_test__val_savanna__0__000017/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000082_test__val_savanna__0__000017/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..beb3430c69f072e832fe48b938917f8f3a84906c --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000082_test__val_savanna__0__000017/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:68727ea5457037a873341091764ea01ff61ec70d8027402a4555d4a289fec826 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000082_test__val_savanna__0__000017/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000082_test__val_savanna__0__000017/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..b4f42d013c863eb39c26d3162f983ad5a27d33a4 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000082_test__val_savanna__0__000017/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "82", + "global_test_index": "82", + "dataset_index": "170", + "relative_path": "test/val_savanna/0/000017.mp4", + "clip_id": "test/val_savanna/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": 42000082, + "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/0/000017_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": 0.012054363265633583, + "psnr": 19.486520767211914, + "lpips": 0.2790786623954773 + }, + "frames_100_200": { + "mse": 0.034419622272253036, + "psnr": 14.738089561462402, + "lpips": 0.34023234248161316 + }, + "frames_200_300": { + "mse": 0.04989715665578842, + "psnr": 13.180026054382324, + "lpips": 0.41504615545272827 + }, + "frames_300_400": { + "mse": 0.06496300548315048, + "psnr": 12.247797012329102, + "lpips": 0.3596496880054474 + }, + "frames_400_500": { + "mse": 0.07712022960186005, + "psnr": 11.865474700927734, + "lpips": 0.46575626730918884 + }, + "overall": { + "mse": 0.047690872102975845, + "psnr": 14.303583145141602, + "lpips": 0.371952623128891 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000083_test__val_plains__2__000017/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000083_test__val_plains__2__000017/case.json new file mode 100644 index 0000000000000000000000000000000000000000..4385d2cc2caadc19281d47d1622542d3731379db --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000083_test__val_plains__2__000017/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "83", + "global_test_index": "83", + "dataset_index": "415", + "relative_path": "test/val_plains/2/000017.mp4", + "clip_id": "test/val_plains/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": 42000083, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.10897652700078, + "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_seed42_job85134/cases/g000083_test__val_plains__2__000017/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000083_test__val_plains__2__000017/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..667de8846c5f8052d9bf8944411a907c96ed923f --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000083_test__val_plains__2__000017/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a53b5f42aee45e05cbe6610bc420110d16aaad79ff2aa3b1161e0bdb4cd48746 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000083_test__val_plains__2__000017/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000083_test__val_plains__2__000017/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..c09a1644b9dd13050e0e3e44b0fb9241a86d24fe --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000083_test__val_plains__2__000017/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "83", + "global_test_index": "83", + "dataset_index": "415", + "relative_path": "test/val_plains/2/000017.mp4", + "clip_id": "test/val_plains/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": 42000083, + "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/2/000017_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": 0.006952962838113308, + "psnr": 22.07476806640625, + "lpips": 0.11937867850065231 + }, + "frames_100_200": { + "mse": 0.022882215678691864, + "psnr": 17.546424865722656, + "lpips": 0.31689921021461487 + }, + "frames_200_300": { + "mse": 0.022935042157769203, + "psnr": 17.054492950439453, + "lpips": 0.32420384883880615 + }, + "frames_300_400": { + "mse": 0.01728488877415657, + "psnr": 17.81946563720703, + "lpips": 0.31152790784835815 + }, + "frames_400_500": { + "mse": 0.013039127923548222, + "psnr": 19.060495376586914, + "lpips": 0.3390658497810364 + }, + "overall": { + "mse": 0.016618845984339714, + "psnr": 18.711130142211914, + "lpips": 0.28221508860588074 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000084_test__val_savanna__0__000006/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000084_test__val_savanna__0__000006/case.json new file mode 100644 index 0000000000000000000000000000000000000000..229f1c6054f6eb31e8dba517445382518c4cd3cb --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000084_test__val_savanna__0__000006/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "84", + "global_test_index": "84", + "dataset_index": "145", + "relative_path": "test/val_savanna/0/000006.mp4", + "clip_id": "test/val_savanna/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": 42000084, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 224.64585186797194, + "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_seed42_job85134/cases/g000084_test__val_savanna__0__000006/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000084_test__val_savanna__0__000006/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..ab188cab650b1ea95fca3eb21a4adfbafcd26069 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000084_test__val_savanna__0__000006/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:16c9b4b938a45ecf5cb1f0ababdc667c870741151ec559e242e9e197849898f0 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000084_test__val_savanna__0__000006/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000084_test__val_savanna__0__000006/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..ee61483c05213972894dc6b0df6678e6c9d28d79 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000084_test__val_savanna__0__000006/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "84", + "global_test_index": "84", + "dataset_index": "145", + "relative_path": "test/val_savanna/0/000006.mp4", + "clip_id": "test/val_savanna/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": 42000084, + "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/0/000006_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": 0.017755519598722458, + "psnr": 18.56708526611328, + "lpips": 0.12994161248207092 + }, + "frames_100_200": { + "mse": 0.02105613797903061, + "psnr": 17.045766830444336, + "lpips": 0.23321153223514557 + }, + "frames_200_300": { + "mse": 0.017575835809111595, + "psnr": 17.9531307220459, + "lpips": 0.4522656202316284 + }, + "frames_300_400": { + "mse": 0.006743900012224913, + "psnr": 21.839197158813477, + "lpips": 0.41311293840408325 + }, + "frames_400_500": { + "mse": 0.005872038193047047, + "psnr": 22.393089294433594, + "lpips": 0.4014129936695099 + }, + "overall": { + "mse": 0.013800686225295067, + "psnr": 19.559654235839844, + "lpips": 0.3259889483451843 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000085_test__val_savanna__1__000002/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000085_test__val_savanna__1__000002/case.json new file mode 100644 index 0000000000000000000000000000000000000000..fb0d3d534719549c59c864972d5ca7c2f8038eec --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000085_test__val_savanna__1__000002/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "85", + "global_test_index": "85", + "dataset_index": "133", + "relative_path": "test/val_savanna/1/000002.mp4", + "clip_id": "test/val_savanna/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": 42000085, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 224.94058923097327, + "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_seed42_job85134/cases/g000085_test__val_savanna__1__000002/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000085_test__val_savanna__1__000002/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..c6fdc0da4fea8922193821afc8ddd7d4c6f3d7b7 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000085_test__val_savanna__1__000002/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:898da334ae85cf8df8a8c721db3da5bf53b296cfc03526f44a0aaacbee38dc7c +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000085_test__val_savanna__1__000002/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000085_test__val_savanna__1__000002/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..0e1fda08e89442f735bf72ef8516b26cc81199fe --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000085_test__val_savanna__1__000002/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "85", + "global_test_index": "85", + "dataset_index": "133", + "relative_path": "test/val_savanna/1/000002.mp4", + "clip_id": "test/val_savanna/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": 42000085, + "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/1/000002_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": 0.011285670101642609, + "psnr": 19.653379440307617, + "lpips": 0.2619589567184448 + }, + "frames_100_200": { + "mse": 0.02196894958615303, + "psnr": 16.801177978515625, + "lpips": 0.29938217997550964 + }, + "frames_200_300": { + "mse": 0.016156572848558426, + "psnr": 18.344593048095703, + "lpips": 0.3523021936416626 + }, + "frames_300_400": { + "mse": 0.012617328204214573, + "psnr": 19.169965744018555, + "lpips": 0.37788963317871094 + }, + "frames_400_500": { + "mse": 0.020221270620822906, + "psnr": 17.951553344726562, + "lpips": 0.39422357082366943 + }, + "overall": { + "mse": 0.016449958086013794, + "psnr": 18.384132385253906, + "lpips": 0.33715134859085083 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000086_test__val_iceplains__2__000039/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000086_test__val_iceplains__2__000039/case.json new file mode 100644 index 0000000000000000000000000000000000000000..74b3f807136a85db4249082875068860d9cb189d --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000086_test__val_iceplains__2__000039/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "86", + "global_test_index": "86", + "dataset_index": "347", + "relative_path": "test/val_iceplains/2/000039.mp4", + "clip_id": "test/val_iceplains/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": 42000086, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 224.91139730100986, + "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_seed42_job85134/cases/g000086_test__val_iceplains__2__000039/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000086_test__val_iceplains__2__000039/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..eaf93d595aca32487d312cdb3cf19de158a83e96 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000086_test__val_iceplains__2__000039/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ce01718c58bc9a746267d55ea6c8fc1451bd34f9ca4cababd69a41c8f954314b +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000086_test__val_iceplains__2__000039/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000086_test__val_iceplains__2__000039/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..2ed3a182a742bee4f5ad0a98778c39f325a94d29 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000086_test__val_iceplains__2__000039/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "86", + "global_test_index": "86", + "dataset_index": "347", + "relative_path": "test/val_iceplains/2/000039.mp4", + "clip_id": "test/val_iceplains/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": 42000086, + "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/2/000039_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": 0.02075340785086155, + "psnr": 17.151208877563477, + "lpips": 0.19061937928199768 + }, + "frames_100_200": { + "mse": 0.014682957902550697, + "psnr": 19.417430877685547, + "lpips": 0.2166585475206375 + }, + "frames_200_300": { + "mse": 0.03729237616062164, + "psnr": 17.206836700439453, + "lpips": 0.34138745069503784 + }, + "frames_300_400": { + "mse": 0.18325267732143402, + "psnr": 7.589336395263672, + "lpips": 0.5897568464279175 + }, + "frames_400_500": { + "mse": 0.21861088275909424, + "psnr": 6.84079122543335, + "lpips": 0.6720837354660034 + }, + "overall": { + "mse": 0.09491845965385437, + "psnr": 13.641119956970215, + "lpips": 0.40210118889808655 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000087_test__val_savanna__0__000038/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000087_test__val_savanna__0__000038/case.json new file mode 100644 index 0000000000000000000000000000000000000000..0ad2a9737a57a785ce996f7b42e68954046764ef --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000087_test__val_savanna__0__000038/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "87", + "global_test_index": "87", + "dataset_index": "209", + "relative_path": "test/val_savanna/0/000038.mp4", + "clip_id": "test/val_savanna/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": 42000087, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 223.59352648293134, + "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_seed42_job85134/cases/g000087_test__val_savanna__0__000038/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000087_test__val_savanna__0__000038/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..165ae83414e5cd499be0445a1b64051730ed4da1 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000087_test__val_savanna__0__000038/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4708873f8488e2b3f2d4d25566093540d724f7053596fd59f63e03e75d26fcd2 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000087_test__val_savanna__0__000038/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000087_test__val_savanna__0__000038/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..f8b19a3805068570dc1d84b826bf398eb35d86e1 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000087_test__val_savanna__0__000038/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "87", + "global_test_index": "87", + "dataset_index": "209", + "relative_path": "test/val_savanna/0/000038.mp4", + "clip_id": "test/val_savanna/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": 42000087, + "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/0/000038_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": 0.015088949352502823, + "psnr": 18.89339256286621, + "lpips": 0.26669082045555115 + }, + "frames_100_200": { + "mse": 0.03187736123800278, + "psnr": 16.236005783081055, + "lpips": 0.36731529235839844 + }, + "frames_200_300": { + "mse": 0.03693845868110657, + "psnr": 15.566036224365234, + "lpips": 0.41454434394836426 + }, + "frames_300_400": { + "mse": 0.007255536504089832, + "psnr": 21.717182159423828, + "lpips": 0.36932212114334106 + }, + "frames_400_500": { + "mse": 0.013100729323923588, + "psnr": 19.635723114013672, + "lpips": 0.42514708638191223 + }, + "overall": { + "mse": 0.020852206274867058, + "psnr": 18.40966796875, + "lpips": 0.3686039447784424 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000088_test__val_plains__1__000013/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000088_test__val_plains__1__000013/case.json new file mode 100644 index 0000000000000000000000000000000000000000..af739b4d43c5f9006af996b590be428313db3c98 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000088_test__val_plains__1__000013/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "88", + "global_test_index": "88", + "dataset_index": "404", + "relative_path": "test/val_plains/1/000013.mp4", + "clip_id": "test/val_plains/1/000013", + "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": 42000088, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 223.40496907092165, + "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_seed42_job85134/cases/g000088_test__val_plains__1__000013/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000088_test__val_plains__1__000013/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..e17dd40b5aca91a02591b992113e86272824e961 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000088_test__val_plains__1__000013/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dc7c465cb640893f05c3190ff351901d36745f3e3df0e925901ff524e2a06b4a +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000088_test__val_plains__1__000013/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000088_test__val_plains__1__000013/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..183a1de5e9c890d1f4ae0837b17a02b6e031aaa4 