Hugging Face's logo

yslan
/
o-intermediate-model

Model card Files
xet
 Community
yslan commited on
Commit
716d97c
·
verified ·
1 Parent(s): 272d988

Add lineage manifest for cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108

Browse files
Files changed (19) hide show
  1. cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/all_source_readback_attempt02.json +31 -0
  2. cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/backup_manifest.json +127 -0
  3. cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/checkpoint_header_metadata.json +12 -0
  4. cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/droid_readback.json +133 -0
  5. cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/experiment_config_module.py +131 -0
  6. cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/first_optimizer_update_receipt.json +18 -0
  7. cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/input_14700_cpu_verified.json +21 -0
  8. cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/launch_plan.json +53 -0
  9. cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/libero_readback.json +194 -0
  10. cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/lineage.json +43 -0
  11. cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/long49_all_source_cpu_gate.json +31 -0
  12. cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/molmo49_readback.json +418 -0
  13. cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/omni4d_vgg_stage_one_next_version_all_sources_plus_hdepic_49_frames_ddp_global32_b4_gas4.yaml +815 -0
  14. cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/pointodyssey49_readback.json +12 -0
  15. cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/robotwin_readback.json +192 -0
  16. cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/source_commit.txt +1 -0
  17. cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/source_windows.yaml +815 -0
  18. cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/source_windows_recipe.yaml +815 -0
  19. cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/syn4d_readback.json +11 -0
cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/all_source_readback_attempt02.json ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "blockers": [],
3
+ "failures": {
4
+ "four_d_pretraining_relaunch_syn4d_sources": "RuntimeError: Cannot read Syn4D index cache /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/syn4d_index/syn4d_index_338563a7462b874f01ea79e31424cb0b.pkl; existing cache preserved. Inspect its source/schema or select an explicit new index_cache_dir.",
5
+ "pointodyssey_train": "ValueError: pointodyssey_train: position_windows keeps no window of the 108 sequences the PointOdysseyTrack3RSparseTrack loader indexed"
6
+ },
7
+ "model_forward_or_training": false,
8
+ "sources_read": [
9
+ "dl3dv",
10
+ "droid",
11
+ "four_d_pretraining_relaunch_dynamic_replica_stored_depth_source",
12
+ "four_d_pretraining_relaunch_midair_within_200_meters_source",
13
+ "four_d_pretraining_relaunch_mvs_synth_1000_then_far_percentile_source",
14
+ "four_d_pretraining_relaunch_omniworld_game_source",
15
+ "four_d_pretraining_relaunch_paralleldomain4d_sky_then_far_percentile_source",
16
+ "four_d_pretraining_relaunch_spring_stored_depth_source",
17
+ "four_d_pretraining_relaunch_synthverse_source",
18
+ "four_d_pretraining_relaunch_tartanair_sky_then_far_percentile_source",
19
+ "four_d_pretraining_relaunch_tartanground_sky_then_far_percentile_source",
20
+ "four_d_pretraining_relaunch_vkitti_source",
21
+ "kubric_train",
22
+ "libero",
23
+ "robotwin",
24
+ "scannet",
25
+ "waymo",
26
+ "wildrgbd"
27
+ ],
28
+ "status": "failed",
29
+ "windows_yaml": "/scratch/shared/engs2907/logs/vgg_forecast_4d/train/cosmos_nano_49_frame_midtraining_all_sources_hdepic_without_cityscapes_from14700_20261107/attempt03/source_windows.yaml",
30
+ "windows_yaml_sha256": "6f4010e7495eafbbbd6a6c74db5433888f3307f8a7381ee863894eeed00bf87e"
31
+ }
cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/backup_manifest.json ADDED
@@ -0,0 +1,127 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "checkpoint": {
3
+ "bytes": 2990807192,
4
+ "filename": "step_014720.safetensors",
5
+ "role": "trainable_checkpoint_only",
6
+ "sha256": "31c7e38a374e2d080f08ea58a413b2e0541c095d83eda97f606688c84eda418f"
7
+ },
8
+ "excluded": [
9
+ "frozen pretrained base weights",
10
+ "optimizer/scheduler/RNG state",
11
+ "15K protected backup"
12
+ ],
13
+ "metadata_files": [
14
+ {
15
+ "bytes": 1675,
16
+ "filename": "all_source_readback_attempt02.json",
17
+ "role": "run_metadata",
18
+ "sha256": "98e9ff00e2915950ff53ecdaf4913edc98e1e8e23cd8c67e6c024c37a7af5acc"
19
+ },
20
+ {
21
+ "bytes": 38530,
22
+ "filename": "checkpoint_header_metadata.json",
23
+ "role": "run_metadata",
24
+ "sha256": "f78a52de393ee803651c553336506594f7cf78a24e63a70baab2bbdb3192d130"
25
+ },
26
+ {
27
+ "bytes": 7880,
28
+ "filename": "droid_readback.json",
29
+ "role": "run_metadata",
30
+ "sha256": "9b37ced309b4145ab50b896eeec430364decfe5a77feaf49e20cd0be0aa33dd3"
31
+ },
32
+ {
33
+ "bytes": 7208,
34
+ "filename": "experiment_config_module.py",
35
+ "role": "run_metadata",
36
+ "sha256": "e497abadf90f4f1ed9235de583ac4ecdd48595ef4225f3dcde71a461821fc99e"
37
+ },
38
+ {
39
+ "bytes": 827,
40
+ "filename": "first_optimizer_update_receipt.json",
41
+ "role": "run_metadata",
42
+ "sha256": "8a1b927f9b6422f42ab0e8d2c4c6351746c28ab55406b9edef81913653129a82"
43
+ },
44
+ {
45
+ "bytes": 39062,
46
+ "filename": "input_14700_cpu_verified.json",
47
+ "role": "run_metadata",
48
+ "sha256": "b2c7fd17d34cdd224d9eab49a978eeb15de55261d245f4eedc5a211a2b64fdbe"
49
+ },
50
+ {
51
+ "bytes": 3435,
52
+ "filename": "launch_plan.json",
53
+ "role": "run_metadata",
54
+ "sha256": "737d3e344e1b38d09e9f50ad183de2d7a8022ed7f6da0f6d29e390944fcab61f"
55
+ },
56
+ {
57
+ "bytes": 3357,
58
+ "filename": "libero_readback.json",
59
+ "role": "run_metadata",
60
+ "sha256": "c8d49b7d87f64b45f1183ab8babc3c17d632e5345e0f697ed4a2aebc6fac4a0f"
61
+ },
62
+ {
63
+ "bytes": 2243,
64
+ "filename": "lineage.json",
65
+ "role": "run_metadata",
66
+ "sha256": "87ae48ef3e12c8b342430f148e9cfab5db857afe1637a9a305c897179c705995"
67
+ },
68
+ {
69
+ "bytes": 1567,
70
+ "filename": "long49_all_source_cpu_gate.json",
71
+ "role": "run_metadata",
72
+ "sha256": "fe8098d6d39d9913c0340bd01e6aca3c23ab6230b6a8235bd48015ec16c305ab"
73
+ },
74
+ {
75
+ "bytes": 9288,
76
+ "filename": "molmo49_readback.json",
77
+ "role": "run_metadata",
78
+ "sha256": "ec0577a78e97db76f14cca0bd4d44ab87bc4e588f615fca242c1d675abb8dbb1"
79
+ },
80
+ {
81
+ "bytes": 41246,
82
+ "filename": "omni4d_vgg_stage_one_next_version_all_sources_plus_hdepic_49_frames_ddp_global32_b4_gas4.yaml",
83
+ "role": "run_metadata",
84
+ "sha256": "ff4f200255a52f69a3f5551d311156258dea1dace5ac4e7af47a6abf266f4ade"
85
+ },
86
+ {
87
+ "bytes": 452,
88
+ "filename": "pointodyssey49_readback.json",
89
+ "role": "run_metadata",
90
+ "sha256": "d35042d9877fd0f5c06d3db64f6c5d021de029c2a592f5887c3e8be5656ba726"
91
+ },
92
+ {
93
+ "bytes": 3291,
94
+ "filename": "robotwin_readback.json",
95
+ "role": "run_metadata",
96
+ "sha256": "785c7dde5f5f69a6a6fe30fe6accf56fe2380ab8903b7c9f2ec625d0501ecf2f"
97
+ },
98
+ {
99
+ "bytes": 41,
100
+ "filename": "source_commit.txt",
101
+ "role": "run_metadata",
102
+ "sha256": "247b5f7227677f9f004d91769bcbdc055fd5d7b90c16779bb3d8b2853597a823"
103
+ },
104
+ {
105
+ "bytes": 41246,
106
+ "filename": "source_windows.yaml",
107
+ "role": "run_metadata",
108
+ "sha256": "ff4f200255a52f69a3f5551d311156258dea1dace5ac4e7af47a6abf266f4ade"
109
+ },
110
+ {
111
+ "bytes": 41246,
112
+ "filename": "source_windows_recipe.yaml",
113
+ "role": "run_metadata",
114
+ "sha256": "ff4f200255a52f69a3f5551d311156258dea1dace5ac4e7af47a6abf266f4ade"
115
+ },
116
+ {
117
+ "bytes": 458,
118
+ "filename": "syn4d_readback.json",
119
+ "role": "run_metadata",
120
+ "sha256": "cc8c4082509fcf2c5e6cb23f50477ab1d2c07151ac34f96cbdbbe345d96f8ac4"
121
+ }
122
+ ],
123
+ "remote_prefix": "cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720",
124
+ "repository": "yslan/o-intermediate-model",
125
+ "run_id": "cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108",
126
+ "schema": "vgg_trainable_checkpoint_hf_backup_manifest_v1"
127
+ }
cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/checkpoint_header_metadata.json ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "file_bytes": 2990807192,
3
+ "metadata": {
4
+ "base_identity": "nvidia/Cosmos3-Nano@e59a53c25979a090fa8706c9acc0c254a6e89b92",
5
+ "config_identity": "cosmos3_nano_omni4d_stage_one_lora_r256_plain_ddp_49_frames_temporal_forcing_all_sources_plus_hdepic_from14700_global32_lr2p5e5",
6
+ "exact_training_resume": "false",
7
+ "format": "omni4d_trainable_weights_v1",
8
+ "parameter_names": "[\"language_model.model.layers.0.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.0.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.0.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.0.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.0.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.0.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.0.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.0.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.0.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.0.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.0.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.0.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.0.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.0.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.1.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.1.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.1.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.1.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.1.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.1.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.1.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.1.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.1.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.1.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.1.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.1.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.1.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.1.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.2.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.2.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.2.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.2.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.2.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.2.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.2.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.2.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.2.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.2.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.2.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.2.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.2.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.2.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.3.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.3.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.3.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.3.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.3.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.3.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.3.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.3.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.3.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.3.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.3.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.3.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.3.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.3.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.4.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.4.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.4.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.4.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.4.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.4.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.4.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.4.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.4.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.4.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.4.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.4.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.4.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.4.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.5.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.5.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.5.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.5.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.5.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.5.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.5.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.5.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.5.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.5.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.5.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.5.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.5.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.5.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.6.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.6.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.6.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.6.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.6.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.6.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.6.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.6.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.6.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.6.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.6.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.6.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.6.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.6.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.7.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.7.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.7.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.7.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.7.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.7.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.7.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.7.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.7.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.7.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.7.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.7.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.7.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.7.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.8.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.8.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.8.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.8.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.8.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.8.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.8.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.8.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.8.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.8.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.8.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.8.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.8.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.8.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.9.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.9.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.9.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.9.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.9.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.9.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.9.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.9.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.9.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.9.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.9.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.9.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.9.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.9.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.10.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.10.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.10.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.10.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.10.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.10.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.10.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.10.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.10.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.10.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.10.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.10.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.10.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.10.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.11.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.11.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.11.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.11.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.11.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.11.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.11.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.11.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.11.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.11.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.11.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.11.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.11.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.11.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.12.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.12.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.12.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.12.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.12.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.12.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.12.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.12.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.12.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.12.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.12.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.12.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.12.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.12.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.13.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.13.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.13.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.13.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.13.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.13.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.13.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.13.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.13.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.13.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.13.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.13.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.13.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.13.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.14.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.14.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.14.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.14.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.14.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.14.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.14.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.14.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.14.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.14.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.14.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.14.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.14.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.14.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.15.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.15.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.15.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.15.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.15.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.15.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.15.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.15.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.15.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.15.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.15.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.15.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.15.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.15.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.16.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.16.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.16.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.16.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.16.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.16.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.16.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.16.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.16.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.16.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.16.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.16.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.16.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.16.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.17.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.17.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.17.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.17.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.17.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.17.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.17.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.17.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.17.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.17.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.17.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.17.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.17.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.17.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.18.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.18.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.18.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.18.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.18.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.18.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.18.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.18.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.18.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.18.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.18.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.18.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.18.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.18.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.19.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.19.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.19.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.19.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.19.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.19.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.19.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.19.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.19.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.19.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.19.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.19.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.19.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.19.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.20.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.20.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.20.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.20.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.20.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.20.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.20.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.20.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.20.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.20.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.20.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.20.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.20.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.20.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.21.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.21.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.21.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.21.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.21.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.21.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.21.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.21.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.21.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.21.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.21.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.21.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.21.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.21.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.22.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.22.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.22.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.22.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.22.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.22.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.22.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.22.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.22.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.22.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.22.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.22.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.22.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.22.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.23.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.23.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.23.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.23.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.23.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.23.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.23.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.23.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.23.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.23.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.23.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.23.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.23.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.23.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.24.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.24.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.24.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.24.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.24.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.24.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.24.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.24.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.24.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.24.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.24.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.24.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.24.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.24.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.25.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.25.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.25.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.25.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.25.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.25.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.25.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.25.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.25.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.25.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.25.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.25.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.25.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.25.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.26.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.26.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.26.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.26.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.26.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.26.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.26.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.26.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.26.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.26.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.26.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.26.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.26.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.26.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.27.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.27.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.27.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.27.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.27.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.27.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.27.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.27.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.27.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.27.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.27.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.27.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.27.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.27.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.28.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.28.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.28.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.28.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.28.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.28.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.28.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.28.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.28.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.28.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.28.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.28.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.28.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.28.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.29.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.29.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.29.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.29.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.29.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.29.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.29.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.29.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.29.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.29.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.29.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.29.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.29.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.29.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.30.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.30.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.30.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.30.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.30.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.30.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.30.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.30.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.30.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.30.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.30.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.30.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.30.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.30.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.31.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.31.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.31.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.31.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.31.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.31.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.31.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.31.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.31.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.31.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.31.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.31.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.31.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.31.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.32.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.32.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.32.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.32.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.32.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.32.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.32.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.32.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.32.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.32.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.32.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.32.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.32.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.32.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.33.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.33.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.33.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.33.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.33.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.33.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.33.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.33.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.33.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.33.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.33.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.33.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.33.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.33.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.34.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.34.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.34.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.34.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.34.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.34.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.34.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.34.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.34.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.34.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.34.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.34.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.34.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.34.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.35.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.35.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.35.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.35.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.35.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.35.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.35.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.35.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.35.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.35.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.35.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.35.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.35.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.35.mlp_moe_gen.down_proj.lora_B.weight\", \"time_embedder.mlp.0.weight\", \"time_embedder.mlp.0.bias\", \"time_embedder.mlp.2.weight\", \"time_embedder.mlp.2.bias\", \"omni4d_stream_projector.input.video.weight\", \"omni4d_stream_projector.input.video.bias\", \"omni4d_stream_projector.input.pointmap.weight\", \"omni4d_stream_projector.input.pointmap.bias\", \"omni4d_stream_projector.input.camera.weight\", \"omni4d_stream_projector.input.camera.bias\", \"omni4d_stream_projector.input.dino.weight\", \"omni4d_stream_projector.input.dino.bias\", \"omni4d_stream_projector.input.track_video.weight\", \"omni4d_stream_projector.input.track_video.bias\", \"omni4d_stream_projector.dino_feature_norm.weight\", \"omni4d_stream_projector.dino_feature_norm.bias\", \"omni4d_stream_projector.output.video.weight\", \"omni4d_stream_projector.output.video.bias\", \"omni4d_stream_projector.output.pointmap.weight\", \"omni4d_stream_projector.output.pointmap.bias\", \"omni4d_stream_projector.output.camera.weight\", \"omni4d_stream_projector.output.camera.bias\", \"omni4d_stream_projector.output.dino.weight\", \"omni4d_stream_projector.output.dino.bias\", \"omni4d_stream_projector.output.track_video.weight\", \"omni4d_stream_projector.output.track_video.bias\"]",
9
+ "update": "14720"
10
+ },
11
+ "tensor_count": 530
12
+ }
cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/droid_readback.json ADDED
@@ -0,0 +1,133 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "caption_coverage": null,
3
+ "dino_feature_shape": [
4
+ 1,
5
+ 3072,
6
+ 13,
7
+ 7,
8
+ 7
9
+ ],
10
+ "frames": 49,
11
+ "model_forward_or_training": false,
12
+ "role_replay": null,
13
+ "rows": [
14
+ {
15
+ "caption_characters": 1512,
16
+ "caption_sha256": "6abec49e92ae0bdf8db04a32e85885bec7f6f0ce168e8ce3d5bb337692cb4a1a",
17
+ "caption_tier": "global",
18
+ "condition_only_streams": [
19
+ "pointmap",
20
+ "camera"
21
+ ],
22
+ "frame_ids": [
23
+ 0,
24
+ 2,
25
+ 4,
26
+ 6,
27
+ 8,
28
+ 10,
29
+ 12,
30
+ 14,
31
+ 16,
32
+ 18,
33
+ 20,
34
+ 22,
35
+ 24,
36
+ 26,
37
+ 28,
38
+ 30,
39
+ 32,
40
+ 34,
41
+ 36,
42
+ 38,
43
+ 40,
44
+ 42,
45
+ 44,
46
+ 46,
47
+ 48,
48
+ 50,
49
+ 52,
50
+ 54,
51
+ 56,
52
+ 58,
53
+ 60,
54
+ 62,
55
+ 64,
56
+ 66,
57
+ 68,
58
+ 70,
59
+ 72,
60
+ 74,
61
+ 76,
62
+ 78,
63
+ 80,
64
+ 82,
65
+ 84,
66
+ 86,
67
+ 88,
68
+ 90,
69
+ 92,
70
+ 94,
71
+ 96
72
+ ],
73
+ "labelled": {
74
+ "camera_valid": true,
75
+ "pointmap_valid": true,
76
+ "tracking_video_valid": true
77
+ },
78
+ "sequence": "{\"canonical_video_id\": \"droid_vp/RPL+32cba90c+2023-12-01-21h-19m-51s/episode_005373\", \"caption_export_episode_index\": 5373, \"caption_source_uuid\": \"RPL+32cba90c+2023-12-01-21h-19m-51s\", \"dataset\": \"droid\", \"exposure\": 0, \"frames\": [\"droid_05373_exterior_1_left/000322\", \"droid_05373_exterior_1_left/000324\", \"droid_05373_exterior_1_left/000326\", \"droid_05373_exterior_1_left/000328\", \"droid_05373_exterior_1_left/000330\", \"droid_05373_exterior_1_left/000332\", \"droid_05373_exterior_1_left/000334\", \"droid_05373_exterior_1_left/000336\", \"droid_05373_exterior_1_left/000338\", \"droid_05373_exterior_1_left/000340\", \"droid_05373_exterior_1_left/000342\", \"droid_05373_exterior_1_left/000344\", \"droid_05373_exterior_1_left/000346\", \"droid_05373_exterior_1_left/000348\", \"droid_05373_exterior_1_left/000350\", \"droid_05373_exterior_1_left/000352\", \"droid_05373_exterior_1_left/000354\", \"droid_05373_exterior_1_left/000356\", \"droid_05373_exterior_1_left/000358\", \"droid_05373_exterior_1_left/000360\", \"droid_05373_exterior_1_left/000362\", \"droid_05373_exterior_1_left/000364\", \"droid_05373_exterior_1_left/000366\", \"droid_05373_exterior_1_left/000368\", \"droid_05373_exterior_1_left/000370\", \"droid_05373_exterior_1_left/000372\", \"droid_05373_exterior_1_left/000374\", \"droid_05373_exterior_1_left/000376\", \"droid_05373_exterior_1_left/000378\", \"droid_05373_exterior_1_left/000380\", \"droid_05373_exterior_1_left/000382\", \"droid_05373_exterior_1_left/000384\", \"droid_05373_exterior_1_left/000386\", \"droid_05373_exterior_1_left/000388\", \"droid_05373_exterior_1_left/000390\", \"droid_05373_exterior_1_left/000392\", \"droid_05373_exterior_1_left/000394\", \"droid_05373_exterior_1_left/000396\", \"droid_05373_exterior_1_left/000398\", \"droid_05373_exterior_1_left/000400\", \"droid_05373_exterior_1_left/000402\", \"droid_05373_exterior_1_left/000404\", \"droid_05373_exterior_1_left/000406\", \"droid_05373_exterior_1_left/000408\", \"droid_05373_exterior_1_left/000410\", \"droid_05373_exterior_1_left/000412\", \"droid_05373_exterior_1_left/000414\", \"droid_05373_exterior_1_left/000416\", \"droid_05373_exterior_1_left/000418\"], \"labels\": [\"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000322\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000324\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000326\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000328\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000330\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000332\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000334\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000336\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000338\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000340\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000342\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000344\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000346\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000348\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000350\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000352\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000354\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000356\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000358\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000360\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000362\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000364\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000366\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000368\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000370\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000372\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000374\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000376\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000378\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000380\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000382\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000384\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000386\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000388\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000390\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000392\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000394\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000396\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000398\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000400\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000402\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000404\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000406\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000408\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000410\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000412\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000414\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000416\", \"RPL+32cba90c+2023-12-01-21h-19m-51s_exterior_1_left_000418\"], \"source_ego_id\": \"RPL+32cba90c+2023-12-01-21h-19m-51s\", \"vpm_source_episode_index\": 45638}",
79
+ "streams": {
80
+ "pointmap_image": {
81
+ "dtype": "torch.float32",
82
+ "finite": true,
83
+ "max": -0.1592797040939331,
84
+ "min": -0.819167971611023,
85
+ "shape": [
86
+ 49,
87
+ 3,
88
+ 288,
89
+ 512
90
+ ]
91
+ },
92
+ "raymap_camera_image": {
93
+ "dtype": "torch.float32",
94
+ "finite": true,
95
+ "max": 0.9772403836250305,
96
+ "min": -0.7611461877822876,
97
+ "shape": [
98
+ 49,
99
+ 3,
100
+ 288,
101
+ 512
102
+ ]
103
+ },
104
+ "rgb": {
105
+ "dtype": "torch.uint8",
106
+ "nonzero_count": 21307817,
107
+ "shape": [
108
+ 49,
109
+ 3,
110
+ 288,
111
+ 512
112
+ ]
113
+ },
114
+ "tracking_video": {
115
+ "dtype": "torch.uint8",
116
+ "nonzero_count": 166431,
117
+ "shape": [
118
+ 49,
119
+ 3,
120
+ 288,
121
+ 512
122
+ ]
123
+ }
124
+ }
125
+ }
126
+ ],
127
+ "source": "droid",
128
+ "status": "passed",
129
+ "vgg_commit": "f6992c4b67441dc9778642a252447a2687856116",
130
+ "window_identity": null,
131
+ "windows": 1,
132
+ "windows_yaml_sha256": "f0b5f4dc89dc4face44830c3b16eb1f0699ca8a1d27a5fb9a98bcb06502064a4"
133
+ }
cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/experiment_config_module.py ADDED
@@ -0,0 +1,131 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
2
+ # SPDX-License-Identifier: OpenMDW-1.1
3
+
4
+ """49-frame all-source midtraining from the verified update-14,700 weights-only overlay."""
