anhdao69 commited on
Commit
9257a7e
·
verified ·
1 Parent(s): 04ae52a

Index libero-all-suites/init_variants/core16_21

Browse files
Files changed (2) hide show
  1. README.md +10 -1
  2. checkpoint_index.json +36 -1
README.md CHANGED
@@ -9,7 +9,7 @@ tags:
9
 
10
  # LoopWAM final checkpoints
11
 
12
- Final trained policy weights from the LoopWAM experiments. This repository contains **19 checkpoints**, separated by training data scope. Both original and repeated 4/4 runs are retained. All production checkpoints are from completed ten-epoch runs with global batch 128 and training seed 42.
13
 
14
  ## October 9 additions: all-loop KV
15
 
@@ -73,10 +73,19 @@ The checkpoint completed ten epochs on the four-suite full dataset with global b
73
 
74
  Dense-S12 / Dense-S30 trained on all four suites with the same data, normalization, initialization and schedule as `libero-all-suites/v0-video4-action4` (21,700 updates, global batch 128, seed 42) on two B200 GPUs. Not evaluated.
75
 
 
 
 
 
 
 
 
 
76
  ## Checkpoint index
77
 
78
  | Directory | Model | Video/action loops | Parameters | Final evaluation |
79
  |---|---|---|---:|---|
 
80
  | [libero-all-suites/dense-s30](libero-all-suites/dense-s30) | dense_s30 | 1/1 | 1,416,114,247 | Training only; evaluation not run |
81
  | [libero-all-suites/dense-s12](libero-all-suites/dense-s12) | dense_s12 | 1/1 | 584,536,135 | Training only; evaluation not run |
82
  | [libero-all-suites/v0-video1-action4](libero-all-suites/v0-video1-action4) | v0 aligned | 1/4 | 584536135 | Not evaluated (training only) |
 
9
 
10
  # LoopWAM final checkpoints
11
 
12
+ Final trained policy weights from the LoopWAM experiments. This repository contains **20 checkpoints**, separated by training data scope. Both original and repeated 4/4 runs are retained. All production checkpoints are from completed ten-epoch runs with global batch 128 and training seed 42.
13
 
14
  ## October 9 additions: all-loop KV
15
 
 
73
 
74
  Dense-S12 / Dense-S30 trained on all four suites with the same data, normalization, initialization and schedule as `libero-all-suites/v0-video4-action4` (21,700 updates, global batch 128, seed 42) on two B200 GPUs. Not evaluated.
75
 
76
+ ## October 2026 additions: v0 initialization variants
77
+
78
+ | Directory | Model | Video/action loops | Parameters | Evaluation |
79
+ |---|---|---:|---:|---|
80
+ | [libero-all-suites/init_variants/core16_21](libero-all-suites/init_variants/core16_21) | v0 init `core16_21` | 4/4 | 584,536,135 | Training only; evaluation not run |
81
+
82
+ v0 (video 4 / action 4) retrained on all four suites with the same data, normalization, schedule and precision as `libero-all-suites/v0-video4-action4` (21,700 updates, global batch 128, seed 42) on two B200 GPUs; only the student initialization differs (see each directory's README and `init_variant_manifest.json`). Not evaluated.
83
+
84
  ## Checkpoint index
85
 
86
  | Directory | Model | Video/action loops | Parameters | Final evaluation |
87
  |---|---|---|---:|---|
88
+ | [libero-all-suites/init_variants/core16_21](libero-all-suites/init_variants/core16_21) | v0 init `core16_21` | 4/4 | 584,536,135 | Training only; evaluation not run |
89
  | [libero-all-suites/dense-s30](libero-all-suites/dense-s30) | dense_s30 | 1/1 | 1,416,114,247 | Training only; evaluation not run |
90
  | [libero-all-suites/dense-s12](libero-all-suites/dense-s12) | dense_s12 | 1/1 | 584,536,135 | Training only; evaluation not run |
91
  | [libero-all-suites/v0-video1-action4](libero-all-suites/v0-video1-action4) | v0 aligned | 1/4 | 584536135 | Not evaluated (training only) |
checkpoint_index.json CHANGED
@@ -537,7 +537,42 @@
537
  "gpu": "NVIDIA B200",
538
  "world_size": 2,
539
  "created_utc": "2026-10-26T18:32:27.508039+00:00"
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
540
  }
541
  ],
542
- "updated_utc": "2026-10-26T18:32:27.508039+00:00"
543
  }
 
537
  "gpu": "NVIDIA B200",
538
  "world_size": 2,
539
  "created_utc": "2026-10-26T18:32:27.508039+00:00"
540
+ },
541
+ {
542
+ "id": "v0_core16_21_full_libero_hpg",
543
+ "repo_path": "libero-all-suites/init_variants/core16_21",
544
+ "training_path": "/blue/yu954774.ucf/an221229.ucf/checkpoints/LoopWAM/runs/v0_init_variants/core16_21/train",
545
+ "source_sha256": "e1d4a1951d04b94fc1be93e25507db3bbc2d9b1fa7923495b5f8a89cc3053f15",
546
+ "source_bytes": 7015638231,
547
+ "export_sha256": "7810a43430d3bd2845a216bb7db42d9266aefac741a599f3e9dfbda7453e60b2",
548
+ "export_bytes": 2338711947,
549
+ "format": "loopwam-s-v1",
550
+ "optimizer_included": false,
551
+ "step": 21700,
552
+ "version": "v0",
553
+ "video_loops": 4,
554
+ "action_loops": 4,
555
+ "parameters": 584536135,
556
+ "init_variant": "core16_21",
557
+ "initialization_sha256": "3eb17dcb6e76904ff1dc0d00a224d71028336f65e2887a00a35c57769780bd2d",
558
+ "initialization_mode": "wan_init_variant_core16_21_fresh_optimizer",
559
+ "dataset_scope": "full_libero",
560
+ "epochs": 10,
561
+ "global_batch": 128,
562
+ "training_seed": 42,
563
+ "evaluation_status": "not_run_training_only",
564
+ "success_rate": null,
565
+ "successes": null,
566
+ "episodes": 0,
567
+ "training_seconds": 45906.3459946448,
568
+ "tensor_equality_verified": true,
569
+ "source_revision": "4ebf38a3c65aa5cc9d6c07d94c52f91c51551b57",
570
+ "normalization_sha256": "d5ad351eb3272f416d34082e8af5483ba1118ceca355a37d9f268fdb269fb3e4",
571
+ "fairness_check_passed": true,
572
+ "gpu": "NVIDIA B200",
573
+ "world_size": 2,
574
+ "created_utc": "2026-10-27T15:22:03.001353+00:00"
575
  }
576
  ],
577
+ "updated_utc": "2026-10-27T15:22:03.001353+00:00"
578
  }