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000088_test__val_plains__1__000013/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "88", + "global_test_index": "88", + "dataset_index": "404", + "relative_path": "test/val_plains/1/000013.mp4", + "clip_id": "test/val_plains/1/000013", + "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": 42000088, + "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/1/000013_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": 0.001979007152840495, + "psnr": 27.465681076049805, + "lpips": 0.16871009767055511 + }, + "frames_100_200": { + "mse": 0.0058663384988904, + "psnr": 23.0067138671875, + "lpips": 0.25231167674064636 + }, + "frames_200_300": { + "mse": 0.012074603699147701, + "psnr": 19.828468322753906, + "lpips": 0.3021427392959595 + }, + "frames_300_400": { + "mse": 0.0562514066696167, + "psnr": 12.542461395263672, + "lpips": 0.2533787190914154 + }, + "frames_400_500": { + "mse": 0.061275020241737366, + "psnr": 12.147039413452148, + "lpips": 0.32059016823768616 + }, + "overall": { + "mse": 0.027489272877573967, + "psnr": 18.998071670532227, + "lpips": 0.25942668318748474 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000089_test__val_savanna__2__000038/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000089_test__val_savanna__2__000038/case.json new file mode 100644 index 0000000000000000000000000000000000000000..06f81c92e60628f31b1d6a786a73dde0a830422c --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000089_test__val_savanna__2__000038/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "89", + "global_test_index": "89", + "dataset_index": "211", + "relative_path": "test/val_savanna/2/000038.mp4", + "clip_id": "test/val_savanna/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": 42000089, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.1836247870233, + "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_seed42_job85134/cases/g000089_test__val_savanna__2__000038/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000089_test__val_savanna__2__000038/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..36346aa76abb140543238992d4946227a661b25e --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000089_test__val_savanna__2__000038/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:68df6b17746420d5ae31bc6243f53e02a59f3ce50237424ad8665f64d7926de7 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000089_test__val_savanna__2__000038/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000089_test__val_savanna__2__000038/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..c1c929d0585adaaaa7689d446671389f23637a86 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000089_test__val_savanna__2__000038/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "89", + "global_test_index": "89", + "dataset_index": "211", + "relative_path": "test/val_savanna/2/000038.mp4", + "clip_id": "test/val_savanna/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": 42000089, + "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/2/000038_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": 0.015563503839075565, + "psnr": 18.328170776367188, + "lpips": 0.36856383085250854 + }, + "frames_100_200": { + "mse": 0.005765731446444988, + "psnr": 22.473243713378906, + "lpips": 0.36543339490890503 + }, + "frames_200_300": { + "mse": 0.006060788407921791, + "psnr": 22.259056091308594, + "lpips": 0.3785349130630493 + }, + "frames_300_400": { + "mse": 0.00762468995526433, + "psnr": 21.29523468017578, + "lpips": 0.4031403362751007 + }, + "frames_400_500": { + "mse": 0.00978897511959076, + "psnr": 20.2990665435791, + "lpips": 0.4536948800086975 + }, + "overall": { + "mse": 0.00896073691546917, + "psnr": 20.93095588684082, + "lpips": 0.3938734829425812 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000090_test__val_plains__2__000044/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000090_test__val_plains__2__000044/case.json new file mode 100644 index 0000000000000000000000000000000000000000..116e49ecaf45eec3f81b38403e18c198bf19c225 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000090_test__val_plains__2__000044/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "90", + "global_test_index": "90", + "dataset_index": "464", + "relative_path": "test/val_plains/2/000044.mp4", + "clip_id": "test/val_plains/2/000044", + "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": 42000090, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 224.67127126909327, + "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_seed42_job85134/cases/g000090_test__val_plains__2__000044/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000090_test__val_plains__2__000044/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..b793c58c3d40fe8873d598ee8c5b15151957c742 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000090_test__val_plains__2__000044/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1d165058f701ccd2e8ce0b5b6104afb943c2d30bcb10ef004ce15fea81350e10 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000090_test__val_plains__2__000044/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000090_test__val_plains__2__000044/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..c5dd6094492a8959627e467bf0610be34428d2e8 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000090_test__val_plains__2__000044/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "90", + "global_test_index": "90", + "dataset_index": "464", + "relative_path": "test/val_plains/2/000044.mp4", + "clip_id": "test/val_plains/2/000044", + "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": 42000090, + "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/2/000044_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": 0.009854784235358238, + "psnr": 21.697132110595703, + "lpips": 0.30292484164237976 + }, + "frames_100_200": { + "mse": 0.032724425196647644, + "psnr": 15.166298866271973, + "lpips": 0.42703452706336975 + }, + "frames_200_300": { + "mse": 0.05092819035053253, + "psnr": 13.726238250732422, + "lpips": 0.3965069651603699 + }, + "frames_300_400": { + "mse": 0.03217763453722, + "psnr": 15.189897537231445, + "lpips": 0.3740811049938202 + }, + "frames_400_500": { + "mse": 0.018094651401042938, + "psnr": 17.873842239379883, + "lpips": 0.38506314158439636 + }, + "overall": { + "mse": 0.02875593677163124, + "psnr": 16.730682373046875, + "lpips": 0.37712210416793823 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000091_test__val_savanna__1__000014/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000091_test__val_savanna__1__000014/case.json new file mode 100644 index 0000000000000000000000000000000000000000..5c043040268ec9baba374d798cc9626993cba959 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000091_test__val_savanna__1__000014/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "91", + "global_test_index": "91", + "dataset_index": "162", + "relative_path": "test/val_savanna/1/000014.mp4", + "clip_id": "test/val_savanna/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": 42000091, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 224.6168820100138, + "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_seed42_job85134/cases/g000091_test__val_savanna__1__000014/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000091_test__val_savanna__1__000014/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..9a6dd02167aa6421c8060793cc4de5e66fb8ecef --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000091_test__val_savanna__1__000014/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a0582fc2ef3ee207894bf1f2f0015f2336d5b1e1a55bf2ccb38f341fe0ae8d77 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000091_test__val_savanna__1__000014/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000091_test__val_savanna__1__000014/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..7b0445277b51f0e8ff634ba17663a5328e80924d --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000091_test__val_savanna__1__000014/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "91", + "global_test_index": "91", + "dataset_index": "162", + "relative_path": "test/val_savanna/1/000014.mp4", + "clip_id": "test/val_savanna/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": 42000091, + "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/1/000014_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": 0.02314971387386322, + "psnr": 17.313108444213867, + "lpips": 0.3340839743614197 + }, + "frames_100_200": { + "mse": 0.07382805645465851, + "psnr": 12.138628959655762, + "lpips": 0.5771018862724304 + }, + "frames_200_300": { + "mse": 0.010941821150481701, + "psnr": 20.837522506713867, + "lpips": 0.415218710899353 + }, + "frames_300_400": { + "mse": 0.0055582569912076, + "psnr": 22.63018226623535, + "lpips": 0.3847825229167938 + }, + "frames_400_500": { + "mse": 0.008172615431249142, + "psnr": 20.992660522460938, + "lpips": 0.443231463432312 + }, + "overall": { + "mse": 0.02433009259402752, + "psnr": 18.782421112060547, + "lpips": 0.4308837354183197 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000092_test__val_plains__2__000019/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000092_test__val_plains__2__000019/case.json new file mode 100644 index 0000000000000000000000000000000000000000..764e35ec6d56189f17fab9797e2292e60f48bcb5 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000092_test__val_plains__2__000019/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "92", + "global_test_index": "92", + "dataset_index": "419", + "relative_path": "test/val_plains/2/000019.mp4", + "clip_id": "test/val_plains/2/000019", + "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": 42000092, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 224.13612580299377, + "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_seed42_job85134/cases/g000092_test__val_plains__2__000019/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000092_test__val_plains__2__000019/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..7146e0db2cef294fb45f1c2f90eb122b99f6a33f --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000092_test__val_plains__2__000019/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a405e949a84c6410d68a1173b29892ec84f725b13a7c994d1e4d7f6bc5f75087 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000092_test__val_plains__2__000019/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000092_test__val_plains__2__000019/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..efe1489a96b33580ed28bcc14a9d43c978dc9e23 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000092_test__val_plains__2__000019/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "92", + "global_test_index": "92", + "dataset_index": "419", + "relative_path": "test/val_plains/2/000019.mp4", + "clip_id": "test/val_plains/2/000019", + "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": 42000092, + "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/2/000019_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": 0.012321366928517818, + "psnr": 19.893596649169922, + "lpips": 0.25994691252708435 + }, + "frames_100_200": { + "mse": 0.010026111267507076, + "psnr": 20.5347900390625, + "lpips": 0.3756827116012573 + }, + "frames_200_300": { + "mse": 0.01013142429292202, + "psnr": 20.443151473999023, + "lpips": 0.3708946108818054 + }, + "frames_300_400": { + "mse": 0.00863606482744217, + "psnr": 21.373281478881836, + "lpips": 0.15324433147907257 + }, + "frames_400_500": { + "mse": 0.0031472037080675364, + "psnr": 25.679616928100586, + "lpips": 0.03747815266251564 + }, + "overall": { + "mse": 0.00885243434458971, + "psnr": 21.584888458251953, + "lpips": 0.2394493669271469 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000093_test__val_desert__3__000049/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000093_test__val_desert__3__000049/case.json new file mode 100644 index 0000000000000000000000000000000000000000..71530286d083275332bf7187b0f4ce7abf52d040 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000093_test__val_desert__3__000049/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "93", + "global_test_index": "93", + "dataset_index": "127", + "relative_path": "test/val_desert/3/000049.mp4", + "clip_id": "test/val_desert/3/000049", + "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": 42000093, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 224.2161980040837, + "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_seed42_job85134/cases/g000093_test__val_desert__3__000049/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000093_test__val_desert__3__000049/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..944df2104e54a3faa84eaad151a0903e8ef33a23 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000093_test__val_desert__3__000049/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:12c78b7402f50fd6d30989dacac86c5176cdebc9176850fa2238b3468318f134 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000093_test__val_desert__3__000049/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000093_test__val_desert__3__000049/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..f6bf62c6f8054cb695098e7a06eb598e6376b60a --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000093_test__val_desert__3__000049/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "93", + "global_test_index": "93", + "dataset_index": "127", + "relative_path": "test/val_desert/3/000049.mp4", + "clip_id": "test/val_desert/3/000049", + "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": 42000093, + "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/3/000049_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": 0.005808006972074509, + "psnr": 23.5024356842041, + "lpips": 0.13743585348129272 + }, + "frames_100_200": { + "mse": 0.016257118433713913, + "psnr": 19.560550689697266, + "lpips": 0.2375103235244751 + }, + "frames_200_300": { + "mse": 0.01671038381755352, + "psnr": 18.94465446472168, + "lpips": 0.2576060891151428 + }, + "frames_300_400": { + "mse": 0.01879621483385563, + "psnr": 18.297801971435547, + "lpips": 0.3120792508125305 + }, + "frames_400_500": { + "mse": 0.016487326472997665, + "psnr": 18.208364486694336, + "lpips": 0.27908065915107727 + }, + "overall": { + "mse": 0.014811811037361622, + "psnr": 19.702762603759766, + "lpips": 0.24474243819713593 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000094_test__val_iceplains__2__000048/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000094_test__val_iceplains__2__000048/case.json new file mode 100644 index 0000000000000000000000000000000000000000..3ec5200e2fbf22eaaf6a6d27dd1c4da37a50122d --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000094_test__val_iceplains__2__000048/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "94", + "global_test_index": "94", + "dataset_index": "370", + "relative_path": "test/val_iceplains/2/000048.mp4", + "clip_id": "test/val_iceplains/2/000048", + "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": 42000094, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.19615069800057, + "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_seed42_job85134/cases/g000094_test__val_iceplains__2__000048/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000094_test__val_iceplains__2__000048/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..11b827a69c3fb0d2590d23705d9b2d126f403828 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000094_test__val_iceplains__2__000048/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:acb9cdd6df07ea4065d1ad25331087ecd09a5c0b433e5934ea252e9e91713d7f +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000094_test__val_iceplains__2__000048/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000094_test__val_iceplains__2__000048/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..91279991a122349da4a9ca162b6475ff59c5c27f --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000094_test__val_iceplains__2__000048/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "94", + "global_test_index": "94", + "dataset_index": "370", + "relative_path": "test/val_iceplains/2/000048.mp4", + "clip_id": "test/val_iceplains/2/000048", + "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": 42000094, + "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/2/000048_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": 0.006527842953801155, + "psnr": 22.012739181518555, + "lpips": 0.08346416801214218 + }, + "frames_100_200": { + "mse": 0.01270291768014431, + "psnr": 19.371593475341797, + "lpips": 0.09923094511032104 + }, + "frames_200_300": { + "mse": 0.008414355106651783, + "psnr": 21.78388023376465, + "lpips": 0.059944283217191696 + }, + "frames_300_400": { + "mse": 0.02148391678929329, + "psnr": 21.02130699157715, + "lpips": 0.10497161746025085 + }, + "frames_400_500": { + "mse": 0.12538494169712067, + "psnr": 9.17703628540039, + "lpips": 0.5502245426177979 + }, + "overall": { + "mse": 0.034902796149253845, + "psnr": 18.673309326171875, + "lpips": 0.1795671135187149 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000095_test__val_iceplains__2__000018/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000095_test__val_iceplains__2__000018/case.json new file mode 100644 index 0000000000000000000000000000000000000000..7f74b60729c1d0f95352b03e6b910a2621583e66 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000095_test__val_iceplains__2__000018/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "95", + "global_test_index": "95", + "dataset_index": "293", + "relative_path": "test/val_iceplains/2/000018.mp4", + "clip_id": "test/val_iceplains/2/000018", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_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": 42000095, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 224.0368143110536, + "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_seed42_job85134/cases/g000095_test__val_iceplains__2__000018/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000095_test__val_iceplains__2__000018/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..adda6b5a401e67a6ae4edde55a4ac1e6d6d4e0be --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000095_test__val_iceplains__2__000018/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c60e113268cc6e70492b85a62088eefaa8110fb9470e5b7f92641fa7aea41a81 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000095_test__val_iceplains__2__000018/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000095_test__val_iceplains__2__000018/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..3d02a759225046eb08a747dba5258687d43a52b1 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000095_test__val_iceplains__2__000018/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "95", + "global_test_index": "95", + "dataset_index": "293", + "relative_path": "test/val_iceplains/2/000018.mp4", + "clip_id": "test/val_iceplains/2/000018", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_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": 42000095, + "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/2/000018_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": 0.0037903846241533756, + "psnr": 24.240293502807617, + "lpips": 0.17606912553310394 + }, + "frames_100_200": { + "mse": 0.007717910222709179, + "psnr": 22.431800842285156, + "lpips": 0.18211966753005981 + }, + "frames_200_300": { + "mse": 0.012126042507588863, + "psnr": 19.961694717407227, + "lpips": 0.22617489099502563 + }, + "frames_300_400": { + "mse": 0.013583356514573097, + "psnr": 19.793384552001953, + "lpips": 0.23104484379291534 + }, + "frames_400_500": { + "mse": 0.021284740418195724, + "psnr": 18.451427459716797, + "lpips": 0.16046279668807983 + }, + "overall": { + "mse": 0.01170048676431179, + "psnr": 20.97572135925293, + "lpips": 0.19517426192760468 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000096_test__val_iceplains__1__000003/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000096_test__val_iceplains__1__000003/case.json new file mode 100644 index 0000000000000000000000000000000000000000..f7b64818de17b5faefcab4ea5a374d52a776d265 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000096_test__val_iceplains__1__000003/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "96", + "global_test_index": "96", + "dataset_index": "251", + "relative_path": "test/val_iceplains/1/000003.mp4", + "clip_id": "test/val_iceplains/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": 42000096, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 224.20748409000225, + "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_seed42_job85134/cases/g000096_test__val_iceplains__1__000003/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000096_test__val_iceplains__1__000003/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..f760d315df14e71ed2711b6ab5b19fcaa30e2a54 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000096_test__val_iceplains__1__000003/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e8b50ee65c0a22d020cf6ebb7349afd3268055c0b30a87d8f6038c64e8617d7a +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000096_test__val_iceplains__1__000003/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000096_test__val_iceplains__1__000003/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..9e3ce987a8b4d18f5a71568bd7cc7e6d52d21fb3 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000096_test__val_iceplains__1__000003/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "96", + "global_test_index": "96", + "dataset_index": "251", + "relative_path": "test/val_iceplains/1/000003.mp4", + "clip_id": "test/val_iceplains/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": 42000096, + "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/1/000003_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": 0.006837934255599976, + "psnr": 23.857084274291992, + "lpips": 0.056156180799007416 + }, + "frames_100_200": { + "mse": 0.00524763111025095, + "psnr": 24.820171356201172, + "lpips": 0.044947732239961624 + }, + "frames_200_300": { + "mse": 0.004913405980914831, + "psnr": 25.06283187866211, + "lpips": 0.042589493095874786 + }, + "frames_300_400": { + "mse": 0.0008500135154463351, + "psnr": 30.733633041381836, + "lpips": 0.011915426701307297 + }, + "frames_400_500": { + "mse": 0.0018243644153699279, + "psnr": 28.681333541870117, + "lpips": 0.021000981330871582 + }, + "overall": { + "mse": 0.003934669774025679, + "psnr": 26.631011962890625, + "lpips": 0.03532196208834648 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000097_test__val_plains__1__000031/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000097_test__val_plains__1__000031/case.json new file mode 100644 index 0000000000000000000000000000000000000000..84ab1169771faeb16efe42fd4e0399705d478155 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000097_test__val_plains__1__000031/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "97", + "global_test_index": "97", + "dataset_index": "441", + "relative_path": "test/val_plains/1/000031.mp4", + "clip_id": "test/val_plains/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": 42000097, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 226.19590765610337, + "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_seed42_job85134/cases/g000097_test__val_plains__1__000031/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000097_test__val_plains__1__000031/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..a85601fe529d29d2a118842b9d4f2af2c42fb19e --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000097_test__val_plains__1__000031/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4b9f27266105c0cb4369cf6f201e15ebdac177566e47f751c0b4a2cf7e432274 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000097_test__val_plains__1__000031/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000097_test__val_plains__1__000031/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..942e4d5844af05cfc94bd2d9770084afad99b2dc --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000097_test__val_plains__1__000031/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "97", + "global_test_index": "97", + "dataset_index": "441", + "relative_path": "test/val_plains/1/000031.mp4", + "clip_id": "test/val_plains/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": 42000097, + "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/1/000031_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": 0.009681429713964462, + "psnr": 23.5019474029541, + "lpips": 0.0760703980922699 + }, + "frames_100_200": { + "mse": 0.015562593936920166, + "psnr": 19.23906707763672, + "lpips": 0.11067785322666168 + }, + "frames_200_300": { + "mse": 0.0485471710562706, + "psnr": 13.307043075561523, + "lpips": 0.38715052604675293 + }, + "frames_300_400": { + "mse": 0.06833904981613159, + "psnr": 11.854065895080566, + "lpips": 0.4324353337287903 + }, + "frames_400_500": { + "mse": 0.06893973797559738, + "psnr": 11.730586051940918, + "lpips": 0.5307671427726746 + }, + "overall": { + "mse": 0.04221399873495102, + "psnr": 15.926543235778809, + "lpips": 0.3074202537536621 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000098_test__val_desert__2__000026/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000098_test__val_desert__2__000026/case.json new file mode 100644 index 0000000000000000000000000000000000000000..d783189bb2ef3f511820cc9d125712bb4d58cc4d --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000098_test__val_desert__2__000026/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "98", + "global_test_index": "98", + "dataset_index": "69", + "relative_path": "test/val_desert/2/000026.mp4", + "clip_id": "test/val_desert/2/000026", + "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": 42000098, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.43990384007338, + "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_seed42_job85134/cases/g000098_test__val_desert__2__000026/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000098_test__val_desert__2__000026/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..0e5a8e377877b1db189548c7bc88d94f1cec23cf --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000098_test__val_desert__2__000026/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9bec88a2c7c9716f5f605f609b7c16f4e360b57869d3c8abf3d76800ccfad83a +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000098_test__val_desert__2__000026/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000098_test__val_desert__2__000026/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..9a1ba8fe8212d159d8b6739eebcf9d25cd2fd2f4 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000098_test__val_desert__2__000026/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "98", + "global_test_index": "98", + "dataset_index": "69", + "relative_path": "test/val_desert/2/000026.mp4", + "clip_id": "test/val_desert/2/000026", + "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": 42000098, + "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/2/000026_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": 0.006772654131054878, + "psnr": 21.804445266723633, + "lpips": 0.21439053118228912 + }, + "frames_100_200": { + "mse": 0.008204908110201359, + "psnr": 21.102800369262695, + "lpips": 0.34267762303352356 + }, + "frames_200_300": { + "mse": 0.007740702014416456, + "psnr": 21.327783584594727, + "lpips": 0.3789578676223755 + }, + "frames_300_400": { + "mse": 0.007694942411035299, + "psnr": 21.156936645507812, + "lpips": 0.37082168459892273 + }, + "frames_400_500": { + "mse": 0.006963862106204033, + "psnr": 21.5853328704834, + "lpips": 0.4089972674846649 + }, + "overall": { + "mse": 0.007475413382053375, + "psnr": 21.395458221435547, + "lpips": 0.3431689441204071 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000099_test__val_iceplains__2__000015/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000099_test__val_iceplains__2__000015/case.json new file mode 100644 index 0000000000000000000000000000000000000000..a07cf17d3e724254bb44d457acbdda1a6cabf7a9 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000099_test__val_iceplains__2__000015/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "99", + "global_test_index": "99", + "dataset_index": "286", + "relative_path": "test/val_iceplains/2/000015.mp4", + "clip_id": "test/val_iceplains/2/000015", + "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": 42000099, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.11404380993918, + "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_seed42_job85134/cases/g000099_test__val_iceplains__2__000015/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000099_test__val_iceplains__2__000015/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..fc073d89c7521e3ef45c45100bbc436d292bbad3 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000099_test__val_iceplains__2__000015/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0aa8f955b0fb68e76430bf4238e168739fb311249f4442e7c5dbffd64f589097 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000099_test__val_iceplains__2__000015/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000099_test__val_iceplains__2__000015/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..edd79ac83170f8642f241b79469a2fd3d74887b2 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000099_test__val_iceplains__2__000015/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "99", + "global_test_index": "99", + "dataset_index": "286", + "relative_path": "test/val_iceplains/2/000015.mp4", + "clip_id": "test/val_iceplains/2/000015", + "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": 42000099, + "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/2/000015_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": 0.020865054801106453, + "psnr": 17.503881454467773, + "lpips": 0.2061655968427658 + }, + "frames_100_200": { + "mse": 0.02510390803217888, + "psnr": 17.542644500732422, + "lpips": 0.2778812348842621 + }, + "frames_200_300": { + "mse": 0.04723326861858368, + "psnr": 13.672036170959473, + "lpips": 0.37518078088760376 + }, + "frames_300_400": { + "mse": 0.04576940834522247, + "psnr": 14.488913536071777, + "lpips": 0.3897820711135864 + }, + "frames_400_500": { + "mse": 0.025316011160612106, + "psnr": 17.392793655395508, + "lpips": 0.3479126989841461 + }, + "overall": { + "mse": 0.03285753354430199, + "psnr": 16.120054244995117, + "lpips": 0.31938445568084717 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000100_test__val_plains__0__000029/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000100_test__val_plains__0__000029/case.json new file mode 100644 index 0000000000000000000000000000000000000000..28284d00d51a49a96802ffb42fc73cbb6d6956ab --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000100_test__val_plains__0__000029/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "100", + "global_test_index": "100", + "dataset_index": "438", + "relative_path": "test/val_plains/0/000029.mp4", + "clip_id": "test/val_plains/0/000029", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_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": 42000100, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 224.62832075194456, + "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_seed42_job85134/cases/g000100_test__val_plains__0__000029/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000100_test__val_plains__0__000029/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..b558d72602e1b8062467b43a1d0ee723cb2df553 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000100_test__val_plains__0__000029/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9bfadc1e58eb48c1c6edc64d4e0d36f543022fc4b0d8371d615f20bed32ee253 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000100_test__val_plains__0__000029/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000100_test__val_plains__0__000029/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..fd94a34d69504538c707043523834262fe87639e --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000100_test__val_plains__0__000029/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "100", + "global_test_index": "100", + "dataset_index": "438", + "relative_path": "test/val_plains/0/000029.mp4", + "clip_id": "test/val_plains/0/000029", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_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": 42000100, + "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/0/000029_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": 0.013046170584857464, + "psnr": 20.828197479248047, + "lpips": 0.17720767855644226 + }, + "frames_100_200": { + "mse": 0.012936729937791824, + "psnr": 25.10344886779785, + "lpips": 0.053137537091970444 + }, + "frames_200_300": { + "mse": 0.028541788458824158, + "psnr": 23.256948471069336, + "lpips": 0.08209577202796936 + }, + "frames_300_400": { + "mse": 0.11823341250419617, + "psnr": 