5
+
6
+ import hashlib
7
+ import os
8
+ from copy import deepcopy
9
+ from pathlib import Path
10
+
11
+ from hydra.core.config_store import ConfigStore
12
+
13
+ from cosmos_framework.configs.base.experiment.sft.omni4d_stage_one_lora_r256_next_version_all_sources_21_frames_temporal_forcing_plain_ddp_from_8200 import (
14
+ WORLD_SIZE,
15
+ recipe as base_recipe,
16
+ )
17
+ from cosmos_framework.inference.omni4d_overlay_config_identities import (
18
+ CITYSCAPES_AND_MOLMO_FROM_OVERLAY_CONFIG_IDENTITY,
19
+ FORTY_NINE_FRAME_FROM_14700_CONFIG_IDENTITY,
20
+ )
21
+
22
+ SOURCE_FRAMES = 49
23
+ START_UPDATE = 14700
24
+ END_UPDATE = 19700
25
+ MICROBATCH_PER_RANK = 4
26
+ GRAD_ACCUM_ITER = 4
27
+ DEFAULT_WINDOWS_YAML = Path(__file__).resolve().parents[4] / "configs/omni4d_vgg_stage_one_next_version_all_sources_plus_hdepic_49_frames_ddp_global32_b4_gas4.yaml"
28
+ WINDOWS_YAML = Path(os.environ.get("COSMOS_VGG_WINDOWS_CONFIG", str(DEFAULT_WINDOWS_YAML)))
29
+ START_OVERLAY = "/scratch/shared/engs2907/logs/vgg_forecast_4d/train/cosmos_next_version_cityscapes_molmo_22_sources_main_process_loader_20261106/attempt02/trainable_overlays/step_014700.safetensors"
30
+ START_OVERLAY_SHA256 = "ae92b4b891df0a6f44e22027139f48f8f6537a43f78a78f92c54a589853209fb"
31
+ START_OVERLAY_CONFIG_IDENTITY = CITYSCAPES_AND_MOLMO_FROM_OVERLAY_CONFIG_IDENTITY
32
+ OUTPUT = "/scratch/shared/engs2907/logs/vgg_forecast_4d/train/cosmos_nano_49_frame_midtraining_all_sources_hdepic_without_cityscapes_from14700_20261107/attempt01"
33
+ WRITER_CONFIG_IDENTITY = FORTY_NINE_FRAME_FROM_14700_CONFIG_IDENTITY
34
+
35
+ recipe = deepcopy(base_recipe)
36
+ recipe.job.name = "cosmos_nano_49_frame_midtraining_all_sources_hdepic_without_cityscapes_from14700_20261107"
37
+ recipe.trainer.max_iter = END_UPDATE
38
+ recipe.optimizer.lr = 2.5e-5
39
+ recipe.model.model_overrides.omni4d_training_recipe.source_frames = SOURCE_FRAMES
40
+ recipe.trainer.grad_accum_iter = GRAD_ACCUM_ITER
41
+ recipe.dataloader_train.batch_size = MICROBATCH_PER_RANK
42
+ recipe.dataloader_train.num_workers = 12
43
+ recipe.dataloader_train.prefetch_factor = 1
44
+ recipe.dataloader_train.dataset.windows_config_path = str(WINDOWS_YAML)
45
+ recipe.dataloader_train.dataset.expected_windows_config_sha256 = hashlib.sha256(WINDOWS_YAML.read_bytes()).hexdigest()
46
+ recipe.dataloader_train.dataset.expected_vgg_commit = "f6992c4b67441dc9778642a252447a2687856116"
47
+ recipe.dataloader_train.dataset.expected_source_frames = SOURCE_FRAMES
48
+ recipe.dataloader_train.dataset.expected_world_size = WORLD_SIZE
49
+ recipe.dataloader_train.dataset.sample_offset_updates = START_UPDATE
50
+ recipe.dataloader_train.dataset.batch_size = MICROBATCH_PER_RANK
51
+ recipe.dataloader_train.dataset.gradient_accumulation_steps = GRAD_ACCUM_ITER
52
+ recipe.model.trainable_overlay_path = START_OVERLAY
53
+ recipe.model.trainable_overlay_config_identity = START_OVERLAY_CONFIG_IDENTITY
54
+ recipe.model.trainable_overlay_base_identity = "nvidia/Cosmos3-Nano@e59a53c25979a090fa8706c9acc0c254a6e89b92"
55
+ recipe.checkpoint.load_training_state = False
56
+ recipe.checkpoint.only_load_scheduler_state = False
57
+ recipe.checkpoint.model_only_overlay_clock_handoff = True
58
+ recipe.checkpoint.model_only_overlay_start_iteration = START_UPDATE
59
+ recipe.trainer.callbacks.omni4d_trainable_overlay.output_dir = f"{OUTPUT}/trainable_overlays"
60
+ recipe.trainer.callbacks.omni4d_trainable_overlay.config_identity = WRITER_CONFIG_IDENTITY
61
+ recipe.trainer.callbacks.omni4d_trainable_overlay.save_updates = sorted(
62
+ {START_UPDATE + 20, *range(START_UPDATE + 100, END_UPDATE + 1, 100)}
63
+ )
64
+ recipe.trainer.callbacks.omni4d_trainable_overlay.keep_last = 1_000_000
65
+ recipe.trainer.callbacks.omni4d_trainable_overlay.milestone_interval = 1000
66
+ recipe.trainer.callbacks.iter_speed.native_phase_timing = {
67
+ "output_dir": f"{OUTPUT}/native_phase_timings",
68
+ "start_after_updates": 1,
69
+ "num_updates": 50,
70
+ "max_pending": 32,
71
+ "max_age_updates": 8,
72
+ }
73
+
74
+
75
+ def require_launch_ready() -> None:
76
+ import yaml
77
+
78
+ from cosmos_framework.data.omni4d_next_version_readiness import check_mixture_arithmetic, launch_blockers
79
+ from cosmos_framework.data.omni4d_window_contract import validate_omni4d_window_shape
80
+
81
+ if not Path(START_OVERLAY).is_file():
82
+ raise RuntimeError(f"the 14,700 input overlay is missing: {START_OVERLAY}")
83
+ digest = hashlib.sha256()
84
+ with Path(START_OVERLAY).open("rb") as handle:
85
+ for chunk in iter(lambda: handle.read(8 * 1024 * 1024), b""):
86
+ digest.update(chunk)
87
+ if digest.hexdigest() != START_OVERLAY_SHA256:
88
+ raise RuntimeError("the 14,700 input overlay SHA-256 differs from the verified checkpoint")
89
+ from safetensors import safe_open
90
+
91
+ with safe_open(START_OVERLAY, framework="pt", device="cpu") as checkpoint:
92
+ metadata = checkpoint.metadata() or {}
93
+ expected_metadata = {
94
+ "format": "omni4d_trainable_weights_v1",
95
+ "config_identity": START_OVERLAY_CONFIG_IDENTITY,
96
+ "base_identity": recipe.model.trainable_overlay_base_identity,
97
+ "update": str(START_UPDATE),
98
+ "exact_training_resume": "false",
99
+ }
100
+ for key, expected in expected_metadata.items():
101
+ if metadata.get(key) != expected:
102
+ raise RuntimeError(f"the 14,700 overlay metadata {key!r} is {metadata.get(key)!r}, expected {expected!r}")
103
+ payload = yaml.safe_load(WINDOWS_YAML.read_text(encoding="utf-8"))
104
+ validate_omni4d_window_shape(SOURCE_FRAMES, *payload["window_block"]["raster_hw"])
105
+ if payload["window_block"]["frames"] != SOURCE_FRAMES:
106
+ raise RuntimeError("the 49-frame model and pinned source window length differ")
107
+ if payload.get("real_video_sources"):
108
+ raise RuntimeError("Cityscapes and other City real-video sources are excluded from this 49-frame mixture")
109
+ robot_names = {item["name"] for item in payload.get("robot_sources", [])}
110
+ if robot_names != {"robotwin", "libero"}:
111
+ raise RuntimeError(f"the 49-frame recipe needs RoboTwin and LIBERO robot sources, got {robot_names}")
112
+ if "droid" not in {item["name"] for item in payload["window_block"]["sources"]}:
113
+ raise RuntimeError("the 49-frame VGG source registry must retain DROID")
114
+ molmo = payload.get("molmo_motion_sources", [])
115
+ if len(molmo) != 1 or set(molmo[0]["reader"]["subsets"]) != {"egodex", "xperience", "stereo4d", "hepic"}:
116
+ raise RuntimeError("the 49-frame recipe must include HD-EPIC with the three already admitted human-video subsets")
117
+ check_mixture_arithmetic(payload)
118
+ blockers = launch_blockers(payload)
119
+ if blockers:
120
+ raise RuntimeError("the 49-frame mixture is not ready to launch:\n" + "\n".join(f"- {item}" for item in blockers))
121
+ actual_yaml_sha = hashlib.sha256(WINDOWS_YAML.read_bytes()).hexdigest()
122
+ if recipe.dataloader_train.dataset.expected_windows_config_sha256 != actual_yaml_sha:
123
+ raise RuntimeError("the composed window config digest changed after the module was imported")
124
+
125
+
126
+ ConfigStore.instance().store(
127
+ group="experiment",
128
+ package="_global_",
129
+ name="omni4d_stage_one_lora_r256_next_version_all_sources_plus_hdepic_49_frames_plain_ddp_from_14700",
130
+ node=recipe,
131
+ )
cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/first_optimizer_update_receipt.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "status": "optimizer_updates_verified",
3
+ "observed_at_utc": "2026-11-08T01:01:00.330032+00:00",
4
+ "job_step": "73012.9",
5
+ "source_commit": "9cb0ceb0167ba45dc89abfe29ae740a59e91ac11",
6
+ "first_new_update": 14701,
7
+ "first_new_update_utc": "2026-11-08T00:59:04Z",
8
+ "second_new_update": 14702,
9
+ "second_new_update_utc": "2026-11-08T00:59:55Z",
10
+ "first_update_seconds": 380.97,
11
+ "second_update_seconds": 50.42,
12
+ "cold_and_steady_not_conflated": true,
13
+ "losses_finite": true,
14
+ "model_quality_claim": false,
15
+ "last_gpu_telemetry": "2026-11-08T01:00:12Z: UUID82bebc16...100%/78402MiB; UUID6f4e57c4...89%/77296MiB",
16
+ "source_CPU_gate": "/scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_nano_two_training_restart_20261107/attempt01/long49_all_source_cpu_gate_for_attempt04.json",
17
+ "first_save_pending": 14720
18
+ }
cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/input_14700_cpu_verified.json ADDED
@@ -0,0 +1,21 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "path": "/scratch/shared/engs2907/logs/vgg_forecast_4d/train/cosmos_next_version_cityscapes_molmo_22_sources_main_process_loader_20261106/attempt02/trainable_overlays/step_014700.safetensors",
3
+ "sha256": "ae92b4b891df0a6f44e22027139f48f8f6537a43f78a78f92c54a589853209fb",
4
+ "stat": {
5
+ "size": 2990807208,
6
+ "mtime_ns": 1794051940677173078
7
+ },
8
+ "metadata": {
9
+ "parameter_names": "[\"language_model.model.layers.0.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.0.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.0.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.0.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.0.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.0.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.0.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.0.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.0.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.0.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.0.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.0.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.0.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.0.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.1.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.1.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.1.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.1.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.1.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.1.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.1.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.1.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.1.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.1.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.1.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.1.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.1.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.1.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.2.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.2.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.2.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.2.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.2.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.2.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.2.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.2.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.2.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.2.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.2.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.2.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.2.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.2.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.3.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.3.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.3.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.3.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.3.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.3.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.3.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.3.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.3.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.3.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.3.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.3.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.3.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.3.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.4.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.4.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.4.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.4.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.4.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.4.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.4.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.4.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.4.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.4.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.4.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.4.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.4.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.4.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.5.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.5.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.5.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.5.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.5.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.5.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.5.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.5.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.5.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.5.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.5.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.5.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.5.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.5.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.6.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.6.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.6.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.6.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.6.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.6.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.6.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.6.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.6.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.6.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.6.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.6.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.6.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.6.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.7.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.7.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.7.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.7.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.7.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.7.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.7.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.7.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.7.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.7.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.7.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.7.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.7.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.7.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.8.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.8.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.8.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.8.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.8.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.8.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.8.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.8.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.8.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.8.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.8.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.8.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.8.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.8.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.9.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.9.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.9.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.9.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.9.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.9.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.9.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.9.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.9.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.9.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.9.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.9.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.9.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.9.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.10.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.10.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.10.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.10.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.10.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.10.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.10.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.10.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.10.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.10.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.10.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.10.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.10.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.10.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.11.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.11.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.11.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.11.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.11.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.11.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.11.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.11.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.11.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.11.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.11.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.11.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.11.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.11.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.12.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.12.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.12.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.12.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.12.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.12.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.12.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.12.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.12.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.12.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.12.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.12.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.12.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.12.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.13.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.13.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.13.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.13.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.13.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.13.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.13.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.13.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.13.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.13.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.13.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.13.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.13.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.13.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.14.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.14.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.14.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.14.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.14.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.14.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.14.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.14.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.14.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.14.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.14.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.14.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.14.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.14.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.15.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.15.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.15.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.15.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.15.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.15.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.15.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.15.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.15.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.15.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.15.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.15.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.15.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.15.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.16.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.16.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.16.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.16.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.16.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.16.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.16.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.16.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.16.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.16.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.16.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.16.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.16.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.16.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.17.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.17.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.17.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.17.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.17.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.17.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.17.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.17.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.17.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.17.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.17.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.17.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.17.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.17.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.18.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.18.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.18.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.18.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.18.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.18.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.18.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.18.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.18.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.18.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.18.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.18.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.18.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.18.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.19.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.19.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.19.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.19.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.19.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.19.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.19.