9.378386497497559, + "lpips": 0.290657639503479 + }, + "frames_400_500": { + "mse": 0.05700356885790825, + "psnr": 14.556885719299316, + "lpips": 0.3547360301017761 + }, + "overall": { + "mse": 0.04595233127474785, + "psnr": 18.624773025512695, + "lpips": 0.19156692922115326 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000101_test__val_plains__0__000002/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000101_test__val_plains__0__000002/case.json new file mode 100644 index 0000000000000000000000000000000000000000..e9b91ed89adac2715de8f425e7d27a417b5b141e --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000101_test__val_plains__0__000002/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "101", + "global_test_index": "101", + "dataset_index": "376", + "relative_path": "test/val_plains/0/000002.mp4", + "clip_id": "test/val_plains/0/000002", + "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": 42000101, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 226.50797669799067, + "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_seed42_job85134/cases/g000101_test__val_plains__0__000002/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000101_test__val_plains__0__000002/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..a870094b9f1e6af526739341c1f19c83f35b3f7f --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000101_test__val_plains__0__000002/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0a88f44314bf20dbf3ba18d2abc26d6c8283456755f89c786a958610d23c3d48 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000101_test__val_plains__0__000002/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000101_test__val_plains__0__000002/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..06da2a4fc480e25e1af1432da67b8250c6125177 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000101_test__val_plains__0__000002/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "101", + "global_test_index": "101", + "dataset_index": "376", + "relative_path": "test/val_plains/0/000002.mp4", + "clip_id": "test/val_plains/0/000002", + "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": 42000101, + "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/0/000002_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": 0.005571022629737854, + "psnr": 22.673851013183594, + "lpips": 0.17147918045520782 + }, + "frames_100_200": { + "mse": 0.0064170099794864655, + "psnr": 22.065214157104492, + "lpips": 0.235605850815773 + }, + "frames_200_300": { + "mse": 0.007795399054884911, + "psnr": 21.223772048950195, + "lpips": 0.294155478477478 + }, + "frames_300_400": { + "mse": 0.013274528086185455, + "psnr": 19.822895050048828, + "lpips": 0.3103310465812683 + }, + "frames_400_500": { + "mse": 0.013261131942272186, + "psnr": 19.810155868530273, + "lpips": 0.2501787841320038 + }, + "overall": { + "mse": 0.009263819083571434, + "psnr": 21.119176864624023, + "lpips": 0.2523500919342041 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000102_test__val_savanna__2__000014/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000102_test__val_savanna__2__000014/case.json new file mode 100644 index 0000000000000000000000000000000000000000..10ba0f8fb62f4d34fd74367fcaa629263647a3cc --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000102_test__val_savanna__2__000014/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "102", + "global_test_index": "102", + "dataset_index": "163", + "relative_path": "test/val_savanna/2/000014.mp4", + "clip_id": "test/val_savanna/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": 42000102, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.62948471400887, + "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_seed42_job85134/cases/g000102_test__val_savanna__2__000014/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000102_test__val_savanna__2__000014/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..baa5cc15a0cb474e49470aa330ddf0a286f95199 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000102_test__val_savanna__2__000014/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5f031efd8c629d8e65d012231bab91f019fadd0ee33278055f704f15ebfc21be +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000102_test__val_savanna__2__000014/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000102_test__val_savanna__2__000014/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..e9c032134135be2dc730da85b713cb18158ffd80 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000102_test__val_savanna__2__000014/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "102", + "global_test_index": "102", + "dataset_index": "163", + "relative_path": "test/val_savanna/2/000014.mp4", + "clip_id": "test/val_savanna/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": 42000102, + "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/2/000014_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": 0.006207610946148634, + "psnr": 22.13510513305664, + "lpips": 0.33687224984169006 + }, + "frames_100_200": { + "mse": 0.006520447786897421, + "psnr": 21.935258865356445, + "lpips": 0.33432698249816895 + }, + "frames_200_300": { + "mse": 0.007515307050198317, + "psnr": 21.315418243408203, + "lpips": 0.40561163425445557 + }, + "frames_300_400": { + "mse": 0.008030231110751629, + "psnr": 21.384973526000977, + "lpips": 0.41841205954551697 + }, + "frames_400_500": { + "mse": 0.008166293613612652, + "psnr": 20.990785598754883, + "lpips": 0.4489426016807556 + }, + "overall": { + "mse": 0.007287978194653988, + "psnr": 21.552309036254883, + "lpips": 0.38883310556411743 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000103_test__val_desert__1__000013/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000103_test__val_desert__1__000013/case.json new file mode 100644 index 0000000000000000000000000000000000000000..c0fe348663225e05c7f027315e90640140f44a86 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000103_test__val_desert__1__000013/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "103", + "global_test_index": "103", + "dataset_index": "31", + "relative_path": "test/val_desert/1/000013.mp4", + "clip_id": "test/val_desert/1/000013", + "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": 42000103, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 224.06713179906365, + "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_seed42_job85134/cases/g000103_test__val_desert__1__000013/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000103_test__val_desert__1__000013/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..616637354f01d090df8aeca7a8bad1c1fa7ad631 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000103_test__val_desert__1__000013/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ae5ff91d7f93bb7bdb296bffd128cfb55757e619c5ac18415952fcc8be461153 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000103_test__val_desert__1__000013/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000103_test__val_desert__1__000013/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..f53c4145703870b06688ebb65d459dbfa8f7a35b --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000103_test__val_desert__1__000013/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "103", + "global_test_index": "103", + "dataset_index": "31", + "relative_path": "test/val_desert/1/000013.mp4", + "clip_id": "test/val_desert/1/000013", + "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": 42000103, + "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/1/000013_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": 0.000820260145701468, + "psnr": 30.873126983642578, + "lpips": 0.010849556885659695 + }, + "frames_100_200": { + "mse": 0.0007975264452397823, + "psnr": 30.994190216064453, + "lpips": 0.0104972580447793 + }, + "frames_200_300": { + "mse": 0.0020676443818956614, + "psnr": 28.06914520263672, + "lpips": 0.03529307246208191 + }, + "frames_300_400": { + "mse": 0.003439509542658925, + "psnr": 25.178983688354492, + "lpips": 0.08703485131263733 + }, + "frames_400_500": { + "mse": 0.0016396321589127183, + "psnr": 28.41912078857422, + "lpips": 0.030461018905043602 + }, + "overall": { + "mse": 0.0017529145115986466, + "psnr": 28.706912994384766, + "lpips": 0.03482715040445328 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000104_test__val_savanna__0__000049/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000104_test__val_savanna__0__000049/case.json new file mode 100644 index 0000000000000000000000000000000000000000..45479c0d2a2918f90c49af813a7a0e623dde37c7 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000104_test__val_savanna__0__000049/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "104", + "global_test_index": "104", + "dataset_index": "241", + "relative_path": "test/val_savanna/0/000049.mp4", + "clip_id": "test/val_savanna/0/000049", + "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": 42000104, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 224.86217594100162, + "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_seed42_job85134/cases/g000104_test__val_savanna__0__000049/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000104_test__val_savanna__0__000049/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..c3320064e8877543b381048ea0df7e800e810f82 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000104_test__val_savanna__0__000049/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b781e2b9f173d6f93b6cf98e9bf38955b523b057f59f7caea180729eb3576288 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000104_test__val_savanna__0__000049/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000104_test__val_savanna__0__000049/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..8e259b8181e3168e443f641a3eb0600271d89203 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000104_test__val_savanna__0__000049/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "104", + "global_test_index": "104", + "dataset_index": "241", + "relative_path": "test/val_savanna/0/000049.mp4", + "clip_id": "test/val_savanna/0/000049", + "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": 42000104, + "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/0/000049_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": 0.021190134808421135, + "psnr": 17.286006927490234, + "lpips": 0.16933630406856537 + }, + "frames_100_200": { + "mse": 0.06258803606033325, + "psnr": 12.706128120422363, + "lpips": 0.34157755970954895 + }, + "frames_200_300": { + "mse": 0.07939502596855164, + "psnr": 11.328330993652344, + "lpips": 0.34757310152053833 + }, + "frames_300_400": { + "mse": 0.06034852936863899, + "psnr": 13.265539169311523, + "lpips": 0.2691700756549835 + }, + "frames_400_500": { + "mse": 0.08533813804388046, + "psnr": 10.850201606750488, + "lpips": 0.3386074900627136 + }, + "overall": { + "mse": 0.06177197024226189, + "psnr": 13.087241172790527, + "lpips": 0.2932528853416443 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000105_test__val_plains__0__000040/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000105_test__val_plains__0__000040/case.json new file mode 100644 index 0000000000000000000000000000000000000000..d6f5ccbf747d2f43301335563aeeb5803ea28a9d --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000105_test__val_plains__0__000040/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "105", + "global_test_index": "105", + "dataset_index": "458", + "relative_path": "test/val_plains/0/000040.mp4", + "clip_id": "test/val_plains/0/000040", + "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": 42000105, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 224.47621724905912, + "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_seed42_job85134/cases/g000105_test__val_plains__0__000040/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000105_test__val_plains__0__000040/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..001a0f8e0cb10d87e7eed0480830d83b048358f3 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000105_test__val_plains__0__000040/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f9d3663dfb0854c278410b12deea085f81d67429358d0c3cc7b76bab5e7c2fbb +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000105_test__val_plains__0__000040/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000105_test__val_plains__0__000040/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..b6275c777a86dd1efa2c0a2319f0d539c61ae431 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000105_test__val_plains__0__000040/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "105", + "global_test_index": "105", + "dataset_index": "458", + "relative_path": "test/val_plains/0/000040.mp4", + "clip_id": "test/val_plains/0/000040", + "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": 42000105, + "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/0/000040_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": 0.009438485838472843, + "psnr": 21.06991195678711, + "lpips": 0.2927570641040802 + }, + "frames_100_200": { + "mse": 0.005997783970087767, + "psnr": 22.509958267211914, + "lpips": 0.3470095098018646 + }, + "frames_200_300": { + "mse": 0.006232433952391148, + "psnr": 22.3701171875, + "lpips": 0.36618107557296753 + }, + "frames_300_400": { + "mse": 0.005108079873025417, + "psnr": 23.10940933227539, + "lpips": 0.38502857089042664 + }, + "frames_400_500": { + "mse": 0.005315364804118872, + "psnr": 23.177318572998047, + "lpips": 0.40892264246940613 + }, + "overall": { + "mse": 0.0064184293150901794, + "psnr": 22.447343826293945, + "lpips": 0.3599798083305359 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000106_test__val_iceplains__1__000049/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000106_test__val_iceplains__1__000049/case.json new file mode 100644 index 0000000000000000000000000000000000000000..e5e80692b88f70567c71660205e57166353d3100 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000106_test__val_iceplains__1__000049/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "106", + "global_test_index": "106", + "dataset_index": "371", + "relative_path": "test/val_iceplains/1/000049.mp4", + "clip_id": "test/val_iceplains/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": 42000106, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 224.08740106702317, + "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_seed42_job85134/cases/g000106_test__val_iceplains__1__000049/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000106_test__val_iceplains__1__000049/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..fcb97763f7f1aefc53157bb26e24b2fb7a6de7eb --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000106_test__val_iceplains__1__000049/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e13930f0f9dc2c391ecbcd97e5df02442627a1794403a746407f2317f7bfdc82 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000106_test__val_iceplains__1__000049/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000106_test__val_iceplains__1__000049/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..20e43ce79c7270ed2323914f8f4841a0eca07424 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000106_test__val_iceplains__1__000049/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "106", + "global_test_index": "106", + "dataset_index": "371", + "relative_path": "test/val_iceplains/1/000049.mp4", + "clip_id": "test/val_iceplains/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": 42000106, + "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/1/000049_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": 0.007830791175365448, + "psnr": 23.186986923217773, + "lpips": 0.05688408017158508 + }, + "frames_100_200": { + "mse": 0.0008175984839908779, + "psnr": 30.893102645874023, + "lpips": 0.012055592611432076 + }, + "frames_200_300": { + "mse": 0.0007891819695942104, + "psnr": 31.03765106201172, + "lpips": 0.011244868859648705 + }, + "frames_300_400": { + "mse": 0.0007792140822857618, + "psnr": 31.10092544555664, + "lpips": 0.011252577416598797 + }, + "frames_400_500": { + "mse": 0.0010892950231209397, + "psnr": 30.28287124633789, + "lpips": 0.013855848461389542 + }, + "overall": { + "mse": 0.00226121605373919, + "psnr": 29.300304412841797, + "lpips": 0.021058594807982445 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000107_test__val_savanna__3__000045/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000107_test__val_savanna__3__000045/case.json new file mode 100644 index 0000000000000000000000000000000000000000..fe6a5defc0f7746051d9995ef415c2341c5359e8 