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.19.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.19.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.19.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.19.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.19.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.19.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.19.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.20.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.20.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.20.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.20.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.20.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.20.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.20.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.20.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.20.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.20.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.20.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.20.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.20.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.20.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.21.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.21.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.21.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.21.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.21.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.21.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.21.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.21.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.21.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.21.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.21.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.21.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.21.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.21.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.22.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.22.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.22.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.22.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.22.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.22.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.22.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.22.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.22.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.22.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.22.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.22.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.22.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.22.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.23.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.23.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.23.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.23.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.23.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.23.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.23.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.23.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.23.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.23.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.23.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.23.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.23.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.23.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.24.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.24.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.24.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.24.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.24.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.24.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.24.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.24.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.24.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.24.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.24.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.24.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.24.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.24.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.25.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.25.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.25.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.25.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.25.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.25.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.25.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.25.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.25.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.25.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.25.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.25.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.25.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.25.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.26.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.26.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.26.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.26.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.26.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.26.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.26.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.26.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.26.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.26.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.26.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.26.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.26.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.26.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.27.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.27.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.27.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.27.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.27.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.27.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.27.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.27.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.27.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.27.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.27.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.27.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.27.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.27.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.28.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.28.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.28.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.28.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.28.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.28.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.28.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.28.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.28.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.28.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.28.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.28.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.28.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.28.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.29.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.29.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.29.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.29.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.29.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.29.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.29.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.29.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.29.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.29.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.29.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.29.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.29.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.29.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.30.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.30.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.30.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.30.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.30.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.30.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.30.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.30.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.30.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.30.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.30.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.30.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.30.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.30.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.31.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.31.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.31.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.31.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.31.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.31.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.31.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.31.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.31.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.31.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.31.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.31.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.31.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.31.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.32.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.32.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.32.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.32.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.32.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.32.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.32.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.32.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.32.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.32.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.32.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.32.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.32.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.32.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.33.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.33.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.33.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.33.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.33.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.33.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.33.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.33.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.33.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.33.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.33.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.33.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.33.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.33.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.34.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.34.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.34.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.34.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.34.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.34.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.34.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.34.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.34.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.34.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.34.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.34.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.34.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.34.mlp_moe_gen.down_proj.lora_B.weight\", \"language_model.model.layers.35.self_attn.q_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.35.self_attn.q_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.35.self_attn.k_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.35.self_attn.k_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.35.self_attn.v_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.35.self_attn.v_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.35.self_attn.o_proj_moe_gen.lora_A.weight\", \"language_model.model.layers.35.self_attn.o_proj_moe_gen.lora_B.weight\", \"language_model.model.layers.35.mlp_moe_gen.gate_proj.lora_A.weight\", \"language_model.model.layers.35.mlp_moe_gen.gate_proj.lora_B.weight\", \"language_model.model.layers.35.mlp_moe_gen.up_proj.lora_A.weight\", \"language_model.model.layers.35.mlp_moe_gen.up_proj.lora_B.weight\", \"language_model.model.layers.35.mlp_moe_gen.down_proj.lora_A.weight\", \"language_model.model.layers.35.mlp_moe_gen.down_proj.lora_B.weight\", \"time_embedder.mlp.0.weight\", \"time_embedder.mlp.0.bias\", \"time_embedder.mlp.2.weight\", \"time_embedder.mlp.2.bias\", \"omni4d_stream_projector.input.video.weight\", \"omni4d_stream_projector.input.video.bias\", \"omni4d_stream_projector.input.pointmap.weight\", \"omni4d_stream_projector.input.pointmap.bias\", \"omni4d_stream_projector.input.camera.weight\", \"omni4d_stream_projector.input.camera.bias\", \"omni4d_stream_projector.input.dino.weight\", \"omni4d_stream_projector.input.dino.bias\", \"omni4d_stream_projector.input.track_video.weight\", \"omni4d_stream_projector.input.track_video.bias\", \"omni4d_stream_projector.dino_feature_norm.weight\", \"omni4d_stream_projector.dino_feature_norm.bias\", \"omni4d_stream_projector.output.video.weight\", \"omni4d_stream_projector.output.video.bias\", \"omni4d_stream_projector.output.pointmap.weight\", \"omni4d_stream_projector.output.pointmap.bias\", \"omni4d_stream_projector.output.camera.weight\", \"omni4d_stream_projector.output.camera.bias\", \"omni4d_stream_projector.output.dino.weight\", \"omni4d_stream_projector.output.dino.bias\", \"omni4d_stream_projector.output.track_video.weight\", \"omni4d_stream_projector.output.track_video.bias\"]",
10
+ "base_identity": "nvidia/Cosmos3-Nano@e59a53c25979a090fa8706c9acc0c254a6e89b92",
11
+ "format": "omni4d_trainable_weights_v1",
12
+ "config_identity": "cosmos3_nano_omni4d_stage_one_lora_r256_plain_ddp_21_frames_temporal_forcing_full20_plus_cityscapes_and_molmo_motion_from_overlay_global32_lr2p5e5",
13
+ "update": "14700",
14
+ "exact_training_resume": "false"
15
+ },
16
+ "tensor_count": 530,
17
+ "elapsed_seconds": 2.951379560981877,
18
+ "node": "civo-a100-1",
19
+ "gpu_used": false,
20
+ "purpose": "Single streamed input verification shared by both owner-authorized formal continuations"
21
+ }
cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/launch_plan.json ADDED
@@ -0,0 +1,53 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "source_root": "/scratch/shared/engs2907/cosmos_long49_hdepic_20261107",
3
+ "source_commit": "9cb0ceb0167ba45dc89abfe29ae740a59e91ac11",
4
+ "attempt_root": "/scratch/shared/engs2907/logs/vgg_forecast_4d/train/cosmos_nano_49_frame_midtraining_all_sources_hdepic_without_cityscapes_from14700_20261107/attempt04",
5
+ "experiment_module": "cosmos_framework.configs.base.experiment.sft.omni4d_stage_one_lora_r256_next_version_all_sources_plus_hdepic_49_frames_plain_ddp_from_14700",
6
+ "windows_config": "/scratch/shared/engs2907/cosmos_long49_hdepic_20261107/cosmos_framework/configs/omni4d_vgg_stage_one_next_version_all_sources_plus_hdepic_49_frames_ddp_global32_b4_gas4.yaml",
7
+ "windows_sha256": "ff4f200255a52f69a3f5551d311156258dea1dace5ac4e7af47a6abf266f4ade",
8
+ "overlay_path": "/scratch/shared/engs2907/logs/vgg_forecast_4d/train/cosmos_next_version_cityscapes_molmo_22_sources_main_process_loader_20261106/attempt02/trainable_overlays/step_014700.safetensors",
9
+ "overlay_sha256": "ae92b4b891df0a6f44e22027139f48f8f6537a43f78a78f92c54a589853209fb",
10
+ "overlay_stat": {
11
+ "size": 2990807208,
12
+ "mtime_ns": 1794051940677173078
13
+ },
14
+ "overlay_config_identity": "cosmos3_nano_omni4d_stage_one_lora_r256_plain_ddp_21_frames_temporal_forcing_full20_plus_cityscapes_and_molmo_motion_from_overlay_global32_lr2p5e5",
15
+ "input_update": 14700,
16
+ "target_update": 19700,
17
+ "input_verification": "/scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_nano_two_training_restart_20261107/attempt01/input_14700_cpu_verified.json",
18
+ "allocation": 73012,
19
+ "job_name": "cosmos_nano_49frame_midtraining",
20
+ "slurm_output": "/scratch/shared/engs2907/logs/vgg_forecast_4d/slurm/train/cosmos_nano_49_frame_midtraining_all_sources_hdepic_without_cityscapes_from14700_20261107/attempt04/slurm-%j-%s.out",
21
+ "experiment_id": "cosmos_nano_49_frame_midtraining_all_sources_hdepic_without_cityscapes_from14700_20261107_attempt04",
22
+ "cpu_preflight_status": "PASS",
23
+ "cpu_evidence": "/scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_nano_two_training_restart_20261107/attempt01/long49_all_source_cpu_gate_for_attempt04.json",
24
+ "compute_script": "/scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_nano_two_training_restart_20261107/attempt01/long49_compute_attempt04.sh",
25
+ "compute_script_sha256": "80efc03508d83ea0703ff69b00bc1a9963a8a3203e1d27f44560c737e43e15a0",
26
+ "owner_authorization": "Ownerauthorized49formaltrainingandfix/relaunchiffailed; onlyexplicitPointOdyssey49temporalstride1neededtocoverauthentic96framecaptions. Preserveothermixtureandsourceweights,exceptalreadyauthorizedCityscapesremovalandHDEpicaddition.",
27
+ "recipe": {
28
+ "source_frames": 49,
29
+ "latent_frames": 13,
30
+ "world_size": 2,
31
+ "batch_per_rank": 4,
32
+ "gradient_accumulation": 4,
33
+ "global_batch": 32,
34
+ "lr": 2.5e-05,
35
+ "source_count": 21,
36
+ "total_weight": 139,
37
+ "cityscapes_included": false,
38
+ "hdepic_included": true,
39
+ "robot_sources": [
40
+ "robotwin",
41
+ "droid",
42
+ "libero"
43
+ ],
44
+ "action_stream": false,
45
+ "omega_stream": false,
46
+ "optimizer": "fresh weights-only continuation",
47
+ "checkpoint_retention": "all saves retained for this5000-update run",
48
+ "pointodyssey_49_temporal_stride_choices": [
49
+ 1
50
+ ]
51
+ },
52
+ "retry_reason": "Attempt03failedPointOdyssey49windowfeasibilitywithstride2:48*2=96 exceeds95-positioncaptionrange. New49-onlystride1preservescoveringcaptionandheldoutids."
53
+ }
cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/libero_readback.json ADDED
@@ -0,0 +1,194 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "caption_coverage": null,
3
+ "dino_feature_shape": [
4
+ 2,
5
+ 3072,
6
+ 13,
7
+ 7,
8
+ 7
9
+ ],
10
+ "frames": 49,
11
+ "model_forward_or_training": false,
12
+ "role_replay": null,
13
+ "rows": [
14
+ {
15
+ "caption_characters": 153,
16
+ "caption_sha256": "e4d688bf20eee14b533476f8e8379de245476461dcf1d7a5e19415489f8fbeb9",
17
+ "caption_tier": "source_task_instruction",
18
+ "condition_only_streams": [],
19
+ "frame_ids": [
20
+ 0,
21
+ 4,
22
+ 8,
23
+ 12,
24
+ 16,
25
+ 20,
26
+ 24,
27
+ 28,
28
+ 32,
29
+ 36,
30
+ 40,
31
+ 44,
32
+ 48,
33
+ 52,
34
+ 56,
35
+ 60,
36
+ 64,
37
+ 68,
38
+ 72,
39
+ 76,
40
+ 80,
41
+ 84,
42
+ 88,
43
+ 92,
44
+ 96,
45
+ 100,
46
+ 104,
47
+ 108,
48
+ 112,
49
+ 116,
50
+ 120,
51
+ 124,
52
+ 128,
53
+ 132,
54
+ 136,
55
+ 140,
56
+ 144,
57
+ 148,
58
+ 152,
59
+ 156,
60
+ 160,
61
+ 164,
62
+ 168,
63
+ 172,
64
+ 176,
65
+ 180,
66
+ 184,
67
+ 188,
68
+ 192
69
+ ],
70
+ "labelled": {
71
+ "camera_valid": true,
72
+ "pointmap_valid": true,
73
+ "tracking_video_valid": true
74
+ },
75
+ "sequence": "libero/libero_spatial/episode_403",
76
+ "streams": {
77
+ "pointmap_image": {
78
+ "dtype": "torch.float32",
79
+ "finite": true,
80
+ "max": 1.0,
81
+ "min": -0.8642057776451111,
82
+ "shape": [
83
+ 49,
84
+ 3,
85
+ 288,
86
+ 512
87
+ ]
88
+ },
89
+ "raymap_camera_image": {
90
+ "dtype": "torch.float32",
91
+ "finite": true,
92
+ "max": 1.0,
93
+ "min": -1.0,
94
+ "shape": [
95
+ 49,
96
+ 3,
97
+ 288,
98
+ 512
99
+ ]
100
+ },
101
+ "rgb": {
102
+ "dtype": "torch.uint8",
103
+ "nonzero_count": 8467200,
104
+ "shape": [
105
+ 49,
106
+ 3,
107
+ 288,
108
+ 512
109
+ ]
110
+ },
111
+ "tracking_video": {
112
+ "dtype": "torch.uint8",
113
+ "nonzero_count": 176490,
114
+ "shape": [
115
+ 49,
116
+ 3,
117
+ 288,
118
+ 512
119
+ ]
120
+ }
121
+ }
122
+ }
123
+ ],
124
+ "source": "libero",
125
+ "status": "passed",
126
+ "vgg_commit": "f6992c4b67441dc9778642a252447a2687856116",
127
+ "window_identity": [
128
+ {
129
+ "episode": 403,
130
+ "frame_ids": [
131
+ 0,
132
+ 4,
133
+ 8,
134
+ 12,
135
+ 16,
136
+ 20,
137
+ 24,
138
+ 28,
139
+ 32,
140
+ 36,
141
+ 40,
142
+ 44,
143
+ 48,
144
+ 52,
145
+ 56,
146
+ 60,
147
+ 64,
148
+ 68,
149
+ 72,
150
+ 76,
151
+ 80,
152
+ 84,
153
+ 88,
154
+ 92,
155
+ 96,
156
+ 100,
157
+ 104,
158
+ 108,
159
+ 112,
160
+ 116,
161
+ 120,
162
+ 124,
163
+ 128,
164
+ 132,
165
+ 136,
166
+ 140,
167
+ 144,
168
+ 148,
169
+ 152,
170
+ 156,
171
+ 160,
172
+ 164,
173
+ 168,
174
+ 172,
175
+ 176,
176
+ 180,
177
+ 184,
178
+ 188,
179
+ 192
180
+ ],
181
+ "native_index": 48987,
182
+ "requested_window": 0,
183
+ "sequence": "libero/libero_spatial/episode_403",
184
+ "underlying_reported": {
185
+ "requested_native_index": 48987,
186
+ "window_episode": 403,
187
+ "window_start": 0
188
+ },
189
+ "window_start": 0
190
+ }
191
+ ],
192
+ "windows": 1,
193
+ "windows_yaml_sha256": "f0b5f4dc89dc4face44830c3b16eb1f0699ca8a1d27a5fb9a98bcb06502064a4"
194
+ }
cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/lineage.json ADDED
@@ -0,0 +1,43 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema": "cosmos_nano_trainable_overlay_lineage_v1",
3
+ "run_id": "cosmos_nano_49_frame_midtraining_all_sources_hdepic_without_cityscapes_from14700_20261107_attempt04",
4
+ "checkpoint": "/scratch/shared/engs2907/logs/vgg_forecast_4d/train/cosmos_nano_49_frame_midtraining_all_sources_hdepic_without_cityscapes_from14700_20261107/attempt04/trainable_overlays/step_014720.safetensors",
5
+ "update": 14720,
6
+ "source_commit": "9cb0ceb0167ba45dc89abfe29ae740a59e91ac11",
7
+ "source_archive_sha256": "0c427fd418c4c1db71e004abac03e9822c5aaf72bf3d1ae7abe1ebe3a556172b",
8
+ "base_identity": "nvidia/Cosmos3-Nano@e59a53c25979a090fa8706c9acc0c254a6e89b92",
9
+ "input_overlay": {
10
+ "update": 14700,
11
+ "path": "/scratch/shared/engs2907/logs/vgg_forecast_4d/train/cosmos_next_version_cityscapes_molmo_22_sources_main_process_loader_20261106/attempt02/trainable_overlays/step_014700.safetensors",
12
+ "sha256": "ae92b4b891df0a6f44e22027139f48f8f6537a43f78a78f92c54a589853209fb",
13
+ "config_identity": "cosmos3_nano_omni4d_stage_one_lora_r256_plain_ddp_21_frames_temporal_forcing_full20_plus_cityscapes_and_molmo_motion_from_overlay_global32_lr2p5e5",
14
+ "resume_semantics": "trainable-weights-only initialization; fresh optimizer, no exact optimizer/RNG resume"
15
+ },
16
+ "recipe": {
17
+ "source_frames": 49,
18
+ "latent_frames": 13,
19
+ "source_count": 21,
20
+ "total_weight": 139,
21
+ "global_batch": 32,
22
+ "batch_per_rank": 4,
23
+ "gradient_accumulation": 4,
24
+ "learning_rate": 0.000025,
25
+ "cityscapes_included": false,
26
+ "hdepic_included": true,
27
+ "action_stream": false,
28
+ "omega_stream": false,
29
+ "pointodyssey_temporal_stride_choices": [1],
30
+ "checkpoint_retention": "all durable saves retained"
31
+ },
32
+ "dataset_lineage": {
33
+ "source_windows_file": "source_windows.yaml",
34
+ "source_windows_sha256": "ff4f200255a52f69a3f5551d311156258dea1dace5ac4e7af47a6abf266f4ade",
35
+ "readback_gate": "PASS, 21 logical source groups; receipts are bundled",
36
+ "dataset_test_or_validation_claim": false
37
+ },
38
+ "evaluation_lineage": {
39
+ "evaluation_performed": false,
40
+ "note": "No model-quality evaluation is included in this checkpoint backup."