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000107_test__val_savanna__3__000045/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "107", + "global_test_index": "107", + "dataset_index": "230", + "relative_path": "test/val_savanna/3/000045.mp4", + "clip_id": "test/val_savanna/3/000045", + "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": 42000107, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 224.61492959002499, + "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_seed42_job85134/cases/g000107_test__val_savanna__3__000045/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000107_test__val_savanna__3__000045/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..6e2a0052f6327f63316e0774e2062aac36103e43 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000107_test__val_savanna__3__000045/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ee4199ec46d2a8eeafbeb868762bf67bcd1d7b07e54517c2e23e0ba50496fb1e +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000107_test__val_savanna__3__000045/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000107_test__val_savanna__3__000045/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..ae0ae6942a107a32cd1a8bc541bdbe9745fdf51b --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000107_test__val_savanna__3__000045/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "107", + "global_test_index": "107", + "dataset_index": "230", + "relative_path": "test/val_savanna/3/000045.mp4", + "clip_id": "test/val_savanna/3/000045", + "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": 42000107, + "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/3/000045_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": 0.011973926797509193, + "psnr": 22.513904571533203, + "lpips": 0.09315023571252823 + }, + "frames_100_200": { + "mse": 0.003919745795428753, + "psnr": 26.193452835083008, + "lpips": 0.029972167685627937 + }, + "frames_200_300": { + "mse": 0.003715503728017211, + "psnr": 26.008955001831055, + "lpips": 0.03214488923549652 + }, + "frames_300_400": { + "mse": 0.0064493203535676, + "psnr": 25.87047576904297, + "lpips": 0.04916377365589142 + }, + "frames_400_500": { + "mse": 0.05886874347925186, + "psnr": 12.36939811706543, + "lpips": 0.4224735498428345 + }, + "overall": { + "mse": 0.016985448077321053, + "psnr": 22.591236114501953, + "lpips": 0.12538093328475952 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000108_test__val_iceplains__2__000049/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000108_test__val_iceplains__2__000049/case.json new file mode 100644 index 0000000000000000000000000000000000000000..8417a8bf8fe2631203bd9ce135e157c7641239cc --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000108_test__val_iceplains__2__000049/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "108", + "global_test_index": "108", + "dataset_index": "372", + "relative_path": "test/val_iceplains/2/000049.mp4", + "clip_id": "test/val_iceplains/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": 42000108, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 224.72657536505722, + "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_seed42_job85134/cases/g000108_test__val_iceplains__2__000049/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000108_test__val_iceplains__2__000049/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..948647f51927cb9cd8d3270e8d175e0af5e077d8 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000108_test__val_iceplains__2__000049/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f97c15aee44397ae224748edc60dbffd1e4ec15fe5484d1cd6500c9ff42ad897 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000108_test__val_iceplains__2__000049/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000108_test__val_iceplains__2__000049/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..7bc485251ecf08d1c094cdd0b5732d9901b8ea43 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000108_test__val_iceplains__2__000049/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "108", + "global_test_index": "108", + "dataset_index": "372", + "relative_path": "test/val_iceplains/2/000049.mp4", + "clip_id": "test/val_iceplains/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": 42000108, + "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/2/000049_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": 0.00998007133603096, + "psnr": 20.69851303100586, + "lpips": 0.1075042337179184 + }, + "frames_100_200": { + "mse": 0.014378969557583332, + "psnr": 18.691299438476562, + "lpips": 0.11490283906459808 + }, + "frames_200_300": { + "mse": 0.05830337107181549, + "psnr": 13.723502159118652, + "lpips": 0.28336387872695923 + }, + "frames_300_400": { + "mse": 0.13575686514377594, + "psnr": 8.822724342346191, + "lpips": 0.5488847494125366 + }, + "frames_400_500": { + "mse": 0.11936222016811371, + "psnr": 9.376544952392578, + "lpips": 0.5506389141082764 + }, + "overall": { + "mse": 0.06755629926919937, + "psnr": 14.262517929077148, + "lpips": 0.3210589289665222 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000109_test__val_savanna__1__000042/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000109_test__val_savanna__1__000042/case.json new file mode 100644 index 0000000000000000000000000000000000000000..4d98d000644a9e763faec65b6640ea1a252772d8 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000109_test__val_savanna__1__000042/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "109", + "global_test_index": "109", + "dataset_index": "223", + "relative_path": "test/val_savanna/1/000042.mp4", + "clip_id": "test/val_savanna/1/000042", + "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": 42000109, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 224.63017290597782, + "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_seed42_job85134/cases/g000109_test__val_savanna__1__000042/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000109_test__val_savanna__1__000042/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..8fe7093c240f8ee225e49192331e2940bf95f4dc --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000109_test__val_savanna__1__000042/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0711167df64b5923f0402e2396c63ae27b2ce17d738c597419e71c9e33007912 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000109_test__val_savanna__1__000042/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000109_test__val_savanna__1__000042/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..f6aa9c1f200a792faff5dc054ff45d2c874a83f7 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000109_test__val_savanna__1__000042/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "109", + "global_test_index": "109", + "dataset_index": "223", + "relative_path": "test/val_savanna/1/000042.mp4", + "clip_id": "test/val_savanna/1/000042", + "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": 42000109, + "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/1/000042_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": 0.011332868598401546, + "psnr": 19.597736358642578, + "lpips": 0.24662373960018158 + }, + "frames_100_200": { + "mse": 0.03086951933801174, + "psnr": 15.442724227905273, + "lpips": 0.2905440032482147 + }, + "frames_200_300": { + "mse": 0.02433810755610466, + "psnr": 17.753868103027344, + "lpips": 0.24268867075443268 + }, + "frames_300_400": { + "mse": 0.02645569108426571, + "psnr": 16.275041580200195, + "lpips": 0.28595975041389465 + }, + "frames_400_500": { + "mse": 0.03716764226555824, + "psnr": 15.346318244934082, + "lpips": 0.43339017033576965 + }, + "overall": { + "mse": 0.026032764464616776, + "psnr": 16.883136749267578, + "lpips": 0.2998412847518921 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000110_test__val_desert__0__000026/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000110_test__val_desert__0__000026/case.json new file mode 100644 index 0000000000000000000000000000000000000000..becfc36aabead36d32f00f29fd4267bf604dbc99 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000110_test__val_desert__0__000026/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "110", + "global_test_index": "110", + "dataset_index": "68", + "relative_path": "test/val_desert/0/000026.mp4", + "clip_id": "test/val_desert/0/000026", + "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": 42000110, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 224.76354036701377, + "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_seed42_job85134/cases/g000110_test__val_desert__0__000026/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000110_test__val_desert__0__000026/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..dc0aecd166a2c82e677b3178b8cc739e06fb6dc5 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000110_test__val_desert__0__000026/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9e23a10ac20ab13650d1000cc2d9e234bbc46d96cd67fe32f62e12fc5be27672 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000110_test__val_desert__0__000026/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000110_test__val_desert__0__000026/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..bf292b9230f13bd6fe6f1b59b9fa29bd9def4056 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000110_test__val_desert__0__000026/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "110", + "global_test_index": "110", + "dataset_index": "68", + "relative_path": "test/val_desert/0/000026.mp4", + "clip_id": "test/val_desert/0/000026", + "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": 42000110, + "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/0/000026_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": 0.0024501821026206017, + "psnr": 26.84090805053711, + "lpips": 0.16052505373954773 + }, + "frames_100_200": { + "mse": 0.0034199052024632692, + "psnr": 24.72600746154785, + "lpips": 0.1419469267129898 + }, + "frames_200_300": { + "mse": 0.00566839100793004, + "psnr": 22.660615921020508, + "lpips": 0.1858409345149994 + }, + "frames_300_400": { + "mse": 0.004394045099616051, + "psnr": 23.957721710205078, + "lpips": 0.1913488358259201 + }, + "frames_400_500": { + "mse": 0.0034582666121423244, + "psnr": 25.5615177154541, + "lpips": 0.15922948718070984 + }, + "overall": { + "mse": 0.0038781575858592987, + "psnr": 24.74935531616211, + "lpips": 0.16777823865413666 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000111_test__val_plains__2__000042/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000111_test__val_plains__2__000042/case.json new file mode 100644 index 0000000000000000000000000000000000000000..0501dc359619f8356054eba821a1250971fac23f --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000111_test__val_plains__2__000042/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "111", + "global_test_index": "111", + "dataset_index": "461", + "relative_path": "test/val_plains/2/000042.mp4", + "clip_id": "test/val_plains/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": 42000111, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 223.4894815919688, + "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_seed42_job85134/cases/g000111_test__val_plains__2__000042/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000111_test__val_plains__2__000042/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..ae09a4699390616c6cf25a8d062710ad3ae8899a --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000111_test__val_plains__2__000042/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:75316f5a53e04962558f5db569e4f6dd951c342b8dd3515ac98e293588ef1ba5 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000111_test__val_plains__2__000042/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000111_test__val_plains__2__000042/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..03ea6ea631e0c072df476e81e48461402e7bc229 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000111_test__val_plains__2__000042/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "111", + "global_test_index": "111", + "dataset_index": "461", + "relative_path": "test/val_plains/2/000042.mp4", + "clip_id": "test/val_plains/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": 42000111, + "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/2/000042_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": 0.0034647537395358086, + "psnr": 26.571277618408203, + "lpips": 0.03744039312005043 + }, + "frames_100_200": { + "mse": 0.00783625803887844, + "psnr": 27.709064483642578, + "lpips": 0.05577164143323898 + }, + "frames_200_300": { + "mse": 0.006955979857593775, + "psnr": 25.358844757080078, + "lpips": 0.05057954788208008 + }, + "frames_300_400": { + "mse": 0.0035483611281961203, + "psnr": 27.324716567993164, + "lpips": 0.03771853446960449 + }, + "frames_400_500": { + "mse": 0.009143620729446411, + "psnr": 24.524551391601562, + "lpips": 0.0799357146024704 + }, + "overall": { + "mse": 0.00618979474529624, + "psnr": 26.297693252563477, + "lpips": 0.05228916555643082 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000112_test__val_savanna__1__000009/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000112_test__val_savanna__1__000009/case.json new file mode 100644 index 0000000000000000000000000000000000000000..d009e425b40135f3213b27571404258e8d29273a --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000112_test__val_savanna__1__000009/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "112", + "global_test_index": "112", + "dataset_index": "151", + "relative_path": "test/val_savanna/1/000009.mp4", + "clip_id": "test/val_savanna/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": 42000112, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 224.86092396103777, + "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_seed42_job85134/cases/g000112_test__val_savanna__1__000009/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000112_test__val_savanna__1__000009/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..2ca3db05aa701a56fa233ff7290dca5a9586a6a6 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000112_test__val_savanna__1__000009/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:da009dac81f1acff84455eb047ae9b336415e1916f576947acdecc1ce702cded +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000112_test__val_savanna__1__000009/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000112_test__val_savanna__1__000009/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..3fe99d1ed9053f21806c2731e231128ae888be87 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000112_test__val_savanna__1__000009/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "112", + "global_test_index": "112", + "dataset_index": "151", + "relative_path": "test/val_savanna/1/000009.mp4", + "clip_id": "test/val_savanna/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": 42000112, + "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/1/000009_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": 0.006553394719958305, + "psnr": 22.388277053833008, + "lpips": 0.2529396712779999 + }, + "frames_100_200": { + "mse": 0.007808459457010031, + "psnr": 21.258811950683594, + "lpips": 0.35489705204963684 + }, + "frames_200_300": { + "mse": 0.006613118574023247, + "psnr": 21.863637924194336, + "lpips": 0.4314395785331726 + }, + "frames_300_400": { + "mse": 0.00640931585803628, + "psnr": 22.011890411376953, + "lpips": 0.4249429702758789 + }, + "frames_400_500": { + "mse": 0.004638263955712318, + "psnr": 23.376249313354492, + "lpips": 0.3330015242099762 + }, + "overall": { + "mse": 0.006404511164873838, + "psnr": 22.179773330688477, + "lpips": 0.35944411158561707 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000113_test__val_desert__3__000027/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000113_test__val_desert__3__000027/case.json new file mode 100644 index 0000000000000000000000000000000000000000..8eee9fff07ab7115ffa69d10931d9477bec94b80 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000113_test__val_desert__3__000027/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "113", + "global_test_index": "113", + "dataset_index": "72", + "relative_path": "test/val_desert/3/000027.mp4", + "clip_id": "test/val_desert/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": 42000113, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 224.61715015792288, + "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_seed42_job85134/cases/g000113_test__val_desert__3__000027/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000113_test__val_desert__3__000027/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..39b8f50ee5bf06027de50e5abbd84690315496b9 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000113_test__val_desert__3__000027/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b6d9c1894f410c80d1897f9c17966b71780da49c17c49c898ec41741da32c2f6 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000113_test__val_desert__3__000027/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000113_test__val_desert__3__000027/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..c34cd1e6fdd7aebcecb0bf07228f374931fb50d4 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000113_test__val_desert__3__000027/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "113", + "global_test_index": "113", + "dataset_index": "72", + "relative_path": "test/val_desert/3/000027.mp4", + "clip_id": "test/val_desert/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": 42000113, + "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/3/000027_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": 0.0029521516989916563, + "psnr": 26.0258731842041, + "lpips": 0.07162731885910034 + }, + "frames_100_200": { + "mse": 0.0010811970569193363, + "psnr": 30.247804641723633, + "lpips": 0.016599560156464577 + }, + "frames_200_300": { + "mse": 0.0008303674403578043, + "psnr": 30.825550079345703, + "lpips": 0.0110757602378726 + }, + "frames_300_400": { + "mse": 0.0008529965416528285, + "psnr": 30.719228744506836, + "lpips": 0.011597874574363232 + }, + "frames_400_500": { + "mse": 0.0008991160430014133, + "psnr": 30.542896270751953, + "lpips": 0.013344638049602509 + }, + "overall": { + "mse": 0.0013231656048446894, + "psnr": 29.672271728515625, + "lpips": 0.024849029257893562 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000114_test__val_plains__0__000024/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000114_test__val_plains__0__000024/case.json new file mode 100644 index 0000000000000000000000000000000000000000..a697a7b06c97212fd3efc6f9ed25588a5d504d00 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000114_test__val_plains__0__000024/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "114", + "global_test_index": "114", + "dataset_index": "427", + "relative_path": "test/val_plains/0/000024.mp4", + "clip_id": "test/val_plains/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": 42000114, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.09896080498584, + "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_seed42_job85134/cases/g000114_test__val_plains__0__000024/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000114_test__val_plains__0__000024/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..ea27e675136a3f12a0bbb76ae6294a4d47d86dff --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000114_test__val_plains__0__000024/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a7f0d8895c9b6ecad0a06df649f764c01966747068f1305d6d1c8621e6cdcd6f +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000114_test__val_plains__0__000024/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000114_test__val_plains__0__000024/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..fe06d320bec5fc53105d7bfb6ca4f670db45d7c5 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000114_test__val_plains__0__000024/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "114", + "global_test_index": "114", + "dataset_index": "427", + "relative_path": "test/val_plains/0/000024.mp4", + "clip_id": "test/val_plains/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": 42000114, + "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/0/000024_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": 0.0036065580788999796, + "psnr": 24.840988159179688, + "lpips": 0.2514917552471161 + }, + "frames_100_200": { + "mse": 0.007619716692715883, + "psnr": 21.435710906982422, + "lpips": 0.37156376242637634 + }, + "frames_200_300": { + "mse": 0.00757497688755393, + "psnr": 21.392719268798828, + "lpips": 0.3687761723995209 + }, + "frames_300_400": { + "mse": 0.019204648211598396, + "psnr": 17.835609436035156, + "lpips": 0.5403845310211182 + }, + "frames_400_500": { + "mse": 0.05145234242081642, + "psnr": 14.534814834594727, + "lpips": 0.48176106810569763 + }, + "overall": { + "mse": 0.017891651019454002, + "psnr": 20.007966995239258, + "lpips": 0.4027954339981079 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000115_test__val_iceplains__3__000029/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000115_test__val_iceplains__3__000029/case.json new file mode 100644 index 0000000000000000000000000000000000000000..5f930b25154a390bf363e83112b4b5a163ba6a42 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000115_test__val_iceplains__3__000029/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "115", + "global_test_index": "115", + "dataset_index": "317", + "relative_path": "test/val_iceplains/3/000029.mp4", + "clip_id": "test/val_iceplains/3/000029", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_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": 42000115, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.28111911495216, + "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_seed42_job85134/cases/g000115_test__val_iceplains__3__000029/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000115_test__val_iceplains__3__000029/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..fb022b9fa935574d1d14f8cf1d1995da7c3ba8af --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000115_test__val_iceplains__3__000029/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c9d68136d6ef4b67da1ffe96e3ed27ce65e10494108ea21c646555a90f4c6b87 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000115_test__val_iceplains__3__000029/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000115_test__val_iceplains__3__000029/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..e5e524f99e04a1d3d30f1c7ef3fa7ba3babd4be0 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000115_test__val_iceplains__3__000029/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "115", + "global_test_index": "115", + "dataset_index": "317", + "relative_path": "test/val_iceplains/3/000029.mp4", + "clip_id": "test/val_iceplains/3/000029", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_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": 42000115, + "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/3/000029_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": 0.008050039410591125, + "psnr": 22.196569442749023, + "lpips": 0.11570991575717926 + }, + "frames_100_200": { + "mse": 0.005875909235328436, + "psnr": 22.755352020263672, + "lpips": 0.07740610092878342 + }, + "frames_200_300": { + "mse": 0.12687048316001892, + "psnr": 11.706250190734863, + "lpips": 0.30799347162246704 + }, + "frames_300_400": { + "mse": 0.10898943245410919, + "psnr": 10.29513168334961, + "lpips": 0.33254513144493103 + }, + "frames_400_500": { + "mse": 0.06569962948560715, + "psnr": 13.174166679382324, + "lpips": 0.1996360719203949 + }, + "overall": { + "mse": 0.06309710443019867, + "psnr": 16.025493621826172, + "lpips": 0.20665811002254486 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000116_test__val_iceplains__2__000001/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000116_test__val_iceplains__2__000001/case.json new file mode 100644 index 0000000000000000000000000000000000000000..5d6000f8fae5a6c8c20d62e04fe379233e01fdae --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000116_test__val_iceplains__2__000001/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "116", + "global_test_index": "116", + "dataset_index": "249", + "relative_path": "test/val_iceplains/2/000001.mp4", + "clip_id": "test/val_iceplains/2/000001", + "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": 42000116, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 224.10163770406507, + "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_seed42_job85134/cases/g000116_test__val_iceplains__2__000001/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000116_test__val_iceplains__2__000001/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..369f942475f831f631595e96d58f9bcc105dbbec --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000116_test__val_iceplains__2__000001/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d498a94ef2ca23aa2e9b37960e0dd5fb13504e8dc5f8f2f722c46baf04885e28 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000116_test__val_iceplains__2__000001/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000116_test__val_iceplains__2__000001/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..619c13096f5e2f28001dfeeef0ee7c05984db9cf --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000116_test__val_iceplains__2__000001/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "116", + "global_test_index": "116", + "dataset_index": "249", + "relative_path": "test/val_iceplains/2/000001.mp4", + "clip_id": "test/val_iceplains/2/000001", + "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": 42000116, + "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/2/000001_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": 0.0050717806443572044, + "psnr": 24.92470932006836, + "lpips": 0.04472696781158447 + }, + "frames_100_200": { + "mse": 0.1007758304476738, + "psnr": 10.81521987915039, + "lpips": 0.4268266260623932 + }, + "frames_200_300": { + "mse": 0.1268777698278427, + "psnr": 9.258744239807129, + "lpips": 0.5090939998626709 + }, + "frames_300_400": { + "mse": 0.08911297470331192, + "psnr": 11.728388786315918, + "lpips": 0.47606658935546875 + }, + "frames_400_500": { + "mse": 0.08120406419038773, + "psnr": 11.527795791625977, + "lpips": 0.48087626695632935 + }, + "overall": { + "mse": 0.08060848712921143, + "psnr": 13.650970458984375, + "lpips": 0.38751810789108276 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000117_test__val_iceplains__2__000046/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000117_test__val_iceplains__2__000046/case.json new file mode 100644 index 0000000000000000000000000000000000000000..bdf1f6047dcf3cfb57039ad8770830ed51b374bd --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000117_test__val_iceplains__2__000046/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "117", + "global_test_index": "117", + "dataset_index": "364", + "relative_path": "test/val_iceplains/2/000046.mp4", + "clip_id": "test/val_iceplains/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": 42000117, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.23305268096738, + "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_seed42_job85134/cases/g000117_test__val_iceplains__2__000046/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000117_test__val_iceplains__2__000046/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..1997ad8d78f68359a072695423e4fb2a99a0d5ee --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000117_test__val_iceplains__2__000046/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9a6e0ebda8cf3d922fad6cb956288d4eed05d6eb6f136b7f2d2e9329a537f127 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000117_test__val_iceplains__2__000046/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000117_test__val_iceplains__2__000046/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..2813eb15a5ae9c3927767eacf414950c0a2856c3 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000117_test__val_iceplains__2__000046/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "117", + "global_test_index": "117", + "dataset_index": "364", + "relative_path": "test/val_iceplains/2/000046.mp4", + "clip_id": "test/val_iceplains/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": 42000117, + "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/2/000046_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": 0.01581919938325882, + "psnr": 18.348896026611328, + "lpips": 0.1549374908208847 + }, + "frames_100_200": { + "mse": 0.02715037390589714, + "psnr": 16.329490661621094, + "lpips": 0.2181043028831482 + }, + "frames_200_300": { + "mse": 0.00877596065402031, + "psnr": 21.82852554321289, + "lpips": 0.18737727403640747 + }, + "frames_300_400": { + "mse": 0.023711448535323143, + "psnr": 16.574098587036133, + "lpips": 0.2038159966468811 + }, + "frames_400_500": { + "mse": 0.019228100776672363, + "psnr": 18.11199378967285, + "lpips": 0.13975372910499573 + }, + "overall": { + "mse": 0.018937015905976295, + "psnr": 18.238601684570312, + "lpips": 0.18079774081707 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000118_test__val_desert__2__000037/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000118_test__val_desert__2__000037/case.json new file mode 100644 index 0000000000000000000000000000000000000000..f5769f20739f03f0251093fb779e283c53d1d310 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000118_test__val_desert__2__000037/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "118", + "global_test_index": "118", + "dataset_index": "96", + "relative_path": "test/val_desert/2/000037.mp4", + "clip_id": "test/val_desert/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": 42000118, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.0000265489798, + "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_seed42_job85134/cases/g000118_test__val_desert__2__000037/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000118_test__val_desert__2__000037/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..1f08f3ea114b5b5e09ca5c6965269240d03b2724 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000118_test__val_desert__2__000037/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d88c9caa399c435b19e675d00449e446705b41a9eb81d359debd3700a134e1ab +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000118_test__val_desert__2__000037/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000118_test__val_desert__2__000037/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..b7cfb6ec23b898464b033b4b3da6e3b5cc906613 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000118_test__val_desert__2__000037/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "118", + "global_test_index": "118", + "dataset_index": "96", + "relative_path": "test/val_desert/2/000037.mp4", + "clip_id": "test/val_desert/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": 42000118, + "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/2/000037_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": 0.00074342469451949, + "psnr": 31.2935791015625, + "lpips": 0.010082283057272434 + }, + "frames_100_200": { + "mse": 0.0013380161253735423, + "psnr": 30.095592498779297, + "lpips": 0.018674444407224655 + }, + "frames_200_300": { + "mse": 0.0009635143796913326, + "psnr": 30.788969039916992, + "lpips": 0.012648504227399826 + }, + "frames_300_400": { + "mse": 0.0010534935863688588, + "psnr": 30.612733840942383, + "lpips": 0.014251100830733776 + }, + "frames_400_500": { + "mse": 0.0008002668037079275, + "psnr": 30.988563537597656, + "lpips": 0.011212456971406937 + }, + "overall": { + "mse": 0.0009797430830076337, + "psnr": 30.755889892578125, + "lpips": 0.01337375957518816 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000119_test__val_savanna__3__000029/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000119_test__val_savanna__3__000029/case.json new file mode 100644 index 0000000000000000000000000000000000000000..d178fa92fa267f0da0d774c423701b5a0bbee77b --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000119_test__val_savanna__3__000029/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "119", + "global_test_index": "119", + "dataset_index": "190", + "relative_path": "test/val_savanna/3/000029.mp4", + "clip_id": "test/val_savanna/3/000029", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_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": 42000119, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 223.55340887000784, + "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_seed42_job85134/cases/g000119_test__val_savanna__3__000029/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000119_test__val_savanna__3__000029/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..74973e6f8a8cd88268a1ed984721e9d005036c26 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000119_test__val_savanna__3__000029/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:65d593792ae6c6dee4e928f6123b62b596c5238523365d27f891f3dec32aed72 