41
+ },
42
+ "exclusions": ["frozen pretrained base weights", "optimizer/scheduler/RNG state"]
43
+ }
cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/long49_all_source_cpu_gate.json ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "status": "PASS",
3
+ "source_commit": "9cb0ceb0167ba45dc89abfe29ae740a59e91ac11",
4
+ "tested_code_commit": "bf0cf4b06db0891baca6244a67d9265828b6537b",
5
+ "logical_source_count": 21,
6
+ "non_molmo_source_count": 20,
7
+ "method": "18 original source readbacks plus production-environment Syn4D retry plus corrected PointOdyssey49 stride1; separate Molmo49 group including HD-EPIC eligibility/direct readback",
8
+ "receipts": [
9
+ {
10
+ "path": "/scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_long49_all_source_readback_20261108/attempt02/results/readback.json",
11
+ "sha256": "98e9ff00e2915950ff53ecdaf4913edc98e1e8e23cd8c67e6c024c37a7af5acc"
12
+ },
13
+ {
14
+ "path": "/scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_long49_all_source_readback_20261108/attempt03_native_syn4d/results/readback.json",
15
+ "sha256": "cc8c4082509fcf2c5e6cb23f50477ab1d2c07151ac34f96cbdbbe345d96f8ac4"
16
+ },
17
+ {
18
+ "path": "/scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_nano_49_pointodyssey_caption_stride_fix_20261108/attempt01/results_attempt02/readback.json",
19
+ "sha256": "d35042d9877fd0f5c06d3db64f6c5d021de029c2a592f5887c3e8be5656ba726"
20
+ },
21
+ {
22
+ "path": "/scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_molmo_49frame_source_readback_20261107/attempt01/native_group_readback/molmo_motion_readback.json",
23
+ "sha256": "ec0577a78e97db76f14cca0bd4d44ab87bc4e588f615fca242c1d675abb8dbb1"
24
+ }
25
+ ],
26
+ "pointodyssey_stride_override": [
27
+ 1
28
+ ],
29
+ "source_weights_changed": false,
30
+ "model_quality_claim": false
31
+ }
cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/molmo49_readback.json ADDED
@@ -0,0 +1,418 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "blockers_of_the_yaml": [
3
+ "molmo_motion: readback receipt /scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_city_molmo_pin_migration_20261106/attempt06/molmo_motion/molmo_motion_readback.json is not a complete real molmo_motion window readback"
4
+ ],
5
+ "coverage": {
6
+ "droid": {
7
+ "drawn_by_this_source": false,
8
+ "fps_of_entries": {
9
+ "15.0": 22363
10
+ },
11
+ "in_training_mix": false,
12
+ "train_entries": 22363
13
+ },
14
+ "egodex": {
15
+ "drawn_by_this_source": true,
16
+ "fps_of_entries": {
17
+ "15.0": 95798
18
+ },
19
+ "in_training_mix": true,
20
+ "index": {
21
+ "drawable": 71,
22
+ "excluded_object": 0,
23
+ "no_caption": 0,
24
+ "records": 1438,
25
+ "too_few_frames": 1367
26
+ },
27
+ "split_census": {
28
+ "entries": 95798,
29
+ "files": 95798,
30
+ "records": 1438,
31
+ "required_files": {
32
+ "tracks": {
33
+ "found": 95798,
34
+ "of": 95798
35
+ },
36
+ "videos": {
37
+ "found": 1228,
38
+ "of": 95798
39
+ }
40
+ },
41
+ "split": "train"
42
+ },
43
+ "train_entries": 95798
44
+ },
45
+ "egodex_hand": {
46
+ "drawn_by_this_source": false,
47
+ "fps_of_entries": {
48
+ "15.0": 196576
49
+ },
50
+ "in_training_mix": false,
51
+ "train_entries": 196576
52
+ },
53
+ "hepic": {
54
+ "drawn_by_this_source": true,
55
+ "fps_of_entries": {
56
+ "15.0": 12109
57
+ },
58
+ "in_training_mix": true,
59
+ "index": {
60
+ "drawable": 5149,
61
+ "excluded_object": 0,
62
+ "no_caption": 0,
63
+ "records": 19706,
64
+ "too_few_frames": 14557
65
+ },
66
+ "split_census": {
67
+ "entries": 12109,
68
+ "files": 12109,
69
+ "records": 19706,
70
+ "required_files": {
71
+ "tracks": {
72
+ "found": 12109,
73
+ "of": 12109
74
+ },
75
+ "videos": {
76
+ "found": 12109,
77
+ "of": 12109
78
+ }
79
+ },
80
+ "split": "train"
81
+ },
82
+ "train_entries": 12109
83
+ },
84
+ "molmospaces": {
85
+ "drawn_by_this_source": false,
86
+ "fps_of_entries": {
87
+ "15.0": 132941
88
+ },
89
+ "in_training_mix": false,
90
+ "train_entries": 132941
91
+ },
92
+ "stereo4d": {
93
+ "drawn_by_this_source": true,
94
+ "fps_of_entries": {
95
+ "30.0": 20803
96
+ },
97
+ "in_training_mix": true,
98
+ "index": {
99
+ "drawable": 11825,
100
+ "excluded_object": 0,
101
+ "no_caption": 0,
102
+ "records": 63089,
103
+ "too_few_frames": 51264
104
+ },
105
+ "split_census": {
106
+ "entries": 20803,
107
+ "files": 20803,
108
+ "records": 63089,
109
+ "required_files": {
110
+ "camera_poses": {
111
+ "found": 20728,
112
+ "of": 20803
113
+ },
114
+ "tracks": {
115
+ "found": 20728,
116
+ "of": 20803
117
+ },
118
+ "videos": {
119
+ "found": 20803,
120
+ "of": 20803
121
+ }
122
+ },
123
+ "split": "train"
124
+ },
125
+ "train_entries": 20803
126
+ },
127
+ "xperience": {
128
+ "drawn_by_this_source": true,
129
+ "fps_of_entries": {
130
+ "10.0": 98921
131
+ },
132
+ "in_training_mix": true,
133
+ "index": {
134
+ "drawable": 138864,
135
+ "excluded_object": 250235,
136
+ "no_caption": 0,
137
+ "records": 473631,
138
+ "too_few_frames": 84532
139
+ },
140
+ "split_census": {
141
+ "entries": 98921,
142
+ "files": 98921,
143
+ "records": 473631,
144
+ "required_files": {
145
+ "camera_poses": {
146
+ "found": 98921,
147
+ "of": 98921
148
+ },
149
+ "tracks": {
150
+ "found": 205696,
151
+ "of": 205696
152
+ },
153
+ "videos": {
154
+ "found": 98921,
155
+ "of": 98921
156
+ }
157
+ },
158
+ "split": "train"
159
+ },
160
+ "train_entries": 98921
161
+ },
162
+ "xperience_hand": {
163
+ "drawn_by_this_source": false,
164
+ "fps_of_entries": {
165
+ "10.0": 87997
166
+ },
167
+ "in_training_mix": false,
168
+ "train_entries": 87997
169
+ },
170
+ "ytvis": {
171
+ "drawn_by_this_source": false,
172
+ "fps_of_entries": {
173
+ "12.0": 330,
174
+ "6.0": 324,
175
+ "8.0": 277
176
+ },
177
+ "in_training_mix": true,
178
+ "train_entries": 931
179
+ }
180
+ },
181
+ "exclusion_counts": {
182
+ "point_motion_bench_clip_ids": 2875,
183
+ "release_test_file_ids": {
184
+ "droid": 1169,
185
+ "egodex": 5046,
186
+ "egodex_hand": 5044,
187
+ "hepic": 637,
188
+ "molmospaces": 7760,
189
+ "stereo4d": 2208,
190
+ "xperience": 5206,
191
+ "xperience_hand": 4625,
192
+ "ytvis": 104
193
+ },
194
+ "test_cohort_records": 120,
195
+ "train_file_ids": {
196
+ "egodex": 95798,
197
+ "hepic": 12109,
198
+ "stereo4d": 20803,
199
+ "xperience": 98921
200
+ }
201
+ },
202
+ "frames": 49,
203
+ "model_forward_or_training": false,
204
+ "reader_digest": "6ca3229f4b89c99d4158ef8286105d950ab57c89161eabe9caa9f552bb89ce84",
205
+ "role_replay": {
206
+ "camera": {
207
+ "window_or_frame_role_counts_absent_given_generated": {
208
+ "0": 299,
209
+ "1": 533,
210
+ "2": 0
211
+ }
212
+ },
213
+ "dino": {
214
+ "window_or_frame_role_counts_absent_given_generated": {
215
+ "0": 0,
216
+ "1": 466,
217
+ "2": 366
218
+ }
219
+ },
220
+ "pointmap": {
221
+ "window_or_frame_role_counts_absent_given_generated": {
222
+ "0": 832,
223
+ "1": 0,
224
+ "2": 0
225
+ }
226
+ },
227
+ "track_video": {
228
+ "window_or_frame_role_counts_absent_given_generated": {
229
+ "0": 0,
230
+ "1": 376,
231
+ "2": 456
232
+ }
233
+ },
234
+ "video": {
235
+ "window_or_frame_role_counts_absent_given_generated": {
236
+ "0": 0,
237
+ "1": 403,
238
+ "2": 429
239
+ }
240
+ }
241
+ },
242
+ "rows": [
243
+ {
244
+ "caption_characters": 46,
245
+ "caption_sha256": "3576d20fd436060b3b994bf3dab0a0bc65d342e21524776b53e963b061063d05",
246
+ "caption_tier": "native_annotation_caption",
247
+ "checks": {
248
+ "camera_valid": true,
249
+ "caption_equals_release_entry": true,
250
+ "depth_absent": true,
251
+ "frames_consecutive": true,
252
+ "no_cuda": true,
253
+ "rgb_equals_independent_decode": true,
254
+ "subset_in_training_mix": true,
255
+ "tracking_video_matches_independent_paint": true,
256
+ "tracking_video_valid": true
257
+ },
258
+ "condition_only_streams": [
259
+ "camera"
260
+ ],
261
+ "frame_ids": [
262
+ 9,
263
+ 10,
264
+ 11,
265
+ 12,
266
+ 13,
267
+ 14,
268
+ 15,
269
+ 16,
270
+ 17,
271
+ 18,
272
+ 19,
273
+ 20,
274
+ 21,
275
+ 22,
276
+ 23,
277
+ 24,
278
+ 25,
279
+ 26,
280
+ 27,
281
+ 28,
282
+ 29,
283
+ 30,
284
+ 31,
285
+ 32,
286
+ 33,
287
+ 34,
288
+ 35,
289
+ 36,
290
+ 37,
291
+ 38,
292
+ 39,
293
+ 40,
294
+ 41,
295
+ 42,
296
+ 43,
297
+ 44,
298
+ 45,
299
+ 46,
300
+ 47,
301
+ 48,
302
+ 49,
303
+ 50,
304
+ 51,
305
+ 52,
306
+ 53,
307
+ 54,
308
+ 55,
309
+ 56,
310
+ 57
311
+ ],
312
+ "labelled": {
313
+ "camera_valid": true,
314
+ "pointmap_valid": false,
315
+ "tracking_video_valid": true
316
+ },
317
+ "projection": {
318
+ "fps": 7.5,
319
+ "frame0_median_px": 0.21846943246718498,
320
+ "kept_tracks": 80,
321
+ "later_cells": 3828,
322
+ "later_median_px": 0.2384225186913857,
323
+ "later_p90_px": 0.33664261176933946,
324
+ "mp4_frames": [
325
+ 18,
326
+ 114
327
+ ],
328
+ "mp4_step": 2.0,
329
+ "points_frame": "world",
330
+ "tracks_total": 99
331
+ },
332
+ "sequence": "egodex/part4_pick_place_food_1006/object",
333
+ "streams": {
334
+ "pointmap_image": {
335
+ "dtype": "torch.float32",
336
+ "finite": true,
337
+ "max": 0.0,
338
+ "min": 0.0,
339
+ "shape": [
340
+ 49,
341
+ 3,
342
+ 288,
343
+ 512
344
+ ]
345
+ },
346
+ "raymap_camera_image": {
347
+ "dtype": "torch.float32",
348
+ "finite": true,
349
+ "max": 0.9833903312683105,
350
+ "min": -0.8238287568092346,
351
+ "shape": [
352
+ 49,
353
+ 3,
354
+ 288,
355
+ 512
356
+ ]
357
+ },
358
+ "rgb": {
359
+ "dtype": "torch.uint8",
360
+ "nonzero_count": 21674368,
361
+ "shape": [
362
+ 49,
363
+ 3,
364
+ 288,
365
+ 512
366
+ ]
367
+ },
368
+ "tracking_video": {
369
+ "dtype": "torch.uint8",
370
+ "nonzero_count": 52272,
371
+ "shape": [
372
+ 49,
373
+ 3,
374
+ 288,
375
+ 512
376
+ ]
377
+ }
378
+ },
379
+ "tracks_painted": [
380
+ 17424,
381
+ 80,
382
+ 80
383
+ ]
384
+ }
385
+ ],
386
+ "source": "molmo_motion",
387
+ "status": "passed",
388
+ "summary": {
389
+ "camera_valid_windows": 1,
390
+ "caption_coverage": "1/1",
391
+ "identity_checks_passed": "1/1",
392
+ "reprojection_px_by_subset": {
393
+ "egodex": {
394
+ "fps": [
395
+ 7.5
396
+ ],
397
+ "frame0_median_px": 0.21846943246718498,
398
+ "kept_tracks_median": 80.0,
399
+ "later_median_px": 0.2384225186913857,
400
+ "later_p90_px_median": 0.33664261176933946,
401
+ "mp4_step": 2.0,
402
+ "points_frame": [
403
+ "world"
404
+ ],
405
+ "windows": 1
406
+ }
407
+ },
408
+ "subsets": {
409
+ "egodex": 1
410
+ },
411
+ "tracking_video_valid": "1/1",
412
+ "windows": 1,
413
+ "windows_with_a_redraw": 0
414
+ },
415
+ "vgg_commit": "f6992c4b67441dc9778642a252447a2687856116",
416
+ "windows": 1,
417
+ "windows_yaml_sha256_without_the_receipt_pin": "18a73879963edbb91ed188699e57ce432b4c4a0431d47bf8aa581ad0a2b0a89e"
418
+ }
cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/omni4d_vgg_stage_one_next_version_all_sources_plus_hdepic_49_frames_ddp_global32_b4_gas4.yaml ADDED
@@ -0,0 +1,815 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # B4/GAS4 49-frame continuation from the update-14,700 weights-only overlay; no Cityscapes.
2
+ # This is a separate model-only run from an explicit five-stream overlay: 21 sources, total weight 139.
3
+ # The Cityscapes source is deliberately omitted; the remaining real-video sources keep their prior weights.
4
+ # Next Cosmos stage-one version: the owner's weighted mixture at 49 frames (decision 2026-11-07).
5
+ #
6
+ # Weights, not shares. Total 139 over 21 logical sources:
7
+ # synthetic 79 (56.8%) the eleven synthetic sources at the 20,000-update Wan run's weights, unchanged:
8
+ # syn4d 15.2, omniworld_game 8.75, synthverse 11.05, vkitti 6, tartanair 6, dynamic_replica 5,
9
+ # spring 8, midair 3, mvs_synth 3, tartanground 3, paralleldomain4d 10
10
+ # real video 22 (15.8%) ScanNet 8, DL3DV 6, WildRGB-D 4, Waymo 4; Cityscapes excluded.
11
+ # robot 26 (18.7%) RoboTwin 10 (clean-train episodes only), DROID 10 (depth and camera as inputs only),
12
+ # LIBERO 6 (official training demonstrations only); no action stream in stage one
13
+ # tracks 6 ( 4.3%) PointOdyssey 3, Kubric 3
14
+ # human video 6 ( 4.3%) MolmoMotion including HD-EPIC; DINO and sparse tracking video are targets, camera is input-only, depth absent
15
+ # DINO is a generated target on the real-video sources (ScanNet, DL3DV, WildRGB-D, DROID, Waymo): it is computed from the
16
+ # RGB, so it is exact on real data, and the future-feature forecasts the DINO-Foresight and VFMF comparisons make would
17
+ # otherwise never train on real video. This changes the first launch's roles: the owner's rule is that geometry is
18
+ # condition-only on real data, and that rule still holds for depth and camera, which stay inputs only here.
19
+ # Waymo trains on the processed training split only (waymo.training.pickle.gz, SHA-256 6abd3476...: 798 sequences,
20
+ # 158,081 frames; validation never enters training) through the pinned sparse-track reader's waymo stanza: sparse LiDAR
21
+ # depth, the camera and the RGB; no tracking video; strides 1-4 at 21 frames.
22
+ # No segmentation stream, no Cityscapes, no YAM.
23
+ #
24
+ # The VGG window block below is read by the pinned VGG checkout (vgg_commit) through VggStageOneWindowDataset: the
25
+ # synthetic sources are the recorded 500K run's stanzas (omni4d_vgg_sparse_tracking_synthetic_continuation_sources.yaml,
26
+ # SHA-256 ae49ec6d...) with the window length 49 and missing_caption: error. RoboTwin and LIBERO are read through
27
+ # robot_sources (RobotEpisodeStageOneDataset) because the pinned VGG checkout has no Track4D reader for them.
28
+ # The tracking-video query draw keeps the recorded 10-pixel grid (no query_count_range).
29
+ #
30
+ # Train membership of the real-video sources (official split files, 2026-11-05):
31
+ # DL3DV the pinned reader ignores `split` and indexes every processed scene, 85 of which are official
32
+ # DL3DV-Benchmark (test) scenes. The 141 official benchmark scene hashes (Hugging Face
33
+ # DL3DV/DL3DV-Benchmark, dataset revision 9684e838) are in omni4d_dl3dv_official_benchmark_scene_hashes.txt;
34
+ # the 87 of them present in the processed 1K-7K root are added to the stanza's exclude_sequence_ids below
35
+ # (source_reader_overrides), leaving about 5,898 scenes.
36
+ # WildRGB-D the pinned reader reads <category>/selected_seqs_train.json; those 2,244 processed train scenes are all in
37
+ # the official train-of-both lists (nvs_list.json and camera_eval_list.json of hongchi/wildrgbd, CUT3R's
38
+ # get_set_list rule) and overlap the official NVS and camera-evaluation validation scenes in none.
39
+ #
40
+ # launch_blockers names every evidence gap that stops the launch. A blocked source keeps its weight; composition
41
+ # (omni4d_native_config.make_config) and the window dataset both refuse to start while the list is not empty. A robot
42
+ # source additionally needs a passing real-window readback receipt (readback_receipt). WildRGB-D and Kubric train on
43
+ # caption text converted by tools/omni4d/convert_stage_one_caption_texts.py (text-only products: no umT5 cache exists).
44
+ #
45
+ # VGG pin: the 500K run's f5a59cd51 plus the window-reader updates, processed-release EgoDex support, and the
46
+ # native2048 Cityscapes caption-reader migration at f6992c4b. The approved raw EgoDex window and Cityscapes caption
47
+ # product readbacks used this exact pin. City captions point to the completed 2026-11-06 converted text product below.
48
+ # One makes the DROID and LIBERO robot window length the data node's num_frames: the pinned window class and the LIBERO
49
+ # painter fixed the clock at 33 rows in five places, one a literal inside the painter's constructor; a project adapter
50
+ # could only rebind module globals and fake attributes, so the smallest safe change is in the VGG source, with the
51
+ # recorded 33-row nodes reading as before. Two DROID commits (cherry-picked from the VGG main stack, f5a59cd is their
52
+ # ancestor) keep a window whose FK arm has no visible frame-0 query, with its source motion absent, instead of raising.
53
+ # The fourth adds the stanza key position_windows (src/vggforecast/data/position_windows.py): each listed position range is
54
+ # one window and the stride and start are drawn inside it. ScanNet and PointOdyssey use it (source_reader_overrides,
55
+ # caption_windows: true) with the local caption windows of their products, so every window they draw lies inside one
56
+ # caption window and the strict caption join never redraws. The fifth adds a waymo stanza to the stage-one real-source config
57
+ # (no source uses it yet: every real source leaves it out). The pin is a separate checkout (a git worktree of the earlier pin
58
+ # directory, which a running training still reads at 9cb6a626a), so the earlier pin's files never change under a run.
59
+ vgg_repository: /scratch/shared/engs2907/vgg_pins/vgg_forecast_4d_cosmos_source_20261106_f6992c4
60
+ vgg_commit: f6992c4b67441dc9778642a252447a2687856116
61
+ recipe_source_config: /scratch/shared/engs2907/logs/vgg_forecast_4d/train/flexpi_four_d_pretraining_wan22_5b_separate_streams_full_parameters_video_depth_camera_dino_and_sparse_tracking_video_from_the_15000_update_four_d_run_20261101/attempt02/run/config.yaml
62
+ recipe_source_config_sha256: 72590f4598cae1b329c31b3cb45d059bdaeca23642faa10ca3f31372f9ebc0aa
63
+ window_block:
64
+ _target_: four_d_windows.FourDPretrainingWindows
65
+ sources:
66
+ - name: four_d_pretraining_relaunch_syn4d_sources
67
+ kind: track4d_source_config
68
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_rendered_depth_sources_21_frames_512
69
+ weight: 15.2
70
+ left_out_sources:
71
+ - omniworld_game
72
+ - tartanair
73
+ - mvs_synth
74
+ condition_only_streams: []
75
+ - name: four_d_pretraining_relaunch_omniworld_game_source
76
+ kind: track4d_source_config
77
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_rendered_depth_sources_21_frames_512
78
+ weight: 8.75
79
+ left_out_sources:
80
+ - syn4d_v1
81
+ - syn4d_v2_mixed
82
+ - syn4d_v2_mixed_no_sim
83
+ - tartanair
84
+ - mvs_synth
85
+ condition_only_streams: []
86
+ - name: four_d_pretraining_relaunch_synthverse_source
87
+ kind: track4d_source_config
88
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_synthverse_dynamic_replica_vkitti_midair_spring_sources_without_midair_quality_masks_21_frames_512
89
+ weight: 11.05
90
+ left_out_sources:
91
+ - dynamic_replica
92
+ - vkitti
93
+ - midair
94
+ - spring
95
+ condition_only_streams: []
96
+ - name: four_d_pretraining_relaunch_vkitti_source
97
+ kind: track4d_source_config
98
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_synthverse_dynamic_replica_vkitti_midair_spring_sources_without_midair_quality_masks_21_frames_512
99
+ weight: 6.0
100
+ left_out_sources:
101
+ - synthverse_depth
102
+ - dynamic_replica
103
+ - midair
104
+ - spring
105
+ condition_only_streams: []
106
+ - name: four_d_pretraining_relaunch_tartanair_sky_then_far_percentile_source
107
+ kind: track4d_source_config
108
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_sky_and_far_percentile_depth_validity_rules_21_frames_512
109
+ weight: 6.0
110
+ left_out_sources:
111
+ - dynamic_replica
112
+ - spring
113
+ - midair
114
+ - mvs_synth
115
+ - tartanground
116
+ - paralleldomain4d
117
+ condition_only_streams: []
118
+ - name: four_d_pretraining_relaunch_dynamic_replica_stored_depth_source
119
+ kind: track4d_source_config
120
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_sky_and_far_percentile_depth_validity_rules_21_frames_512
121
+ weight: 5.0
122
+ left_out_sources:
123
+ - tartanair
124
+ - spring
125
+ - midair
126
+ - mvs_synth
127
+ - tartanground
128
+ - paralleldomain4d
129
+ condition_only_streams: []
130
+ - name: four_d_pretraining_relaunch_spring_stored_depth_source
131
+ kind: track4d_source_config
132
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_sky_and_far_percentile_depth_validity_rules_21_frames_512
133
+ weight: 8.0
134
+ left_out_sources:
135
+ - tartanair
136
+ - dynamic_replica
137
+ - midair
138
+ - mvs_synth
139
+ - tartanground
140
+ - paralleldomain4d
141
+ condition_only_streams: []
142
+ - name: four_d_pretraining_relaunch_midair_within_200_meters_source
143
+ kind: track4d_source_config
144
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_sky_and_far_percentile_depth_validity_rules_21_frames_512
145
+ weight: 3.0
146
+ left_out_sources:
147
+ - tartanair
148
+ - dynamic_replica
149
+ - spring
150
+ - mvs_synth
151
+ - tartanground
152
+ - paralleldomain4d
153
+ condition_only_streams: []
154
+ - name: four_d_pretraining_relaunch_mvs_synth_1000_then_far_percentile_source
155
+ kind: track4d_source_config
156
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_sky_and_far_percentile_depth_validity_rules_21_frames_512
157
+ weight: 3.0
158
+ left_out_sources:
159
+ - tartanair
160
+ - dynamic_replica
161
+ - spring
162
+ - midair
163
+ - tartanground
164
+ - paralleldomain4d
165
+ condition_only_streams: []
166
+ - name: four_d_pretraining_relaunch_tartanground_sky_then_far_percentile_source
167
+ kind: track4d_source_config
168
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_sky_and_far_percentile_depth_validity_rules_21_frames_512
169
+ weight: 3.0
170
+ left_out_sources:
171
+ - tartanair
172
+ - dynamic_replica
173
+ - spring
174
+ - midair
175
+ - mvs_synth
176
+ - paralleldomain4d
177
+ condition_only_streams: []
178
+ - name: four_d_pretraining_relaunch_paralleldomain4d_sky_then_far_percentile_source
179
+ kind: track4d_source_config
180
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_sky_and_far_percentile_depth_validity_rules_21_frames_512
181
+ weight: 10.0
182
+ left_out_sources:
183
+ - tartanair
184
+ - dynamic_replica
185
+ - spring
186
+ - midair
187
+ - mvs_synth
188
+ - tartanground
189
+ condition_only_streams: []
190
+ - name: scannet
191
+ kind: track4d_source_config
192
+ config: train/mot/stage_one_real_and_native_track_source_windows_9_frames_288_512
193
+ weight: 8.0
194
+ left_out_sources:
195
+ - dl3dv
196
+ - wildrgbd
197
+ - pointodyssey_train
198
+ - kubric_train
199
+ - droid
200
+ - waymo
201
+ condition_only_streams:
202
+ - pointmap
203
+ - camera
204
+ - name: dl3dv
205
+ kind: track4d_source_config
206
+ config: train/mot/stage_one_real_and_native_track_source_windows_9_frames_288_512
207
+ weight: 6.0
208
+ left_out_sources:
209
+ - scannet
210
+ - wildrgbd
211
+ - pointodyssey_train
212
+ - kubric_train
213
+ - droid
214
+ - waymo
215
+ condition_only_streams:
216
+ - pointmap
217
+ - camera
218
+ - name: wildrgbd
219
+ kind: track4d_source_config
220
+ config: train/mot/stage_one_real_and_native_track_source_windows_9_frames_288_512
221
+ weight: 4.0
222
+ left_out_sources:
223
+ - dl3dv
224
+ - scannet
225
+ - pointodyssey_train
226
+ - kubric_train
227
+ - droid
228
+ - waymo
229
+ condition_only_streams:
230
+ - pointmap
231
+ - camera
232
+ - name: waymo
233
+ kind: track4d_source_config
234
+ config: train/mot/stage_one_real_and_native_track_source_windows_9_frames_288_512
235
+ weight: 4.0
236
+ left_out_sources:
237
+ - dl3dv
238
+ - scannet
239
+ - wildrgbd
240
+ - pointodyssey_train
241
+ - kubric_train
242
+ - droid
243
+ condition_only_streams:
244
+ - pointmap
245
+ - camera
246
+ - name: droid
247
+ kind: track4d_source_config
248
+ config: train/mot/stage_one_real_and_native_track_source_windows_9_frames_288_512
249
+ weight: 10.0
250
+ left_out_sources:
251
+ - dl3dv
252
+ - scannet
253
+ - wildrgbd
254
+ - pointodyssey_train
255
+ - kubric_train
256
+ - waymo
257
+ condition_only_streams:
258
+ - pointmap
259
+ - camera
260
+ - name: pointodyssey_train
261
+ kind: track4d_source_config
262
+ config: train/mot/stage_one_real_and_native_track_source_windows_9_frames_288_512
263
+ weight: 3.0
264
+ left_out_sources:
265
+ - dl3dv
266
+ - scannet
267
+ - wildrgbd
268
+ - kubric_train
269
+ - droid
270
+ - waymo
271
+ condition_only_streams: []
272
+ - name: kubric_train
273
+ kind: track4d_source_config
274
+ config: train/mot/stage_one_real_and_native_track_source_windows_9_frames_288_512
275
+ weight: 3.0
276
+ left_out_sources:
277
+ - dl3dv
278
+ - scannet
279
+ - wildrgbd
280
+ - pointodyssey_train
281
+ - droid
282
+ - waymo
283
+ condition_only_streams:
284
+ - pointmap
285
+ frames: 49
286
+ raster_hw:
287
+ - 288
288
+ - 512
289
+ raymap_camera:
290
+ origin_fraction: 0.3333333333333333
291
+ origin_range: 1.0
292
+ decode_margin: 16
293
+ origin_box_grid: 1
294
+ pointmap_log_factor: 0.4
295
+ pointmap_hole_fill:
296
+ synthverse_depth: far_clip
297
+ syn4d: far_clip
298
+ omniworld_game: far_clip
299
+ dynamic_replica: nearest_labelled_depth_of_3x3_neighbours
300
+ tartanair: far_clip
301
+ dl3dv: far_clip
302
+ scannet: telea_log_depth
303
+ virtual_kitti2: far_clip
304
+ midair: far_clip
305
+ pointodyssey: far_clip
306
+ mvs_synth: far_clip
307
+ wildrgbd: telea_log_depth
308
+ hypersim: far_clip
309
+ spring: far_clip
310
+ kubric49: far_clip
311
+ kubric81: far_clip
312
+ droid_stereo_depth: telea_log_depth
313
+ tartanground: far_clip
314
+ paralleldomain4d: far_clip
315
+ syn4d_v2: far_clip
316
+ waymo: far_clip
317
+ pointmap_gauge: window_labelled_median_depth
318
+ geometry_image: genception_log_depth
319
+ context_len: 512
320
+ batch_size: 4
321
+ stream_seed: 42
322
+ epoch_windows: 32000000
323
+ val_set_proportion: 0.0
324
+ resume_state: null
325
+ gradient_accumulation_steps: 4
326
+ loader_workers: 12
327
+ loader_opencv_threads: 1
328
+ window_captions:
329
+ caption_dropout: 0.1
330
+ missing_caption: error
331
+ caches:
332
+ syn4d_v1_gemini_holo4d:
333
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/syn4d_v1_gemini_holo4d_local_and_global_with_repaired_globals_20261031/captions_t5_emb.safetensors
334
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/syn4d_v1_gemini_holo4d_local_and_global_with_repaired_globals_20261031/captions_t5_emb_meta.json
335
+ syn4d_v2_gemini_holo4d:
336
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/syn4d_v2_gemini_holo4d_20261030/captions_t5_emb.safetensors
337
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/syn4d_v2_gemini_holo4d_20261030/captions_t5_emb_meta.json
338
+ tartanair_gemini_holo4d:
339
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/tartanair_gemini_holo4d/captions_t5_emb.safetensors
340
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/tartanair_gemini_holo4d/captions_t5_emb_meta.json
341
+ mvs_synth_gemini_holo4d:
342
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/mvs_synth_gemini_holo4d/captions_t5_emb.safetensors
343
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/mvs_synth_gemini_holo4d/captions_t5_emb_meta.json
344
+ omniworld_game_gemini_local_official_text_chunks_and_gemini_global:
345
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/omniworld_game_gemini_local_official_text_chunks_and_gemini_global_20261030b/captions_t5_emb.safetensors
346
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/omniworld_game_gemini_local_official_text_chunks_and_gemini_global_20261030b/captions_t5_emb_meta.json
347
+ synthverse_gemini_holo4d:
348
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/synthverse_gemini_holo4d_local96_and_global_20261031/captions_t5_emb.safetensors
349
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/synthverse_gemini_holo4d_local96_and_global_20261031/captions_t5_emb_meta.json
350
+ dynamic_replica_gemini_holo4d:
351
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/dynamic_replica_gemini_holo4d/captions_t5_emb.safetensors
352
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/dynamic_replica_gemini_holo4d/captions_t5_emb_meta.json
353
+ vkitti_gemini_holo4d:
354
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/vkitti_gemini_holo4d/captions_t5_emb.safetensors
355
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/vkitti_gemini_holo4d/captions_t5_emb_meta.json
356
+ midair_gemini_holo4d:
357
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/midair_gemini_holo4d_20261030/captions_t5_emb.safetensors
358
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/midair_gemini_holo4d_20261030/captions_t5_emb_meta.json
359
+ spring_gemini_holo4d:
360
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/spring_gemini_holo4d/captions_t5_emb.safetensors
361
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/spring_gemini_holo4d/captions_t5_emb_meta.json
362
+ tartanground_gemini_holo4d:
363
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/tartanground_gemini_holo4d_20261030/captions_t5_emb.safetensors
364
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/tartanground_gemini_holo4d_20261030/captions_t5_emb_meta.json
365
+ paralleldomain4d_gemini_holo4d:
366
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/paralleldomain4d_gemini_holo4d_20261030/captions_t5_emb.safetensors
367
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/paralleldomain4d_gemini_holo4d_20261030/captions_t5_emb_meta.json
368
+ scannet_gemini_holo4d:
369
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/.scannet_gemini_holo4d.version-tovvjwai/captions_t5_emb.safetensors
370
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/.scannet_gemini_holo4d.version-tovvjwai/captions_t5_emb_meta.json
371
+ pointodyssey_gemini_holo4d:
372
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/.pointodyssey_gemini_holo4d.version-6vfyvsbi/captions_t5_emb.safetensors
373
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/.pointodyssey_gemini_holo4d.version-6vfyvsbi/captions_t5_emb_meta.json
374
+ droid_vp_gemini_holo4d:
375
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/.droid_vp_gemini_holo4d.version-e51r5urf/captions_t5_emb.safetensors
376
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/.droid_vp_gemini_holo4d.version-e51r5urf/captions_t5_emb_meta.json
377
+ dl3dv_gemini_holo4d:
378
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/dl3dv_gemini_holo4d_20261031/captions_t5_emb.safetensors
379
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/dl3dv_gemini_holo4d_20261031/captions_t5_emb_meta.json
380
+ wildrgbd_text_captions:
381
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/cosmos_caption_texts_20261105/wildrgbd_text_captions/captions_t5_emb.safetensors
382
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/cosmos_caption_texts_20261105/wildrgbd_text_captions/captions_t5_emb_meta.json
383
+ kubric49_text_captions:
384
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/cosmos_caption_texts_20261105/kubric49_text_captions/captions_t5_emb.safetensors
385
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/cosmos_caption_texts_20261105/kubric49_text_captions/captions_t5_emb_meta.json
386
+ waymo_gemini_holo4d:
387
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/cosmos_caption_texts_20261105/waymo_gemini_holo4d/captions_t5_emb.safetensors
388
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/cosmos_caption_texts_20261105/waymo_gemini_holo4d/captions_t5_emb_meta.json
389
+ stanzas:
390
+ syn4d_v1:
391
+ product: syn4d_v1_gemini_holo4d
392
+ join: local_covering_then_global
393
+ syn4d_v2_mixed:
394
+ product: syn4d_v2_gemini_holo4d
395
+ join: local_covering_then_global
396
+ syn4d_v2_mixed_no_sim:
397
+ product: syn4d_v2_gemini_holo4d
398
+ join: local_covering_then_global
399
+ tartanair:
400
+ product: tartanair_gemini_holo4d
401
+ join: local_covering
402
+ mvs_synth:
403
+ product: mvs_synth_gemini_holo4d
404
+ join: local_covering
405
+ omniworld_game:
406
+ product: omniworld_game_gemini_local_official_text_chunks_and_gemini_global
407
+ join: local_covering_then_second_local_covering_then_global
408
+ pointodyssey_test: none
409
+ kubric49_validation: none
410
+ wildrgbd_test: none
411
+ vkitti_test:
412
+ product: vkitti_gemini_holo4d
413
+ join: local_covering_then_global
414
+ droid_held_out: none
415
+ synthverse_depth:
416
+ product: synthverse_gemini_holo4d
417
+ join: local_covering_then_global
418
+ dynamic_replica:
419
+ product: dynamic_replica_gemini_holo4d
420
+ join: global
421
+ vkitti:
422
+ product: vkitti_gemini_holo4d
423
+ join: local_covering_then_global
424
+ midair:
425
+ product: midair_gemini_holo4d
426
+ join: local_covering_then_global
427
+ spring:
428
+ product: spring_gemini_holo4d
429
+ join: global
430
+ tartanground:
431
+ product: tartanground_gemini_holo4d
432
+ join: local_covering_then_global
433
+ paralleldomain4d:
434
+ product: paralleldomain4d_gemini_holo4d
435
+ join: global
436
+ paralleldomain4d_validation:
437
+ product: paralleldomain4d_gemini_holo4d
438
+ join: global
439
+ scannet:
440
+ product: scannet_gemini_holo4d
441
+ join: local_containing_then_global
442
+ pointodyssey_train:
443
+ product: pointodyssey_gemini_holo4d
444
+ join: local_containing_then_global
445
+ droid:
446
+ product: droid_vp_gemini_holo4d
447
+ join: global
448
+ dl3dv:
449
+ product: dl3dv_gemini_holo4d
450
+ join: local_containing_then_global
451
+ wildrgbd:
452
+ product: wildrgbd_text_captions
453
+ join: local_containing_then_global
454
+ kubric_train:
455
+ product: kubric49_text_captions
456
+ join: global_required
457
+ waymo:
458
+ product: waymo_gemini_holo4d
459
+ join: local_containing_then_global
460
+ training_scene_exclusions:
461
+ reason: rendered depth passes through glass (raw EXR depth through glass panes at 190-260 m), owner 10-29; evidence /scratch/shared/engs2907/logs/vgg_forecast_4d/tools/syn4d_glass_depth_check_20261029;
462
+ scifiroom_bald added 10-30 from /scratch/shared/engs2907/logs/vgg_forecast_4d/tools/depth_clipping_audit_20261030
463
+ stanzas:
464
+ syn4d_v1:
465
+ - bigoffice_v1
466
+ - lab_bald
467
+ - office_bald
468
+ - scifiroom_bald
469
+ syn4d_v2_mixed:
470
+ - citynight
471
+ - mall
472
+ - metro
473
+ syn4d_v2_mixed_no_sim:
474
+ - citynight
475
+ - mall
476
+ - metro
477
+ tracking_video:
478
+ query_grid_px: 10
479
+ disc_radius_px: 2
480
+ occluded_style: ring
481
+ ring_radius_px: 3
482
+ ring_width_px: 1
483
+ xyz_normalization: quantile_0.02_0.98
484
+ colour_floor: 0.2
485
+ min_frame_zero_points: 16
486
+ surface_test:
487
+ relative_tolerance: 0.1
488
+ absolute_tolerance: 0.05
489
+ stanza_rules:
490
+ tartanair: static_scene_points
491
+ midair: static_scene_points
492
+ mvs_synth: static_scene_points
493
+ tartanground: static_scene_points
494
+ syn4d_v1: tracked_points
495
+ syn4d_v2_mixed: tracked_points
496
+ syn4d_v2_mixed_no_sim: tracked_points
497
+ omniworld_game: no_tracking_video
498
+ synthverse_depth: no_tracking_video
499
+ dynamic_replica: tracked_points
500
+ vkitti: no_tracking_video
501
+ spring: no_tracking_video
502
+ paralleldomain4d: no_tracking_video
503
+ scannet: static_scene_points
504
+ wildrgbd_test: static_scene_points
505
+ pointodyssey_test: tracked_points
506
+ kubric49_validation: no_tracking_video
507
+ vkitti_test: no_tracking_video
508
+ droid_held_out: no_tracking_video
509
+ paralleldomain4d_validation: no_tracking_video
510
+ dynamic_replica_test: tracked_points
511
+ pointodyssey_train: tracked_points
512
+ kubric_train: tracked_points
513
+ droid: tracked_points
514
+ dl3dv: static_scene_points
515
+ wildrgbd: static_scene_points
516
+ waymo: no_tracking_video
517
+ stanza_track_reading:
518
+ syn4d_v1:
519
+ stanza:
520
+ capability: dynamic_cross_time
521
+ supervision:
522
+ - pointmap
523
+ - source_motion
524
+ annotation_provenance: syn4d_mesh_surface_forward_track_from_frame0
525
+ allow_missing_tracks: false
526
+ kwargs:
527
+ return_tracks: true
528
+ depth_only: false
529
+ metadata_root: /vgg/chenzhao/syn4d_metadata
530
+ wrapped_face_distance_thresh: 0.1
531
+ query_inconsistent_distance_thresh: 0.1
532
+ pointodyssey_test:
533
+ stanza:
534
+ supervision:
535
+ - pointmap
536
+ - source_motion
537
+ kwargs:
538
+ return_tracks: true
539
+ dynamic_replica_test:
540
+ stanza:
541
+ supervision:
542
+ - pointmap
543
+ - source_motion
544
+ kwargs:
545
+ return_tracks: true
546
+ syn4d_v2_mixed:
547
+ loader: syn4d_v2
548
+ stanza:
549
+ capability: dynamic_cross_time
550
+ supervision:
551
+ - pointmap
552
+ - source_motion
553
+ annotation_provenance: syn4d_v2_mesh_surface_forward_track_from_frame0
554
+ allow_missing_tracks: false
555
+ kwargs:
556
+ return_tracks: true
557
+ depth_only: false
558
+ metadata_root: /vgg/chenzhao/syn4d_metadata
559
+ tracking_format: safetensor
560
+ mask_bg: false
561
+ trusted_background_depth_max: 80.0
562
+ wrapped_face_distance_thresh: 0.1
563
+ query_inconsistent_distance_thresh: 0.1
564
+ syn4d_v2_mixed_no_sim:
565
+ loader: syn4d_v2
566
+ stanza:
567
+ capability: dynamic_cross_time
568
+ supervision:
569
+ - pointmap
570
+ - source_motion
571
+ annotation_provenance: syn4d_v2_mesh_surface_forward_track_from_frame0
572
+ allow_missing_tracks: false
573
+ kwargs:
574
+ return_tracks: true
575
+ depth_only: false
576
+ metadata_root: /vgg/chenzhao/syn4d_metadata
577
+ tracking_format: safetensor
578
+ mask_bg: false
579
+ trusted_background_depth_max: 80.0
580
+ wrapped_face_distance_thresh: 0.1
581
+ query_inconsistent_distance_thresh: 0.1
582
+ dynamic_replica:
583
+ stanza:
584
+ supervision:
585
+ - pointmap
586
+ - source_motion
587
+ kwargs:
588
+ return_tracks: true
589
+ pointodyssey_train:
590
+ stanza:
591
+ allow_missing_tracks: false
592
+ capability: dynamic_cross_time
593
+ supervision:
594
+ - pointmap
595
+ - source_motion
596
+ kwargs:
597
+ return_tracks: true
598
+ kubric_train:
599
+ stanza:
600
+ allow_missing_tracks: false
601
+ capability: dynamic_cross_time
602
+ supervision:
603
+ - pointmap
604
+ - source_motion
605
+ kwargs:
606
+ return_tracks: true
607
+ droid:
608
+ stanza:
609
+ allow_missing_tracks: false
610
+ capability: dynamic_cross_time
611
+ supervision:
612
+ - pointmap
613
+ - source_motion
614
+ kwargs:
615
+ return_tracks: true
616
+ source_reader_overrides:
617
+ dl3dv:
618
+ stanza: dl3dv
619
+ exclude_sequence_ids:
620
+ - 1K/0a1b7c20a92c43c6b8954b1ac909fb2f0fa8b2997b80604bc8bbec80a1cb2da3
621
+ - 1K/14eb48a50e37df548894ab6d8cd628a21dae14bbe6c462e894616fc5962e6c49
622
+ - 1K/1da888bdedfc9629c0fa9f82cf3f5d96f8103baee0ff64d9311aea1224a9f2ae
623
+ - 1K/2991a75d1f0fb2b7a891d424fc2459bb954b61599b52e74a056a7acfb2998b77
624
+ - 1K/2cbfe28643b6636f9c70813cae7625aa858a352109493ac70fb429ce94dd55b3
625
+ - 1K/5a27c00f525f5484b7aa0d0107051253bf7bb395bde0b3a4486ff8c52fb4965e
626
+ - 1K/6d221625614d0515fa44d59a8797a49b64f86bc775e66be22f07a66aa0a0cb3a
627
+ - 1K/917e9c8985d353b0ee4c281f11fa84eb8550814562670cbb82cd0ec9c1194fe7
628
+ - 1K/918c8dad730c3b804306c5da8486124be4aa0612e85fb825338fd350c912e1b0
629
+ - 1K/a17a984ca90a9b5840fdf85b15104b0d18e25975981c1aa90fcdfd6eeeb285f3
630
+ - 1K/a726c1112ad026efc92f59f196c0698e67abd12a04135a7ff6349b1d31b089eb
631
+ - 1K/adb95f29c15b8a89e40c17d10596679be09d7719c6f08a453facd506d056ea8b
632
+ - 1K/d1b3a0b37acc72207295b05e8c359f8b43bf879eb882c60747cfc6d3c7d8efa3
633
+ - 1K/eb4cf52988f805e6fce11d1b239fa9de32eb157364cff06ebac0aa50e0a46567
634
+ - 2K/032dee9fb0a8bc1b90871dc5fe950080d0bcd3caf166447f44e60ca50ac04ec7
635
+ - 2K/0853979305f7ecb80bd8fc2c8df916410d471ef04ed5f1a64e9651baa41d7695
636
+ - 2K/093ef327b4e4f9d4ee52c02a354a53558a8652157fb0d58f3b4a708734afb334
637
+ - 2K/0f8ac521439691fe429e9efbed8d5ded1cee35bcf52d47731fb60d5a2ff661a7
638
+ - 2K/1d6a9ed47cce39fd1c4d18f776bcc97e507b81bde921ca596bd91b0b02b5e414
639
+ - 2K/26fd23358fa11fff0fb3180ef0b65591b486e20dcf753ce4a7aae49a37e370c7
640
+ - 2K/280abf7bd93b81b077af1db638229dbb09869052fa0b7b57c81c94d2db893829
641
+ - 2K/2beaca318994c25409dcbb6d0bdd96c3620f2f18aec44ea3f20edd302f18ca78
642
+ - 2K/341b4ff3dfd3d377d7167bd81f443bedafbff003bf04881b99760fc0aeb69510
643
+ - 2K/374ffd0c5ff8e544f588f1da3ce0d5b69dc39ff3af95d7ccec480fe5e533c45e
644
+ - 2K/41036716da7efda334c1d434c4141d15642e0e02f881a01b6c8c36f8bea64c45
645
+ - 2K/484c0aca40c1087f196d2caf3c8cb463864c08e60b98326ca02eb6dae8ce4984
646
+ - 2K/4ff8650b5c1140e98d957fbd1abbe2cfbbb416b37cb9f06f7f376956703801bb
647
+ - 2K/513e4ea2e8477b06f2b32417d5c243aec71d491cb0596a60e9fe304c635f20a1
648
+ - 2K/8b9fb9d9f10e8c64d5034be69809465753b8ba88ef12da82afd47d38ee789934
649
+ - 2K/91afb9910b042f7185c2b8e4b6b24b5785dae4542617b3f8005b5492f6d123f7
650
+ - 2K/ba55c875d20c34ee85ffc72264c4d77710852e5fb7d9ce4b9c26a8442850e98f
651
+ - 2K/c076929db6501cf7ebe386c70e6d77ea3af844a745e794f2ec17c981c465a69b
652
+ - 2K/ceb252f5d4419510655cf9ed7afbf3e8e688825f798d80414c7715dc8ace153a
653
+ - 2K/df29c225863d0173e060b1f3fde4e6eb5a57a4e4edabd397026ba1a028aeb39e
654
+ - 2K/ff592398657b3dfe94153332861985194a3e3c9d199c4a3a27a0ce4038e81ade
655
+ - 3K/35872363e17af5d173b6a0b09fcf5de94627ad5dc5f8a9ad4c579f3e70b4797a
656
+ - 3K/387eeb925ba47af5db9ce6102aaf6b07e6d3573b7e7568dda4cef22d5ab1342f
657
+ - 3K/3bb894d1933f3081134ad2d40e54de5f0636bd8b502b0a8561873bb63b0dce85
658
+ - 3K/5c3af581028068a3c402c7cbe16ecf9471ddf2897c34ab634b7b1b6cf81aba00
659
+ - 3K/669c36225b273e22ce08364f166dce910b708d7abd89e58463a15d2790b534d5
660
+ - 3K/66fd66cbed40a2d11ecd17ae6d6d60cd60d476616864ecdd351fe69e644f30a1
661
+ - 3K/70eac6ff18a1daae4eeccc5eb18723eaf7e029a77e79c79d9573f7bac59ba92b
662
+ - 3K/75fbbe467353cfdeb93bc1872144855576797be83a0c71c51155ee6b355d6d5e
663
+ - 3K/800cf88687a79c232700454c2f5a402e771bc7157037a17b0ce93da365b25575
664
+ - 3K/8cb2e97d26a639f05a571476240a8fa86988e6853f0f13cc05830d1578002aad
665
+ - 3K/9cbc5548643ca9cfe797f871bac58bccd02f0510c49b7cb3697b38dfd780a508
666
+ - 3K/a62f9a1c6362e7c8a3440261ea0921a0a4b60cf5822bd19579836694f26b3abd
667
+ - 3K/adf35184a12d4cfa3f4248b87aa5adb4f39f179df460d6d76136e13d37299a2a
668
+ - 3K/af0d7039e64af3959210c235e59425896ac064f4e0f721c437bbd2cf1d8ea910
669
+ - 3K/b3bf9079b442c681749f6bd95cfb0659c6d844f2228b2f4fa4b602a0e3627b9a
670
+ - 3K/b6d1134cb03c855d246f916beaa2d3d01290cf47937a5176011ff39b850ec615
671
+ - 3K/d8de66037bd03dd0d39d54f9978bac3318d912e30e22c21e1ada82a98ed48c53
672
+ - 3K/ddfdcfdf02d53dd89a055cd9e1ed30272ab3f57c7560ac2c3761bb01d76a7882
673
+ - 3K/e8ce51b6abfe05bf8dca47e29c8be6c1e6de27a8c9fece7a121400b931b2ca0f
674
+ - 3K/e9360e7a89bee835dc847cf8796093e634b759ff582558788dcfe8326f6e8901
675
+ - 3K/f7aaea9ac683bb1945cfeb1daa9abca3375249e0d19281a89a3f575040835464
676
+ - 4K/56452d9cd9e2e034764132a57058517ab085e1352f9664d2d1e10e98741334fc
677
+ - 4K/6d81c5ab0d480fd43d78b75ff372a8113ad38e2c03f1d69627c009883054d4c2
678
+ - 4K/9641a1ed7963ce5ca734cff3e6ccea3dfa8bcb0b0a3ff78f65d32a080de2d71e
679
+ - 4K/d3812aad538261e7f73c75762ff55f23b468bcc76f376d52ac86ca6cf3c44b4b
680
+ - 5K/0a485338bbdaf19ba9090b874bb36ef0599a9c9a475a382c22903cf5981c6ea6
681
+ - 5K/119fd56d3797e2d349ca64ddcc5851463cd13b5974b5b2e4566ed5cf7e02e6c1
682
+ - 5K/c37726ce770ac50a2cf5c0f43022f0268e26da0d777cd8e3a3418c4eed03fd94
683
+ - 5K/dac9796dd69e1c25277e29d7e30af4f21e3b575d62a0a208c2b3d1211e2d5d77
684
+ - 6K/06da796666297fe4c683c231edf56ec00148a6a52ab5bb159fe1be31f53a58df
685
+ - 6K/7a9f97660be8f2421b37dd413c898360532f17c6447c11515e8aded2d80eb99d
686
+ - 6K/7da3db99059a436adf537bc23309a6a4db9e011696d086f3a64037f7497d9df7
687
+ - 6K/e5684b3292bfd77db297839fc37ee4cce7fd59775af1a6a4827e3b4f59c036d3
688
+ - 6K/ec1e44d4dc0f8fa77610866495f9297a7f82158c43e1777668b84fd4b736c7bc
689
+ - 6K/ef59aac437132bfc1dd45a7e1f8e4800978e7bb28bf98c4428d26fb3e1da3e90
690
+ - 6K/f477ffc4b398bed8e0d921f0fba9825ca63f317381c535c84be23be991ae1d7a
691
+ - 7K/0569e83fdc248a51fc0ab082ce5e2baff15755c53c207f545e6d02d91f01d166
692
+ - 7K/073f5a9b983ced6fb28b23051260558b165f328a16b2d33fe20585b7ee4ad561
693
+ - 7K/183dd248f6a86e07c5adf9de8ee2d0abe45b1216331c03678e89634c2e9b1c7f
694
+ - 7K/1ba74c22670ad047981441581d00f26f4a148d1010bcac7468c615adf5fa4d5d
695
+ - 7K/389a460ca1995e0658e85fe8e6b520b4e88b370cd6710dfe728b1564bba31aee
696
+ - 7K/493816813d2d6d248eb3c2b0b77b63e54235266e9a06e270fd0d282f13960493
697
+ - 7K/4ae797d07b6d1644c9db6919c8cc8c0d28d72be45108ac7a3abf8dc21b599d83
698
+ - 7K/50c46cf8b8b22c8d2ffdef8964b05ddbceaef312c9a9ff331d1ecebfd223f72a
699
+ - 7K/565553aa894be621e8b4773cac288e60ad0c2cf7edb621be62b348c9a0f78380
700
+ - 7K/599ca3e04cae3ec83affc426af7d0d7ab36eb91cd8e539edbc13070a4d455792
701
+ - 7K/5c8dafad7d782c76ffad8c14e9e1244ce2b83aa12324c54a3cc10176964acf04
702
+ - 7K/63798f5c6fbfcb4eb686268248b8ecbc8d87d920b2bcce967eeaedfd3b3b6d82
703
+ - 7K/946f49be73928469000baa5ca04d2573137c5ee6a66362bcf8d130354dca8924
704
+ - 7K/9e9a89ae6fed06d6e2f4749b4b0059f35ca97f848cedc4a14345999e746f7884
705
+ - 7K/cd9c981eeb4a9091547af19181b382698e9d9eee0a838c7c9783a8a268af6aee
706
+ - 7K/d4fbeba0168af8fddb2fc695881787aedcd62f477c7dcec9ebca7b8594bbd95b
707
+ scannet:
708
+ stanza: scannet
709
+ caption_windows: true
710
+ pointodyssey_train:
711
+ stanza: pointodyssey_train
712
+ caption_windows: true
713
+ # Its local captions cover 96 native frames. A 49-frame draw must use
714
+ # stride 1 to stay inside that exact caption interval; the inherited
715
+ # [2, 4, 6, 8] choices do not fit any 96-frame caption window.
716
+ caption_compatible_temporal_strides:
717
+ - 1
718
+ robot_sources:
719
+ - name: robotwin
720
+ weight: 10.0
721
+ frames: 49
722
+ stride: 4
723
+ # Simulator ground truth: depth, camera and the arm tracks are generated targets (owner decision 2026-11-05); DINO too.
724
+ condition_only_streams: []
725
+ reader:
726
+ source_split_manifest: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/flexpi_robotwin_robot_arm_track_video_20261023/split_manifest.json
727
+ source_split_sha256: 10467f314d57acf8deba64b94d0842b0dedcc256acf5f055df92203f7bf10f4e
728
+ pretrained_norm_stats: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/robotwin_action_image_clean_train_filtered_20261102/native_stats/dataset_stats.json
729
+ clean_train_split:
730
+ manifest: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/robotwin_action_image_clean_train_filtered_20261102/split_manifest.json
731
+ sha256: 6511ed1a88a1047d74fb80dd11b1304f6d77abce6cbe3f56df4f5cf7b9f9ee7c
732
+ train_episodes: 26288
733
+ held_out_episodes: 150
734
+ excluded_clamped_episodes: 1062
735
+ geometry:
736
+ simulator_ground_truth: true
737
+ log_factor: 0.4
738
+ raymap_camera: {origin_fraction: 0.3333333333333333, origin_range: 1.0, decode_margin: 16, origin_box_grid: 1}
739
+ gauge: window_labelled_median_depth
740
+ release_root: /datasets/engs2907/FlexPi/robotwin_3d
741
+ calibration_root: /datasets/engs2907/FlexPi/robotwin_3d/calibration
742
+ # Two real clean-train windows at 21 frames through reader, DINO and the native batch adapter, on pin f6992c4b.
743
+ readback_receipt:
744
+ path: /scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_nano_two_training_restart_20261107/attempt01/robot49_cpu/results/robotwin_readback.json
745
+ sha256: 785c7dde5f5f69a6a6fe30fe6accf56fe2380ab8903b7c9f2ec625d0501ecf2f
746
+ - name: libero
747
+ weight: 6.0
748
+ frames: 49
749
+ stride: 4
750
+ condition_only_streams: []
751
+ reader:
752
+ hydra_overrides: []
753
+ geometry:
754
+ simulator_ground_truth: true
755
+ readback_receipt:
756
+ path: /scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_nano_two_training_restart_20261107/attempt01/robot49_cpu/results/libero_readback.json
757
+ sha256: c8d49b7d87f64b45f1183ab8babc3c17d632e5345e0f697ed4a2aebc6fac4a0f
758
+ real_video_sources: []
759
+ molmo_motion_sources:
760
+ - name: molmo_motion
761
+ weight: 6.0
762
+ frames: 49
763
+ # The camera is an input only, like every real source's camera; there is no depth. DINO and the tracking video are targets.
764
+ condition_only_streams: [camera]
765
+ reader:
766
+ config: train/mot/wan_camera_omega_video_depth_vae_and_sparse_future_prediction_molmo_motion_21_frames_512
767
+ stanza: molmo_motion
768
+ release_root: /vgg/yushi/dataset/MolmoMotion-1M
769
+ subsets: [egodex, xperience, stereo4d, hepic]
770
+ # Retain the prior relative weights for this initial 49-frame launch; these are not a newly measured 49-frame census.
771
+ subset_weights: {egodex: 133.6, xperience: 432.4, stereo4d: 210.9, hepic: 63.1}
772
+ temporal_stride: {egodex: 2, xperience: 1, stereo4d: 4, hepic: 1}
773
+ processed_release_only_subsets: [egodex, hepic]
774
+ # EgoDex and HD-EPIC use their processed release videos; neither raw-video path is used by this recipe.
775
+ # The pinned config leaves Xperience's `object` records out: their 3D was lifted with the 256-pixel camera matrix.
776
+ excluded_objects: {xperience: [object]}
777
+ max_window_redraws: 64
778
+ exclusions:
779
+ # The registered 120-record test cohort (the VGG held-out spec molmo_motion_test_120_records_droid_egodex_molmospaces_
780
+ # 49_frames_512; tools/omni4d/write_molmo_motion_test_cohort_manifest.py). The release's own test splits (DROID's too) and
781
+ # PointMotionBench (hot3d, worldtrack, davis) are asserted disjoint from the train records in code.
782
+ test_cohort_manifest:
783
+ path: /scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_molmo_motion_20261105/cohort/molmo_motion_test_cohort_120_records.json
784
+ sha256: d1f8e8d8ad8f8f67c8bfd23319cdea026ea8cfcba1194b5f4594366e9f64fd76
785
+ point_motion_bench_root: /scratch/shared/engs2907/PointMotionBench
786
+ readback_receipt:
787
+ path: /scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_molmo_49frame_source_readback_20261107/attempt01/native_group_readback/molmo_motion_readback.json
788
+ sha256: ec0577a78e97db76f14cca0bd4d44ab87bc4e588f615fca242c1d675abb8dbb1
789
+ launch_blockers: {}
790
+ mixture_policy:
791
+ total_weight: 139.0
792
+ group_weights: {synthetic: 79.0, real_video: 22.0, robot: 26.0, tracks: 6.0, human_video: 6.0}
793
+ group_share_percent: {synthetic: 56.8, real_video: 15.8, robot: 18.7, tracks: 4.3, human_video: 4.3}
794
+ source_groups:
795
+ four_d_pretraining_relaunch_syn4d_sources: synthetic
796
+ four_d_pretraining_relaunch_omniworld_game_source: synthetic
797
+ four_d_pretraining_relaunch_synthverse_source: synthetic
798
+ four_d_pretraining_relaunch_vkitti_source: synthetic
799
+ four_d_pretraining_relaunch_tartanair_sky_then_far_percentile_source: synthetic
800
+ four_d_pretraining_relaunch_dynamic_replica_stored_depth_source: synthetic
801
+ four_d_pretraining_relaunch_spring_stored_depth_source: synthetic
802
+ four_d_pretraining_relaunch_midair_within_200_meters_source: synthetic
803
+ four_d_pretraining_relaunch_mvs_synth_1000_then_far_percentile_source: synthetic
804
+ four_d_pretraining_relaunch_tartanground_sky_then_far_percentile_source: synthetic
805
+ four_d_pretraining_relaunch_paralleldomain4d_sky_then_far_percentile_source: synthetic
806
+ scannet: real_video
807
+ dl3dv: real_video
808
+ wildrgbd: real_video
809
+ waymo: real_video
810
+ robotwin: robot
811
+ droid: robot
812
+ libero: robot
813
+ pointodyssey_train: tracks
814
+ kubric_train: tracks
815
+ molmo_motion: human_video
cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/pointodyssey49_readback.json ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "blockers": [],
3
+ "failures": {},
4
+ "model_forward_or_training": false,
5
+ "sources_read": [
6
+ "pointodyssey_train",
7
+ "scannet"
8
+ ],
9
+ "status": "passed",
10
+ "windows_yaml": "/scratch/shared/engs2907/cosmos_long49_hdepic_20261107/cosmos_framework/configs/omni4d_vgg_stage_one_next_version_all_sources_plus_hdepic_49_frames_ddp_global32_b4_gas4.yaml",
11
+ "windows_yaml_sha256": "ff4f200255a52f69a3f5551d311156258dea1dace5ac4e7af47a6abf266f4ade"
12
+ }
cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/robotwin_readback.json ADDED
@@ -0,0 +1,192 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "caption_coverage": null,
3
+ "dino_feature_shape": [
4
+ 2,
5
+ 3072,
6
+ 13,
7
+ 7,
8
+ 7
9
+ ],
10
+ "frames": 49,
11
+ "model_forward_or_training": false,
12
+ "role_replay": null,
13
+ "rows": [
14
+ {
15
+ "caption_characters": 155,
16
+ "caption_sha256": "093a590922a30b657dbfcdb1b11f2f64a2c84455cdfd573e2fae68500ba2b09e",
17
+ "caption_tier": "source_task_instruction",
18
+ "condition_only_streams": [],
19
+ "frame_ids": [
20
+ 0,
21
+ 4,
22
+ 8,
23
+ 12,
24
+ 16,
25
+ 20,
26
+ 24,
27
+ 28,
28
+ 32,
29
+ 36,
30
+ 40,
31
+ 44,
32
+ 48,
33
+ 52,
34
+ 56,
35
+ 60,
36
+ 64,
37
+ 68,
38
+ 72,
39
+ 76,
40
+ 80,
41
+ 84,
42
+ 88,
43
+ 92,
44
+ 96,
45
+ 100,
46
+ 104,
47
+ 108,
48
+ 112,
49
+ 116,
50
+ 120,
51
+ 124,
52
+ 128,
53
+ 132,
54
+ 136,
55
+ 140,
56
+ 144,
57
+ 148,
58
+ 152,
59
+ 156,
60
+ 160,
61
+ 164,
62
+ 168,
63
+ 172,
64
+ 176,
65
+ 180,
66
+ 184,
67
+ 188,
68
+ 192
69
+ ],
70
+ "labelled": {
71
+ "camera_valid": true,
72
+ "pointmap_valid": true,
73
+ "tracking_video_valid": true
74
+ },
75
+ "sequence": "robotwin/episode_02083",
76
+ "streams": {
77
+ "pointmap_image": {
78
+ "dtype": "torch.float32",
79
+ "finite": true,
80
+ "max": 0.7545167803764343,
81
+ "min": -0.7566211223602295,
82
+ "shape": [
83
+ 49,
84
+ 3,
85
+ 288,
86
+ 512
87
+ ]
88
+ },
89
+ "raymap_camera_image": {
90
+ "dtype": "torch.float32",
91
+ "finite": true,
92
+ "max": 1.0,
93
+ "min": -1.0,
94
+ "shape": [
95
+ 49,
96
+ 3,
97
+ 288,
98
+ 512
99
+ ]
100
+ },
101
+ "rgb": {
102
+ "dtype": "torch.uint8",
103
+ "nonzero_count": 10134820,
104
+ "shape": [
105
+ 49,
106
+ 3,
107
+ 288,
108
+ 512
109
+ ]
110
+ },
111
+ "tracking_video": {
112
+ "dtype": "torch.uint8",
113
+ "nonzero_count": 42483,
114
+ "shape": [
115
+ 49,
116
+ 3,
117
+ 288,
118
+ 512
119
+ ]
120
+ }
121
+ }
122
+ }
123
+ ],
124
+ "source": "robotwin",
125
+ "status": "passed",
126
+ "vgg_commit": "f6992c4b67441dc9778642a252447a2687856116",
127
+ "window_identity": [
128
+ {
129
+ "episode": 2083,
130
+ "frame_ids": [
131
+ 0,
132
+ 4,
133
+ 8,
134
+ 12,
135
+ 16,
136
+ 20,
137
+ 24,
138
+ 28,
139
+ 32,
140
+ 36,
141
+ 40,
142
+ 44,
143
+ 48,
144
+ 52,
145
+ 56,
146
+ 60,
147
+ 64,
148
+ 68,
149
+ 72,
150
+ 76,
151
+ 80,
152
+ 84,
153
+ 88,
154
+ 92,
155
+ 96,
156
+ 100,
157
+ 104,
158
+ 108,
159
+ 112,
160
+ 116,
161
+ 120,
162
+ 124,
163
+ 128,
164
+ 132,
165
+ 136,
166
+ 140,
167
+ 144,
168
+ 148,
169
+ 152,
170
+ 156,
171
+ 160,
172
+ 164,
173
+ 168,
174
+ 172,
175
+ 176,
176
+ 180,
177
+ 184,
178
+ 188,
179
+ 192
180
+ ],
181
+ "native_index": 170,
182
+ "requested_window": 0,
183
+ "sequence": "robotwin/episode_02083",
184
+ "underlying_reported": {
185
+ "requested_native_index": 170
186
+ },
187
+ "window_start": 0
188
+ }
189
+ ],
190
+ "windows": 1,
191
+ "windows_yaml_sha256": "f0b5f4dc89dc4face44830c3b16eb1f0699ca8a1d27a5fb9a98bcb06502064a4"
192
+ }
cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/source_commit.txt ADDED
@@ -0,0 +1 @@
 
 
1
+ 9cb0ceb0167ba45dc89abfe29ae740a59e91ac11
cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/source_windows.yaml ADDED
@@ -0,0 +1,815 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # B4/GAS4 49-frame continuation from the update-14,700 weights-only overlay; no Cityscapes.