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000119_test__val_savanna__3__000029/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000119_test__val_savanna__3__000029/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..bb69f9091eb202798320b9030f6ca44a5ad2bed0 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000119_test__val_savanna__3__000029/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "119", + "global_test_index": "119", + "dataset_index": "190", + "relative_path": "test/val_savanna/3/000029.mp4", + "clip_id": "test/val_savanna/3/000029", + "source_offset": "100", + "context_start": "100", + "context_end": "700", + "target_start": "700", + "target_end": "1200", + "n_frames_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": 42000119, + "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/3/000029_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": 0.008202566765248775, + "psnr": 21.361848831176758, + "lpips": 0.3482595384120941 + }, + "frames_100_200": { + "mse": 0.006559397093951702, + "psnr": 21.8889217376709, + "lpips": 0.39431071281433105 + }, + "frames_200_300": { + "mse": 0.008997504599392414, + "psnr": 20.900129318237305, + "lpips": 0.4149100184440613 + }, + "frames_300_400": { + "mse": 0.007306823041290045, + "psnr": 21.506305694580078, + "lpips": 0.4284696578979492 + }, + "frames_400_500": { + "mse": 0.012927297502756119, + "psnr": 19.561025619506836, + "lpips": 0.46512681245803833 + }, + "overall": { + "mse": 0.008798717521131039, + "psnr": 21.043643951416016, + "lpips": 0.4102153480052948 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000120_test__val_savanna__2__000044/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000120_test__val_savanna__2__000044/case.json new file mode 100644 index 0000000000000000000000000000000000000000..9d7d955c78da86200066241771c1123a3b0fcc18 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000120_test__val_savanna__2__000044/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "120", + "global_test_index": "120", + "dataset_index": "227", + "relative_path": "test/val_savanna/2/000044.mp4", + "clip_id": "test/val_savanna/2/000044", + "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": 42000120, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 224.57086759898812, + "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_seed42_job85134/cases/g000120_test__val_savanna__2__000044/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000120_test__val_savanna__2__000044/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..c836d67e21b06c9b8fbf49394dfccc8ed48eedd8 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000120_test__val_savanna__2__000044/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ba079540771f981bafa6c92951ab562f64b67f189d44baedef83d9424b6bd786 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000120_test__val_savanna__2__000044/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000120_test__val_savanna__2__000044/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..04e3ef5418ccec53a6f8f97c707e36ce9e9defff --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000120_test__val_savanna__2__000044/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "120", + "global_test_index": "120", + "dataset_index": "227", + "relative_path": "test/val_savanna/2/000044.mp4", + "clip_id": "test/val_savanna/2/000044", + "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": 42000120, + "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/2/000044_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": 0.009253856725990772, + "psnr": 20.876197814941406, + "lpips": 0.24109375476837158 + }, + "frames_100_200": { + "mse": 0.029303543269634247, + "psnr": 15.97142219543457, + "lpips": 0.29792892932891846 + }, + "frames_200_300": { + "mse": 0.02914339490234852, + "psnr": 15.840256690979004, + "lpips": 0.33532199263572693 + }, + "frames_300_400": { + "mse": 0.05449453368782997, + "psnr": 12.761920928955078, + "lpips": 0.33437812328338623 + }, + "frames_400_500": { + "mse": 0.07077238708734512, + "psnr": 11.571613311767578, + "lpips": 0.36852890253067017 + }, + "overall": { + "mse": 0.03859354555606842, + "psnr": 15.404282569885254, + "lpips": 0.3154503405094147 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000121_test__val_iceplains__0__000000/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000121_test__val_iceplains__0__000000/case.json new file mode 100644 index 0000000000000000000000000000000000000000..a718d8b96fb33892ee9118d744f954901da314ec --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000121_test__val_iceplains__0__000000/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "121", + "global_test_index": "121", + "dataset_index": "244", + "relative_path": "test/val_iceplains/0/000000.mp4", + "clip_id": "test/val_iceplains/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": 42000121, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.6777098890161, + "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_seed42_job85134/cases/g000121_test__val_iceplains__0__000000/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000121_test__val_iceplains__0__000000/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..dc5d0e999aba450c1c3903994b67030682f6bf1f --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000121_test__val_iceplains__0__000000/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d29b8b8edf23d49d4cd7c160553d95337c5aa3678dc539cca8b4e0aafe7415b9 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000121_test__val_iceplains__0__000000/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000121_test__val_iceplains__0__000000/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..edee2eb5482837eab5e363aaea71f27dfd46814e --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000121_test__val_iceplains__0__000000/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "121", + "global_test_index": "121", + "dataset_index": "244", + "relative_path": "test/val_iceplains/0/000000.mp4", + "clip_id": "test/val_iceplains/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": 42000121, + "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/0/000000_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": 0.00525874737650156, + "psnr": 24.367565155029297, + "lpips": 0.042505890130996704 + }, + "frames_100_200": { + "mse": 0.01715528964996338, + "psnr": 19.459611892700195, + "lpips": 0.1244993582367897 + }, + "frames_200_300": { + "mse": 0.008173355832695961, + "psnr": 24.486778259277344, + "lpips": 0.055174559354782104 + }, + "frames_300_400": { + "mse": 0.0030459007248282433, + "psnr": 27.284631729125977, + "lpips": 0.025556158274412155 + }, + "frames_400_500": { + "mse": 0.0035450533032417297, + "psnr": 26.008255004882812, + "lpips": 0.034042488783597946 + }, + "overall": { + "mse": 0.007435668725520372, + "psnr": 24.321367263793945, + "lpips": 0.05635569617152214 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000122_test__val_desert__3__000015/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000122_test__val_desert__3__000015/case.json new file mode 100644 index 0000000000000000000000000000000000000000..ad4e66781a25ac35415eb985d6d150cabdf7fe2d --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000122_test__val_desert__3__000015/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "122", + "global_test_index": "122", + "dataset_index": "38", + "relative_path": "test/val_desert/3/000015.mp4", + "clip_id": "test/val_desert/3/000015", + "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": 42000122, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.26530668896157, + "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_seed42_job85134/cases/g000122_test__val_desert__3__000015/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000122_test__val_desert__3__000015/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..3c6926339c63dd90a1c13ff922daeb6ee7bd742f --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000122_test__val_desert__3__000015/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:927e11e881bb50fa83c78f26d22a591eb241bc3e432ba81e5833939f8ead7eff +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000122_test__val_desert__3__000015/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000122_test__val_desert__3__000015/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..c268d19b64a33ccd4c122e3d93a2d6b0c013b5fb --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000122_test__val_desert__3__000015/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "122", + "global_test_index": "122", + "dataset_index": "38", + "relative_path": "test/val_desert/3/000015.mp4", + "clip_id": "test/val_desert/3/000015", + "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": 42000122, + "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/3/000015_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": 0.01937691494822502, + "psnr": 17.32380485534668, + "lpips": 0.40796732902526855 + }, + "frames_100_200": { + "mse": 0.03233123570680618, + "psnr": 15.053415298461914, + "lpips": 0.4560030996799469 + }, + "frames_200_300": { + "mse": 0.026970149949193, + "psnr": 16.021671295166016, + "lpips": 0.48127028346061707 + }, + "frames_300_400": { + "mse": 0.043933067470788956, + "psnr": 14.969165802001953, + "lpips": 0.4268436133861542 + }, + "frames_400_500": { + "mse": 0.07941179722547531, + "psnr": 11.475624084472656, + "lpips": 0.6406839489936829 + }, + "overall": { + "mse": 0.040404632687568665, + "psnr": 14.968737602233887, + "lpips": 0.4825536608695984 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000123_test__val_iceplains__2__000017/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000123_test__val_iceplains__2__000017/case.json new file mode 100644 index 0000000000000000000000000000000000000000..fdea7b7e213ebbb3d3d2af5d4857060bee23924e --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000123_test__val_iceplains__2__000017/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "123", + "global_test_index": "123", + "dataset_index": "291", + "relative_path": "test/val_iceplains/2/000017.mp4", + "clip_id": "test/val_iceplains/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": 42000123, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.11638649494853, + "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_seed42_job85134/cases/g000123_test__val_iceplains__2__000017/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000123_test__val_iceplains__2__000017/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..1aa78578297167a4814904d40c58f75e2b44d954 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000123_test__val_iceplains__2__000017/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4abba00a4ccfbd2549aef310ab24e35e23f8affc1f3947b422b475463fe87c72 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000123_test__val_iceplains__2__000017/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000123_test__val_iceplains__2__000017/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..ab8ba2068fcda0bd78e07dd759decfd10718827f --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000123_test__val_iceplains__2__000017/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "123", + "global_test_index": "123", + "dataset_index": "291", + "relative_path": "test/val_iceplains/2/000017.mp4", + "clip_id": "test/val_iceplains/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": 42000123, + "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/2/000017_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": 0.004297051578760147, + "psnr": 25.847450256347656, + "lpips": 0.038234591484069824 + }, + "frames_100_200": { + "mse": 0.03827892243862152, + "psnr": 15.125532150268555, + "lpips": 0.29709500074386597 + }, + "frames_200_300": { + "mse": 0.02169654332101345, + "psnr": 17.04098892211914, + "lpips": 0.1897933930158615 + }, + "frames_300_400": { + "mse": 0.021092332899570465, + "psnr": 17.00200843811035, + "lpips": 0.15532498061656952 + }, + "frames_400_500": { + "mse": 0.02693336084485054, + "psnr": 17.17974281311035, + "lpips": 0.21695472300052643 + }, + "overall": { + "mse": 0.022459641098976135, + "psnr": 18.43914222717285, + "lpips": 0.17948052287101746 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000124_test__val_desert__2__000032/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000124_test__val_desert__2__000032/case.json new file mode 100644 index 0000000000000000000000000000000000000000..c7106d480556ae5fd7eed362fb8c6354d44b31ee --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000124_test__val_desert__2__000032/case.json @@ -0,0 +1,2071 @@ +{ + "case": { + "manifest_index": "124", + "global_test_index": "124", + "dataset_index": "85", + "relative_path": "test/val_desert/2/000032.mp4", + "clip_id": "test/val_desert/2/000032", + "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": 42000124, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.13051542604808, + "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_seed42_job85134/cases/g000124_test__val_desert__2__000032/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000124_test__val_desert__2__000032/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..b8e19979eedf71bf4920693e39c165eb9adbf807 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000124_test__val_desert__2__000032/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b34cddf7600aea14dba3a743e6ac16094315a962d11cfe5c0bfd18dd39d49fbe +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000124_test__val_desert__2__000032/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000124_test__val_desert__2__000032/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..ca4d9440d59898c98538a95766f1455d58e18be2 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000124_test__val_desert__2__000032/metrics.json @@ -0,0 +1,80 @@ +{ + "case": { + "manifest_index": "124", + "global_test_index": "124", + "dataset_index": "85", + "relative_path": "test/val_desert/2/000032.mp4", + "clip_id": "test/val_desert/2/000032", + "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": 42000124, + "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/2/000032_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": 0.004919752478599548, + "psnr": 23.460369110107422, + "lpips": 0.12622776627540588 + }, + "frames_100_200": { + "mse": 0.010460968129336834, + "psnr": 19.960670471191406, + "lpips": 0.2102670818567276 + }, + "frames_200_300": { + "mse": 0.012747722677886486, + "psnr": 19.222492218017578, + "lpips": 0.1916816383600235 + }, + "frames_300_400": { + "mse": 0.02343394234776497, + "psnr": 17.45945930480957, + "lpips": 0.26624277234077454 + }, + "frames_400_500": { + "mse": 0.011530867777764797, + "psnr": 21.21770477294922, + "lpips": 0.16168129444122314 + }, + "overall": { + "mse": 0.012618649750947952, + "psnr": 20.26413917541504, + "lpips": 0.19122011959552765 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000125_test__val_iceplains__0__000044/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000125_test__val_iceplains__0__000044/case.json new file mode 100644 index 0000000000000000000000000000000000000000..4925e261e9601a759f76db03dfcfb3a486b260db --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_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": 42000125, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 224.89850302692503, + "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_seed42_job85134/cases/g000125_test__val_iceplains__0__000044/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000125_test__val_iceplains__0__000044/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..21e6d517a0be58631632169cd1e5022572c27d3a --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000125_test__val_iceplains__0__000044/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8c6ed6af1a1686e0db685339ac0d1b8bfa9db6af9343ce92e3d8e431b59a054e +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000125_test__val_iceplains__0__000044/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000125_test__val_iceplains__0__000044/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..e92a5ca63c04e934c05f418896e7dcee11ed42f7 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_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": 42000125, + "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": 0.0009137155138887465, + "psnr": 30.47869873046875, + "lpips": 0.011241997592151165 + }, + "frames_100_200": { + "mse": 