2
+ # This is a separate model-only run from an explicit five-stream overlay: 21 sources, total weight 139.
3
+ # The Cityscapes source is deliberately omitted; the remaining real-video sources keep their prior weights.
4
+ # Next Cosmos stage-one version: the owner's weighted mixture at 49 frames (decision 2026-11-07).
5
+ #
6
+ # Weights, not shares. Total 139 over 21 logical sources:
7
+ # synthetic 79 (56.8%) the eleven synthetic sources at the 20,000-update Wan run's weights, unchanged:
8
+ # syn4d 15.2, omniworld_game 8.75, synthverse 11.05, vkitti 6, tartanair 6, dynamic_replica 5,
9
+ # spring 8, midair 3, mvs_synth 3, tartanground 3, paralleldomain4d 10
10
+ # real video 22 (15.8%) ScanNet 8, DL3DV 6, WildRGB-D 4, Waymo 4; Cityscapes excluded.
11
+ # robot 26 (18.7%) RoboTwin 10 (clean-train episodes only), DROID 10 (depth and camera as inputs only),
12
+ # LIBERO 6 (official training demonstrations only); no action stream in stage one
13
+ # tracks 6 ( 4.3%) PointOdyssey 3, Kubric 3
14
+ # human video 6 ( 4.3%) MolmoMotion including HD-EPIC; DINO and sparse tracking video are targets, camera is input-only, depth absent
15
+ # DINO is a generated target on the real-video sources (ScanNet, DL3DV, WildRGB-D, DROID, Waymo): it is computed from the
16
+ # RGB, so it is exact on real data, and the future-feature forecasts the DINO-Foresight and VFMF comparisons make would
17
+ # otherwise never train on real video. This changes the first launch's roles: the owner's rule is that geometry is
18
+ # condition-only on real data, and that rule still holds for depth and camera, which stay inputs only here.
19
+ # Waymo trains on the processed training split only (waymo.training.pickle.gz, SHA-256 6abd3476...: 798 sequences,
20
+ # 158,081 frames; validation never enters training) through the pinned sparse-track reader's waymo stanza: sparse LiDAR
21
+ # depth, the camera and the RGB; no tracking video; strides 1-4 at 21 frames.
22
+ # No segmentation stream, no Cityscapes, no YAM.
23
+ #
24
+ # The VGG window block below is read by the pinned VGG checkout (vgg_commit) through VggStageOneWindowDataset: the
25
+ # synthetic sources are the recorded 500K run's stanzas (omni4d_vgg_sparse_tracking_synthetic_continuation_sources.yaml,
26
+ # SHA-256 ae49ec6d...) with the window length 49 and missing_caption: error. RoboTwin and LIBERO are read through
27
+ # robot_sources (RobotEpisodeStageOneDataset) because the pinned VGG checkout has no Track4D reader for them.
28
+ # The tracking-video query draw keeps the recorded 10-pixel grid (no query_count_range).
29
+ #
30
+ # Train membership of the real-video sources (official split files, 2026-11-05):
31
+ # DL3DV the pinned reader ignores `split` and indexes every processed scene, 85 of which are official
32
+ # DL3DV-Benchmark (test) scenes. The 141 official benchmark scene hashes (Hugging Face
33
+ # DL3DV/DL3DV-Benchmark, dataset revision 9684e838) are in omni4d_dl3dv_official_benchmark_scene_hashes.txt;
34
+ # the 87 of them present in the processed 1K-7K root are added to the stanza's exclude_sequence_ids below
35
+ # (source_reader_overrides), leaving about 5,898 scenes.
36
+ # WildRGB-D the pinned reader reads <category>/selected_seqs_train.json; those 2,244 processed train scenes are all in
37
+ # the official train-of-both lists (nvs_list.json and camera_eval_list.json of hongchi/wildrgbd, CUT3R's
38
+ # get_set_list rule) and overlap the official NVS and camera-evaluation validation scenes in none.
39
+ #
40
+ # launch_blockers names every evidence gap that stops the launch. A blocked source keeps its weight; composition
41
+ # (omni4d_native_config.make_config) and the window dataset both refuse to start while the list is not empty. A robot
42
+ # source additionally needs a passing real-window readback receipt (readback_receipt). WildRGB-D and Kubric train on
43
+ # caption text converted by tools/omni4d/convert_stage_one_caption_texts.py (text-only products: no umT5 cache exists).
44
+ #
45
+ # VGG pin: the 500K run's f5a59cd51 plus the window-reader updates, processed-release EgoDex support, and the
46
+ # native2048 Cityscapes caption-reader migration at f6992c4b. The approved raw EgoDex window and Cityscapes caption
47
+ # product readbacks used this exact pin. City captions point to the completed 2026-11-06 converted text product below.
48
+ # One makes the DROID and LIBERO robot window length the data node's num_frames: the pinned window class and the LIBERO
49
+ # painter fixed the clock at 33 rows in five places, one a literal inside the painter's constructor; a project adapter
50
+ # could only rebind module globals and fake attributes, so the smallest safe change is in the VGG source, with the
51
+ # recorded 33-row nodes reading as before. Two DROID commits (cherry-picked from the VGG main stack, f5a59cd is their
52
+ # ancestor) keep a window whose FK arm has no visible frame-0 query, with its source motion absent, instead of raising.
53
+ # The fourth adds the stanza key position_windows (src/vggforecast/data/position_windows.py): each listed position range is
54
+ # one window and the stride and start are drawn inside it. ScanNet and PointOdyssey use it (source_reader_overrides,
55
+ # caption_windows: true) with the local caption windows of their products, so every window they draw lies inside one
56
+ # caption window and the strict caption join never redraws. The fifth adds a waymo stanza to the stage-one real-source config
57
+ # (no source uses it yet: every real source leaves it out). The pin is a separate checkout (a git worktree of the earlier pin
58
+ # directory, which a running training still reads at 9cb6a626a), so the earlier pin's files never change under a run.
59
+ vgg_repository: /scratch/shared/engs2907/vgg_pins/vgg_forecast_4d_cosmos_source_20261106_f6992c4
60
+ vgg_commit: f6992c4b67441dc9778642a252447a2687856116
61
+ recipe_source_config: /scratch/shared/engs2907/logs/vgg_forecast_4d/train/flexpi_four_d_pretraining_wan22_5b_separate_streams_full_parameters_video_depth_camera_dino_and_sparse_tracking_video_from_the_15000_update_four_d_run_20261101/attempt02/run/config.yaml
62
+ recipe_source_config_sha256: 72590f4598cae1b329c31b3cb45d059bdaeca23642faa10ca3f31372f9ebc0aa
63
+ window_block:
64
+ _target_: four_d_windows.FourDPretrainingWindows
65
+ sources:
66
+ - name: four_d_pretraining_relaunch_syn4d_sources
67
+ kind: track4d_source_config
68
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_rendered_depth_sources_21_frames_512
69
+ weight: 15.2
70
+ left_out_sources:
71
+ - omniworld_game
72
+ - tartanair
73
+ - mvs_synth
74
+ condition_only_streams: []
75
+ - name: four_d_pretraining_relaunch_omniworld_game_source
76
+ kind: track4d_source_config
77
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_rendered_depth_sources_21_frames_512
78
+ weight: 8.75
79
+ left_out_sources:
80
+ - syn4d_v1
81
+ - syn4d_v2_mixed
82
+ - syn4d_v2_mixed_no_sim
83
+ - tartanair
84
+ - mvs_synth
85
+ condition_only_streams: []
86
+ - name: four_d_pretraining_relaunch_synthverse_source
87
+ kind: track4d_source_config
88
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_synthverse_dynamic_replica_vkitti_midair_spring_sources_without_midair_quality_masks_21_frames_512
89
+ weight: 11.05
90
+ left_out_sources:
91
+ - dynamic_replica
92
+ - vkitti
93
+ - midair
94
+ - spring
95
+ condition_only_streams: []
96
+ - name: four_d_pretraining_relaunch_vkitti_source
97
+ kind: track4d_source_config
98
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_synthverse_dynamic_replica_vkitti_midair_spring_sources_without_midair_quality_masks_21_frames_512
99
+ weight: 6.0
100
+ left_out_sources:
101
+ - synthverse_depth
102
+ - dynamic_replica
103
+ - midair
104
+ - spring
105
+ condition_only_streams: []
106
+ - name: four_d_pretraining_relaunch_tartanair_sky_then_far_percentile_source
107
+ kind: track4d_source_config
108
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_sky_and_far_percentile_depth_validity_rules_21_frames_512
109
+ weight: 6.0
110
+ left_out_sources:
111
+ - dynamic_replica
112
+ - spring
113
+ - midair
114
+ - mvs_synth
115
+ - tartanground
116
+ - paralleldomain4d
117
+ condition_only_streams: []
118
+ - name: four_d_pretraining_relaunch_dynamic_replica_stored_depth_source
119
+ kind: track4d_source_config
120
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_sky_and_far_percentile_depth_validity_rules_21_frames_512
121
+ weight: 5.0
122
+ left_out_sources:
123
+ - tartanair
124
+ - spring
125
+ - midair
126
+ - mvs_synth
127
+ - tartanground
128
+ - paralleldomain4d
129
+ condition_only_streams: []
130
+ - name: four_d_pretraining_relaunch_spring_stored_depth_source
131
+ kind: track4d_source_config
132
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_sky_and_far_percentile_depth_validity_rules_21_frames_512
133
+ weight: 8.0
134
+ left_out_sources:
135
+ - tartanair
136
+ - dynamic_replica
137
+ - midair
138
+ - mvs_synth
139
+ - tartanground
140
+ - paralleldomain4d
141
+ condition_only_streams: []
142
+ - name: four_d_pretraining_relaunch_midair_within_200_meters_source
143
+ kind: track4d_source_config
144
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_sky_and_far_percentile_depth_validity_rules_21_frames_512
145
+ weight: 3.0
146
+ left_out_sources:
147
+ - tartanair
148
+ - dynamic_replica
149
+ - spring
150
+ - mvs_synth
151
+ - tartanground
152
+ - paralleldomain4d
153
+ condition_only_streams: []
154
+ - name: four_d_pretraining_relaunch_mvs_synth_1000_then_far_percentile_source
155
+ kind: track4d_source_config
156
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_sky_and_far_percentile_depth_validity_rules_21_frames_512
157
+ weight: 3.0
158
+ left_out_sources:
159
+ - tartanair
160
+ - dynamic_replica
161
+ - spring
162
+ - midair
163
+ - tartanground
164
+ - paralleldomain4d
165
+ condition_only_streams: []
166
+ - name: four_d_pretraining_relaunch_tartanground_sky_then_far_percentile_source
167
+ kind: track4d_source_config
168
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_sky_and_far_percentile_depth_validity_rules_21_frames_512
169
+ weight: 3.0
170
+ left_out_sources:
171
+ - tartanair
172
+ - dynamic_replica
173
+ - spring
174
+ - midair
175
+ - mvs_synth
176
+ - paralleldomain4d
177
+ condition_only_streams: []
178
+ - name: four_d_pretraining_relaunch_paralleldomain4d_sky_then_far_percentile_source
179
+ kind: track4d_source_config
180
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_sky_and_far_percentile_depth_validity_rules_21_frames_512
181
+ weight: 10.0
182
+ left_out_sources:
183
+ - tartanair
184
+ - dynamic_replica
185
+ - spring
186
+ - midair
187
+ - mvs_synth
188
+ - tartanground
189
+ condition_only_streams: []
190
+ - name: scannet
191
+ kind: track4d_source_config
192
+ config: train/mot/stage_one_real_and_native_track_source_windows_9_frames_288_512
193
+ weight: 8.0
194
+ left_out_sources:
195
+ - dl3dv
196
+ - wildrgbd
197
+ - pointodyssey_train
198
+ - kubric_train
199
+ - droid
200
+ - waymo
201
+ condition_only_streams:
202
+ - pointmap
203
+ - camera
204
+ - name: dl3dv
205
+ kind: track4d_source_config
206
+ config: train/mot/stage_one_real_and_native_track_source_windows_9_frames_288_512
207
+ weight: 6.0
208
+ left_out_sources:
209
+ - scannet
210
+ - wildrgbd
211
+ - pointodyssey_train
212
+ - kubric_train
213
+ - droid
214
+ - waymo
215
+ condition_only_streams:
216
+ - pointmap
217
+ - camera
218
+ - name: wildrgbd
219
+ kind: track4d_source_config
220
+ config: train/mot/stage_one_real_and_native_track_source_windows_9_frames_288_512
221
+ weight: 4.0
222
+ left_out_sources:
223
+ - dl3dv
224
+ - scannet
225
+ - pointodyssey_train
226
+ - kubric_train
227
+ - droid
228
+ - waymo
229
+ condition_only_streams:
230
+ - pointmap
231
+ - camera
232
+ - name: waymo
233
+ kind: track4d_source_config
234
+ config: train/mot/stage_one_real_and_native_track_source_windows_9_frames_288_512
235
+ weight: 4.0
236
+ left_out_sources:
237
+ - dl3dv
238
+ - scannet
239
+ - wildrgbd
240
+ - pointodyssey_train
241
+ - kubric_train
242
+ - droid
243
+ condition_only_streams:
244
+ - pointmap
245
+ - camera
246
+ - name: droid
247
+ kind: track4d_source_config
248
+ config: train/mot/stage_one_real_and_native_track_source_windows_9_frames_288_512
249
+ weight: 10.0
250
+ left_out_sources:
251
+ - dl3dv
252
+ - scannet
253
+ - wildrgbd
254
+ - pointodyssey_train
255
+ - kubric_train
256
+ - waymo
257
+ condition_only_streams:
258
+ - pointmap
259
+ - camera
260
+ - name: pointodyssey_train
261
+ kind: track4d_source_config
262
+ config: train/mot/stage_one_real_and_native_track_source_windows_9_frames_288_512
263
+ weight: 3.0
264
+ left_out_sources:
265
+ - dl3dv
266
+ - scannet
267
+ - wildrgbd
268
+ - kubric_train
269
+ - droid
270
+ - waymo
271
+ condition_only_streams: []
272
+ - name: kubric_train
273
+ kind: track4d_source_config
274
+ config: train/mot/stage_one_real_and_native_track_source_windows_9_frames_288_512
275
+ weight: 3.0
276
+ left_out_sources:
277
+ - dl3dv
278
+ - scannet
279
+ - wildrgbd
280
+ - pointodyssey_train
281
+ - droid
282
+ - waymo
283
+ condition_only_streams:
284
+ - pointmap
285
+ frames: 49
286
+ raster_hw:
287
+ - 288
288
+ - 512
289
+ raymap_camera:
290
+ origin_fraction: 0.3333333333333333
291
+ origin_range: 1.0
292
+ decode_margin: 16
293
+ origin_box_grid: 1
294
+ pointmap_log_factor: 0.4
295
+ pointmap_hole_fill:
296
+ synthverse_depth: far_clip
297
+ syn4d: far_clip
298
+ omniworld_game: far_clip
299
+ dynamic_replica: nearest_labelled_depth_of_3x3_neighbours
300
+ tartanair: far_clip
301
+ dl3dv: far_clip
302
+ scannet: telea_log_depth
303
+ virtual_kitti2: far_clip
304
+ midair: far_clip
305
+ pointodyssey: far_clip
306
+ mvs_synth: far_clip
307
+ wildrgbd: telea_log_depth
308
+ hypersim: far_clip
309
+ spring: far_clip
310
+ kubric49: far_clip
311
+ kubric81: far_clip
312
+ droid_stereo_depth: telea_log_depth
313
+ tartanground: far_clip
314
+ paralleldomain4d: far_clip
315
+ syn4d_v2: far_clip
316
+ waymo: far_clip
317
+ pointmap_gauge: window_labelled_median_depth
318
+ geometry_image: genception_log_depth
319
+ context_len: 512
320
+ batch_size: 4
321
+ stream_seed: 42
322
+ epoch_windows: 32000000
323
+ val_set_proportion: 0.0
324
+ resume_state: null
325
+ gradient_accumulation_steps: 4
326
+ loader_workers: 12
327
+ loader_opencv_threads: 1
328
+ window_captions:
329
+ caption_dropout: 0.1
330
+ missing_caption: error
331
+ caches:
332
+ syn4d_v1_gemini_holo4d:
333
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/syn4d_v1_gemini_holo4d_local_and_global_with_repaired_globals_20261031/captions_t5_emb.safetensors
334
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/syn4d_v1_gemini_holo4d_local_and_global_with_repaired_globals_20261031/captions_t5_emb_meta.json
335
+ syn4d_v2_gemini_holo4d:
336
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/syn4d_v2_gemini_holo4d_20261030/captions_t5_emb.safetensors
337
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/syn4d_v2_gemini_holo4d_20261030/captions_t5_emb_meta.json
338
+ tartanair_gemini_holo4d:
339
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/tartanair_gemini_holo4d/captions_t5_emb.safetensors
340
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/tartanair_gemini_holo4d/captions_t5_emb_meta.json
341
+ mvs_synth_gemini_holo4d:
342
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/mvs_synth_gemini_holo4d/captions_t5_emb.safetensors