0.0008664834313094616, + "psnr": 30.653779983520508, + "lpips": 0.011896973475813866 + }, + "frames_200_300": { + "mse": 0.0008403597166761756, + "psnr": 30.776287078857422, + "lpips": 0.012112928554415703 + }, + "frames_300_400": { + "mse": 0.0011957015376538038, + "psnr": 29.919801712036133, + "lpips": 0.019635695964097977 + }, + "frames_400_500": { + "mse": 0.000896359037142247, + "psnr": 30.538179397583008, + "lpips": 0.017297927290201187 + }, + "overall": { + "mse": 0.0009425238240510225, + "psnr": 30.473350524902344, + "lpips": 0.01443710457533598 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000126_test__val_desert__1__000020/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000126_test__val_desert__1__000020/case.json new file mode 100644 index 0000000000000000000000000000000000000000..9a729cfbf675774f1233661c631ec0e91c107ec9 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_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": 42000126, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 224.65681898104958, + "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_seed42_job85134/cases/g000126_test__val_desert__1__000020/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000126_test__val_desert__1__000020/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..ce96001c2484ddf74525c3e4942b9fc679986b7f --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000126_test__val_desert__1__000020/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6b4c9aa0cbe47a70beca3f94b371a70dfac60237d0eba2f4a37ce4e46e03f64d +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000126_test__val_desert__1__000020/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000126_test__val_desert__1__000020/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..57a0329e1a309bf3b9cc2903f1e233a72263034d --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_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": 42000126, + "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": 0.0014570680214092135, + "psnr": 29.660486221313477, + "lpips": 0.01841784082353115 + }, + "frames_100_200": { + "mse": 0.0016074382001534104, + "psnr": 30.308626174926758, + "lpips": 0.01512676477432251 + }, + "frames_200_300": { + "mse": 0.0008293997962027788, + "psnr": 30.843603134155273, + "lpips": 0.012148465029895306 + }, + "frames_300_400": { + "mse": 0.0020816829055547714, + "psnr": 28.69129180908203, + "lpips": 0.02866637147963047 + }, + "frames_400_500": { + "mse": 0.008755351416766644, + "psnr": 23.410293579101562, + "lpips": 0.061335816979408264 + }, + "overall": { + "mse": 0.002946187974885106, + "psnr": 28.582860946655273, + "lpips": 0.027139052748680115 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000127_test__val_savanna__0__000036/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000127_test__val_savanna__0__000036/case.json new file mode 100644 index 0000000000000000000000000000000000000000..d4766c59bf9c04ac825ddb083cd84766a78a5ccc --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_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": 42000127, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 224.55529581999872, + "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_seed42_job85134/cases/g000127_test__val_savanna__0__000036/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000127_test__val_savanna__0__000036/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..00b2a06391a0a7114c9f3cba4e382d6d2f49f0ca --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000127_test__val_savanna__0__000036/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1257895ca591653fa412c29a450de6d843541617e00911c4252e4ecfd955c967 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000127_test__val_savanna__0__000036/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000127_test__val_savanna__0__000036/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..c53a1173d7c5910c1949de4a4fe5c0a2dbaec207 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_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": 42000127, + "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": 0.006780956871807575, + "psnr": 22.192113876342773, + "lpips": 0.3357742428779602 + }, + "frames_100_200": { + "mse": 0.007062689866870642, + "psnr": 21.606430053710938, + "lpips": 0.417106032371521 + }, + "frames_200_300": { + "mse": 0.006927165202796459, + "psnr": 21.649316787719727, + "lpips": 0.4243547320365906 + }, + "frames_300_400": { + "mse": 0.017266565933823586, + "psnr": 18.7180118560791, + "lpips": 0.5058624148368835 + }, + "frames_400_500": { + "mse": 0.02076978050172329, + "psnr": 17.422788619995117, + "lpips": 0.519626259803772 + }, + "overall": { + "mse": 0.011761431582272053, + "psnr": 20.317729949951172, + "lpips": 0.4405446946620941 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000128_test__val_iceplains__2__000034/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000128_test__val_iceplains__2__000034/case.json new file mode 100644 index 0000000000000000000000000000000000000000..fe326688b20dad45d2aefcf32f2ebe07e8f901b1 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_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": 42000128, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.03828088799492, + "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_seed42_job85134/cases/g000128_test__val_iceplains__2__000034/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000128_test__val_iceplains__2__000034/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..3d49de75046e38aed60ad79780e6dbbe26e5ba4e --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000128_test__val_iceplains__2__000034/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0e096e99ede33f259dcd36b648acc7ce164879ac48ffb4391537ba2877da8322 +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000128_test__val_iceplains__2__000034/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000128_test__val_iceplains__2__000034/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..22cbe320f6a96933f2a9fa1bff76a246ff669df1 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_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": 42000128, + "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": 0.0009485773625783622, + "psnr": 30.357189178466797, + "lpips": 0.015851648524403572 + }, + "frames_100_200": { + "mse": 0.0011449679732322693, + "psnr": 30.47137451171875, + "lpips": 0.0147525230422616 + }, + "frames_200_300": { + "mse": 0.017051996663212776, + "psnr": 24.46533203125, + "lpips": 0.08740698546171188 + }, + "frames_300_400": { + "mse": 0.10127272456884384, + "psnr": 10.936614036560059, + "lpips": 0.46029236912727356 + }, + "frames_400_500": { + "mse": 0.09476761519908905, + "psnr": 10.907756805419922, + "lpips": 0.49915412068367004 + }, + "overall": { + "mse": 0.04303717985749245, + "psnr": 21.42765235900879, + "lpips": 0.21549153327941895 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000129_test__val_savanna__0__000001/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000129_test__val_savanna__0__000001/case.json new file mode 100644 index 0000000000000000000000000000000000000000..ef65ddedcdce1f30ad519a7ec5d63a709c386d0f --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_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": 42000129, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.37298094294965, + "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_seed42_job85134/cases/g000129_test__val_savanna__0__000001/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000129_test__val_savanna__0__000001/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..c59df68e2f7f417ff454d9522d2bc1aaa3226ecc --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000129_test__val_savanna__0__000001/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:38db7533cd40ed8c1d2b7690c4d9f994cadbffa01b25bd231d0dc980ed9829dd +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000129_test__val_savanna__0__000001/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000129_test__val_savanna__0__000001/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..246f1452cd97fab341d73e423b6ee1400b9a4559 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_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": 42000129, + "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": 0.0014679847517982125, + "psnr": 28.761699676513672, + "lpips": 0.022369833663105965 + }, + "frames_100_200": { + "mse": 0.0019005642971023917, + "psnr": 28.341045379638672, + "lpips": 0.027400638908147812 + }, + "frames_200_300": { + "mse": 0.01229972392320633, + "psnr": 20.783615112304688, + "lpips": 0.21529299020767212 + }, + "frames_300_400": { + "mse": 0.01545175351202488, + "psnr": 18.389652252197266, + "lpips": 0.30183231830596924 + }, + "frames_400_500": { + "mse": 0.004633307922631502, + "psnr": 23.91089630126953, + "lpips": 0.07784868031740189 + }, + "overall": { + "mse": 0.007150666322559118, + "psnr": 24.037382125854492, + "lpips": 0.1289488822221756 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000130_test__val_plains__3__000021/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000130_test__val_plains__3__000021/case.json new file mode 100644 index 0000000000000000000000000000000000000000..74a8f864476058f709387c37646cd899d31ac756 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_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": 42000130, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 224.11570577102248, + "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_seed42_job85134/cases/g000130_test__val_plains__3__000021/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000130_test__val_plains__3__000021/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..11f10da1421e150f0a8d3da67c845aed66f7c066 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_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:c02608eb6c257dd4c665ddbe201e0d49795a0038819f33fa45175a8188b363ec +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000130_test__val_plains__3__000021/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000130_test__val_plains__3__000021/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..b9e4568439fc4e6e36e8d90cbf796db92ebd94b3 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_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": 42000130, + "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": 0.003394520375877619, + "psnr": 24.9071102142334, + "lpips": 0.25273019075393677 + }, + "frames_100_200": { + "mse": 0.004122523125261068, + "psnr": 24.032936096191406, + "lpips": 0.3106400966644287 + }, + "frames_200_300": { + "mse": 0.004185073543339968, + "psnr": 23.827421188354492, + "lpips": 0.3425300121307373 + }, + "frames_300_400": { + "mse": 0.011508551426231861, + "psnr": 21.485370635986328, + "lpips": 0.3484044671058655 + }, + "frames_400_500": { + "mse": 0.04598041623830795, + "psnr": 14.385197639465332, + "lpips": 0.3603768050670624 + }, + "overall": { + "mse": 0.013838215731084347, + "psnr": 21.72760772705078, + "lpips": 0.3229362964630127 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000131_test__val_desert__2__000025/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000131_test__val_desert__2__000025/case.json new file mode 100644 index 0000000000000000000000000000000000000000..4ea03f080374c876c5329729a687b30a91b8fc23 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_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": 42000131, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.4318174669752, + "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_seed42_job85134/cases/g000131_test__val_desert__2__000025/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000131_test__val_desert__2__000025/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..68496d25cb5fcfe974fec57bf2ecbb8193865e20 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_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:d8e9d7180d31fa44093d03c585df490dea48bc011eea44593de3ba317a397a1c +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000131_test__val_desert__2__000025/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000131_test__val_desert__2__000025/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..56f01ac2f710691383d63cd1951fcfa756780da2 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_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": 42000131, + "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": 0.010552216321229935, + "psnr": 20.191654205322266, + "lpips": 0.26330387592315674 + }, + "frames_100_200": { + "mse": 0.015796490013599396, + "psnr": 18.168834686279297, + "lpips": 0.26454734802246094 + }, + "frames_200_300": { + "mse": 0.01861884631216526, + "psnr": 18.06878662109375, + "lpips": 0.28123652935028076 + }, + "frames_300_400": { + "mse": 0.027750594541430473, + "psnr": 15.944437026977539, + "lpips": 0.3269141912460327 + }, + "frames_400_500": { + "mse": 0.027100063860416412, + "psnr": 16.224170684814453, + "lpips": 0.3336130380630493 + }, + "overall": { + "mse": 0.019963642582297325, + "psnr": 17.719575881958008, + "lpips": 0.29392296075820923 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000132_test__val_savanna__3__000009/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000132_test__val_savanna__3__000009/case.json new file mode 100644 index 0000000000000000000000000000000000000000..a0af3c17fe597822d590ccd14c153420721e92f9 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_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": 42000132, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.13210037199315, + "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_seed42_job85134/cases/g000132_test__val_savanna__3__000009/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000132_test__val_savanna__3__000009/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..571402971284531a8435c7919bf81bb5587e7386 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000132_test__val_savanna__3__000009/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:24e1a32521f821d181b0413dad7fcab396d15d05d51d4b43126337b593df656a +size 6742 diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000132_test__val_savanna__3__000009/metrics.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000132_test__val_savanna__3__000009/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..11c40dab2c6064c2d1179c9f8f464c2df9969798 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_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": 42000132, + "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": 0.016176737844944, + "psnr": 18.88351058959961, + "lpips": 0.15916186571121216 + }, + "frames_100_200": { + "mse": 0.028227871283888817, + "psnr": 17.636045455932617, + "lpips": 0.16792228817939758 + }, + "frames_200_300": { + "mse": 0.035047128796577454, + "psnr": 14.757153511047363, + "lpips": 0.20405025780200958 + }, + "frames_300_400": { + "mse": 0.04227089881896973, + "psnr": 13.842018127441406, + "lpips": 0.21504321694374084 + }, + "frames_400_500": { + "mse": 0.042886968702077866, + "psnr": 13.762490272521973, + "lpips": 0.24263431131839752 + }, + "overall": { + "mse": 0.0329219214618206, + "psnr": 15.77624225616455, + "lpips": 0.1977623701095581 + } + } +} diff --git a/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000133_test__val_plains__0__000004/case.json b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000133_test__val_plains__0__000004/case.json new file mode 100644 index 0000000000000000000000000000000000000000..f0a8ebfa6164603e3160c87ac1567fa6c6939466 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_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": 42000133, + "native_generated_frames": 500, + "method_key": "decmem", + "master_seed": 42, + "generation_seconds": 225.97933659900445, + "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_seed42_job85134/cases/g000133_test__val_plains__0__000004/frame_metrics.npz b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000133_test__val_plains__0__000004/frame_metrics.npz new file mode 100644 index 0000000000000000000000000000000000000000..f2875a71c1b48296a1a3f35a6890384be6c1be03 --- /dev/null +++ b/results/worldmem-baseline-evals/decmem/decmem_stm6_topk27_minecraft_300cases_seed42_job85134/cases/g000133_test__val_plains__0__000004/frame_metrics.npz @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0924f64a594e5dac4007bbaa13016eba56449de80573387d9d89845e395214a1 +size 6742