343
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/mvs_synth_gemini_holo4d/captions_t5_emb_meta.json
344
+ omniworld_game_gemini_local_official_text_chunks_and_gemini_global:
345
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/omniworld_game_gemini_local_official_text_chunks_and_gemini_global_20261030b/captions_t5_emb.safetensors
346
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/omniworld_game_gemini_local_official_text_chunks_and_gemini_global_20261030b/captions_t5_emb_meta.json
347
+ synthverse_gemini_holo4d:
348
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/synthverse_gemini_holo4d_local96_and_global_20261031/captions_t5_emb.safetensors
349
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/synthverse_gemini_holo4d_local96_and_global_20261031/captions_t5_emb_meta.json
350
+ dynamic_replica_gemini_holo4d:
351
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/dynamic_replica_gemini_holo4d/captions_t5_emb.safetensors
352
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/dynamic_replica_gemini_holo4d/captions_t5_emb_meta.json
353
+ vkitti_gemini_holo4d:
354
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/vkitti_gemini_holo4d/captions_t5_emb.safetensors
355
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/vkitti_gemini_holo4d/captions_t5_emb_meta.json
356
+ midair_gemini_holo4d:
357
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/midair_gemini_holo4d_20261030/captions_t5_emb.safetensors
358
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/midair_gemini_holo4d_20261030/captions_t5_emb_meta.json
359
+ spring_gemini_holo4d:
360
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/spring_gemini_holo4d/captions_t5_emb.safetensors
361
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/spring_gemini_holo4d/captions_t5_emb_meta.json
362
+ tartanground_gemini_holo4d:
363
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/tartanground_gemini_holo4d_20261030/captions_t5_emb.safetensors
364
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/tartanground_gemini_holo4d_20261030/captions_t5_emb_meta.json
365
+ paralleldomain4d_gemini_holo4d:
366
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/paralleldomain4d_gemini_holo4d_20261030/captions_t5_emb.safetensors
367
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/paralleldomain4d_gemini_holo4d_20261030/captions_t5_emb_meta.json
368
+ scannet_gemini_holo4d:
369
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/.scannet_gemini_holo4d.version-tovvjwai/captions_t5_emb.safetensors
370
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/.scannet_gemini_holo4d.version-tovvjwai/captions_t5_emb_meta.json
371
+ pointodyssey_gemini_holo4d:
372
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/.pointodyssey_gemini_holo4d.version-6vfyvsbi/captions_t5_emb.safetensors
373
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/.pointodyssey_gemini_holo4d.version-6vfyvsbi/captions_t5_emb_meta.json
374
+ droid_vp_gemini_holo4d:
375
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/.droid_vp_gemini_holo4d.version-e51r5urf/captions_t5_emb.safetensors
376
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/.droid_vp_gemini_holo4d.version-e51r5urf/captions_t5_emb_meta.json
377
+ dl3dv_gemini_holo4d:
378
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/dl3dv_gemini_holo4d_20261031/captions_t5_emb.safetensors
379
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/dl3dv_gemini_holo4d_20261031/captions_t5_emb_meta.json
380
+ wildrgbd_text_captions:
381
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/cosmos_caption_texts_20261105/wildrgbd_text_captions/captions_t5_emb.safetensors
382
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/cosmos_caption_texts_20261105/wildrgbd_text_captions/captions_t5_emb_meta.json
383
+ kubric49_text_captions:
384
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/cosmos_caption_texts_20261105/kubric49_text_captions/captions_t5_emb.safetensors
385
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/cosmos_caption_texts_20261105/kubric49_text_captions/captions_t5_emb_meta.json
386
+ waymo_gemini_holo4d:
387
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/cosmos_caption_texts_20261105/waymo_gemini_holo4d/captions_t5_emb.safetensors
388
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/cosmos_caption_texts_20261105/waymo_gemini_holo4d/captions_t5_emb_meta.json
389
+ stanzas:
390
+ syn4d_v1:
391
+ product: syn4d_v1_gemini_holo4d
392
+ join: local_covering_then_global
393
+ syn4d_v2_mixed:
394
+ product: syn4d_v2_gemini_holo4d
395
+ join: local_covering_then_global
396
+ syn4d_v2_mixed_no_sim:
397
+ product: syn4d_v2_gemini_holo4d
398
+ join: local_covering_then_global
399
+ tartanair:
400
+ product: tartanair_gemini_holo4d
401
+ join: local_covering
402
+ mvs_synth:
403
+ product: mvs_synth_gemini_holo4d
404
+ join: local_covering
405
+ omniworld_game:
406
+ product: omniworld_game_gemini_local_official_text_chunks_and_gemini_global
407
+ join: local_covering_then_second_local_covering_then_global
408
+ pointodyssey_test: none
409
+ kubric49_validation: none
410
+ wildrgbd_test: none
411
+ vkitti_test:
412
+ product: vkitti_gemini_holo4d
413
+ join: local_covering_then_global
414
+ droid_held_out: none
415
+ synthverse_depth:
416
+ product: synthverse_gemini_holo4d
417
+ join: local_covering_then_global
418
+ dynamic_replica:
419
+ product: dynamic_replica_gemini_holo4d
420
+ join: global
421
+ vkitti:
422
+ product: vkitti_gemini_holo4d
423
+ join: local_covering_then_global
424
+ midair:
425
+ product: midair_gemini_holo4d
426
+ join: local_covering_then_global
427
+ spring:
428
+ product: spring_gemini_holo4d
429
+ join: global
430
+ tartanground:
431
+ product: tartanground_gemini_holo4d
432
+ join: local_covering_then_global
433
+ paralleldomain4d:
434
+ product: paralleldomain4d_gemini_holo4d
435
+ join: global
436
+ paralleldomain4d_validation:
437
+ product: paralleldomain4d_gemini_holo4d
438
+ join: global
439
+ scannet:
440
+ product: scannet_gemini_holo4d
441
+ join: local_containing_then_global
442
+ pointodyssey_train:
443
+ product: pointodyssey_gemini_holo4d
444
+ join: local_containing_then_global
445
+ droid:
446
+ product: droid_vp_gemini_holo4d
447
+ join: global
448
+ dl3dv:
449
+ product: dl3dv_gemini_holo4d
450
+ join: local_containing_then_global
451
+ wildrgbd:
452
+ product: wildrgbd_text_captions
453
+ join: local_containing_then_global
454
+ kubric_train:
455
+ product: kubric49_text_captions
456
+ join: global_required
457
+ waymo:
458
+ product: waymo_gemini_holo4d
459
+ join: local_containing_then_global
460
+ training_scene_exclusions:
461
+ reason: rendered depth passes through glass (raw EXR depth through glass panes at 190-260 m), owner 10-29; evidence /scratch/shared/engs2907/logs/vgg_forecast_4d/tools/syn4d_glass_depth_check_20261029;
462
+ scifiroom_bald added 10-30 from /scratch/shared/engs2907/logs/vgg_forecast_4d/tools/depth_clipping_audit_20261030
463
+ stanzas:
464
+ syn4d_v1:
465
+ - bigoffice_v1
466
+ - lab_bald
467
+ - office_bald
468
+ - scifiroom_bald
469
+ syn4d_v2_mixed:
470
+ - citynight
471
+ - mall
472
+ - metro
473
+ syn4d_v2_mixed_no_sim:
474
+ - citynight
475
+ - mall
476
+ - metro
477
+ tracking_video:
478
+ query_grid_px: 10
479
+ disc_radius_px: 2
480
+ occluded_style: ring
481
+ ring_radius_px: 3
482
+ ring_width_px: 1
483
+ xyz_normalization: quantile_0.02_0.98
484
+ colour_floor: 0.2
485
+ min_frame_zero_points: 16
486
+ surface_test:
487
+ relative_tolerance: 0.1
488
+ absolute_tolerance: 0.05
489
+ stanza_rules:
490
+ tartanair: static_scene_points
491
+ midair: static_scene_points
492
+ mvs_synth: static_scene_points
493
+ tartanground: static_scene_points
494
+ syn4d_v1: tracked_points
495
+ syn4d_v2_mixed: tracked_points
496
+ syn4d_v2_mixed_no_sim: tracked_points
497
+ omniworld_game: no_tracking_video
498
+ synthverse_depth: no_tracking_video
499
+ dynamic_replica: tracked_points
500
+ vkitti: no_tracking_video
501
+ spring: no_tracking_video
502
+ paralleldomain4d: no_tracking_video
503
+ scannet: static_scene_points
504
+ wildrgbd_test: static_scene_points
505
+ pointodyssey_test: tracked_points
506
+ kubric49_validation: no_tracking_video
507
+ vkitti_test: no_tracking_video
508
+ droid_held_out: no_tracking_video
509
+ paralleldomain4d_validation: no_tracking_video
510
+ dynamic_replica_test: tracked_points
511
+ pointodyssey_train: tracked_points
512
+ kubric_train: tracked_points
513
+ droid: tracked_points
514
+ dl3dv: static_scene_points
515
+ wildrgbd: static_scene_points
516
+ waymo: no_tracking_video
517
+ stanza_track_reading:
518
+ syn4d_v1:
519
+ stanza:
520
+ capability: dynamic_cross_time
521
+ supervision:
522
+ - pointmap
523
+ - source_motion
524
+ annotation_provenance: syn4d_mesh_surface_forward_track_from_frame0
525
+ allow_missing_tracks: false
526
+ kwargs:
527
+ return_tracks: true
528
+ depth_only: false
529
+ metadata_root: /vgg/chenzhao/syn4d_metadata
530
+ wrapped_face_distance_thresh: 0.1
531
+ query_inconsistent_distance_thresh: 0.1
532
+ pointodyssey_test:
533
+ stanza:
534
+ supervision:
535
+ - pointmap
536
+ - source_motion
537
+ kwargs:
538
+ return_tracks: true
539
+ dynamic_replica_test:
540
+ stanza:
541
+ supervision:
542
+ - pointmap
543
+ - source_motion
544
+ kwargs:
545
+ return_tracks: true
546
+ syn4d_v2_mixed:
547
+ loader: syn4d_v2
548
+ stanza:
549
+ capability: dynamic_cross_time
550
+ supervision:
551
+ - pointmap
552
+ - source_motion
553
+ annotation_provenance: syn4d_v2_mesh_surface_forward_track_from_frame0
554
+ allow_missing_tracks: false
555
+ kwargs:
556
+ return_tracks: true
557
+ depth_only: false
558
+ metadata_root: /vgg/chenzhao/syn4d_metadata
559
+ tracking_format: safetensor
560
+ mask_bg: false
561
+ trusted_background_depth_max: 80.0
562
+ wrapped_face_distance_thresh: 0.1
563
+ query_inconsistent_distance_thresh: 0.1
564
+ syn4d_v2_mixed_no_sim:
565
+ loader: syn4d_v2
566
+ stanza:
567
+ capability: dynamic_cross_time
568
+ supervision:
569
+ - pointmap
570
+ - source_motion
571
+ annotation_provenance: syn4d_v2_mesh_surface_forward_track_from_frame0
572
+ allow_missing_tracks: false
573
+ kwargs:
574
+ return_tracks: true
575
+ depth_only: false
576
+ metadata_root: /vgg/chenzhao/syn4d_metadata
577
+ tracking_format: safetensor
578
+ mask_bg: false
579
+ trusted_background_depth_max: 80.0
580
+ wrapped_face_distance_thresh: 0.1
581
+ query_inconsistent_distance_thresh: 0.1
582
+ dynamic_replica:
583
+ stanza:
584
+ supervision:
585
+ - pointmap
586
+ - source_motion
587
+ kwargs:
588
+ return_tracks: true
589
+ pointodyssey_train:
590
+ stanza:
591
+ allow_missing_tracks: false
592
+ capability: dynamic_cross_time
593
+ supervision:
594
+ - pointmap
595
+ - source_motion
596
+ kwargs:
597
+ return_tracks: true
598
+ kubric_train:
599
+ stanza:
600
+ allow_missing_tracks: false
601
+ capability: dynamic_cross_time
602
+ supervision:
603
+ - pointmap
604
+ - source_motion
605
+ kwargs:
606
+ return_tracks: true
607
+ droid:
608
+ stanza:
609
+ allow_missing_tracks: false
610
+ capability: dynamic_cross_time
611
+ supervision:
612
+ - pointmap
613
+ - source_motion
614
+ kwargs:
615
+ return_tracks: true
616
+ source_reader_overrides:
617
+ dl3dv:
618
+ stanza: dl3dv
619
+ exclude_sequence_ids:
620
+ - 1K/0a1b7c20a92c43c6b8954b1ac909fb2f0fa8b2997b80604bc8bbec80a1cb2da3
621
+ - 1K/14eb48a50e37df548894ab6d8cd628a21dae14bbe6c462e894616fc5962e6c49
622
+ - 1K/1da888bdedfc9629c0fa9f82cf3f5d96f8103baee0ff64d9311aea1224a9f2ae
623
+ - 1K/2991a75d1f0fb2b7a891d424fc2459bb954b61599b52e74a056a7acfb2998b77
624
+ - 1K/2cbfe28643b6636f9c70813cae7625aa858a352109493ac70fb429ce94dd55b3
625
+ - 1K/5a27c00f525f5484b7aa0d0107051253bf7bb395bde0b3a4486ff8c52fb4965e
626
+ - 1K/6d221625614d0515fa44d59a8797a49b64f86bc775e66be22f07a66aa0a0cb3a
627
+ - 1K/917e9c8985d353b0ee4c281f11fa84eb8550814562670cbb82cd0ec9c1194fe7
628
+ - 1K/918c8dad730c3b804306c5da8486124be4aa0612e85fb825338fd350c912e1b0
629
+ - 1K/a17a984ca90a9b5840fdf85b15104b0d18e25975981c1aa90fcdfd6eeeb285f3
630
+ - 1K/a726c1112ad026efc92f59f196c0698e67abd12a04135a7ff6349b1d31b089eb
631
+ - 1K/adb95f29c15b8a89e40c17d10596679be09d7719c6f08a453facd506d056ea8b
632
+ - 1K/d1b3a0b37acc72207295b05e8c359f8b43bf879eb882c60747cfc6d3c7d8efa3
633
+ - 1K/eb4cf52988f805e6fce11d1b239fa9de32eb157364cff06ebac0aa50e0a46567
634
+ - 2K/032dee9fb0a8bc1b90871dc5fe950080d0bcd3caf166447f44e60ca50ac04ec7
635
+ - 2K/0853979305f7ecb80bd8fc2c8df916410d471ef04ed5f1a64e9651baa41d7695
636
+ - 2K/093ef327b4e4f9d4ee52c02a354a53558a8652157fb0d58f3b4a708734afb334
637
+ - 2K/0f8ac521439691fe429e9efbed8d5ded1cee35bcf52d47731fb60d5a2ff661a7
638
+ - 2K/1d6a9ed47cce39fd1c4d18f776bcc97e507b81bde921ca596bd91b0b02b5e414
639
+ - 2K/26fd23358fa11fff0fb3180ef0b65591b486e20dcf753ce4a7aae49a37e370c7
640
+ - 2K/280abf7bd93b81b077af1db638229dbb09869052fa0b7b57c81c94d2db893829
641
+ - 2K/2beaca318994c25409dcbb6d0bdd96c3620f2f18aec44ea3f20edd302f18ca78
642
+ - 2K/341b4ff3dfd3d377d7167bd81f443bedafbff003bf04881b99760fc0aeb69510
643
+ - 2K/374ffd0c5ff8e544f588f1da3ce0d5b69dc39ff3af95d7ccec480fe5e533c45e
644
+ - 2K/41036716da7efda334c1d434c4141d15642e0e02f881a01b6c8c36f8bea64c45
645
+ - 2K/484c0aca40c1087f196d2caf3c8cb463864c08e60b98326ca02eb6dae8ce4984
646
+ - 2K/4ff8650b5c1140e98d957fbd1abbe2cfbbb416b37cb9f06f7f376956703801bb
647
+ - 2K/513e4ea2e8477b06f2b32417d5c243aec71d491cb0596a60e9fe304c635f20a1
648
+ - 2K/8b9fb9d9f10e8c64d5034be69809465753b8ba88ef12da82afd47d38ee789934
649
+ - 2K/91afb9910b042f7185c2b8e4b6b24b5785dae4542617b3f8005b5492f6d123f7
650
+ - 2K/ba55c875d20c34ee85ffc72264c4d77710852e5fb7d9ce4b9c26a8442850e98f
651
+ - 2K/c076929db6501cf7ebe386c70e6d77ea3af844a745e794f2ec17c981c465a69b
652
+ - 2K/ceb252f5d4419510655cf9ed7afbf3e8e688825f798d80414c7715dc8ace153a
653
+ - 2K/df29c225863d0173e060b1f3fde4e6eb5a57a4e4edabd397026ba1a028aeb39e
654
+ - 2K/ff592398657b3dfe94153332861985194a3e3c9d199c4a3a27a0ce4038e81ade
655
+ - 3K/35872363e17af5d173b6a0b09fcf5de94627ad5dc5f8a9ad4c579f3e70b4797a
656
+ - 3K/387eeb925ba47af5db9ce6102aaf6b07e6d3573b7e7568dda4cef22d5ab1342f
657
+ - 3K/3bb894d1933f3081134ad2d40e54de5f0636bd8b502b0a8561873bb63b0dce85
658
+ - 3K/5c3af581028068a3c402c7cbe16ecf9471ddf2897c34ab634b7b1b6cf81aba00
659
+ - 3K/669c36225b273e22ce08364f166dce910b708d7abd89e58463a15d2790b534d5
660
+ - 3K/66fd66cbed40a2d11ecd17ae6d6d60cd60d476616864ecdd351fe69e644f30a1
661
+ - 3K/70eac6ff18a1daae4eeccc5eb18723eaf7e029a77e79c79d9573f7bac59ba92b
662
+ - 3K/75fbbe467353cfdeb93bc1872144855576797be83a0c71c51155ee6b355d6d5e
663
+ - 3K/800cf88687a79c232700454c2f5a402e771bc7157037a17b0ce93da365b25575
664
+ - 3K/8cb2e97d26a639f05a571476240a8fa86988e6853f0f13cc05830d1578002aad
665
+ - 3K/9cbc5548643ca9cfe797f871bac58bccd02f0510c49b7cb3697b38dfd780a508
666
+ - 3K/a62f9a1c6362e7c8a3440261ea0921a0a4b60cf5822bd19579836694f26b3abd
667
+ - 3K/adf35184a12d4cfa3f4248b87aa5adb4f39f179df460d6d76136e13d37299a2a
668
+ - 3K/af0d7039e64af3959210c235e59425896ac064f4e0f721c437bbd2cf1d8ea910
669
+ - 3K/b3bf9079b442c681749f6bd95cfb0659c6d844f2228b2f4fa4b602a0e3627b9a
670
+ - 3K/b6d1134cb03c855d246f916beaa2d3d01290cf47937a5176011ff39b850ec615
671
+ - 3K/d8de66037bd03dd0d39d54f9978bac3318d912e30e22c21e1ada82a98ed48c53
672
+ - 3K/ddfdcfdf02d53dd89a055cd9e1ed30272ab3f57c7560ac2c3761bb01d76a7882
673
+ - 3K/e8ce51b6abfe05bf8dca47e29c8be6c1e6de27a8c9fece7a121400b931b2ca0f
674
+ - 3K/e9360e7a89bee835dc847cf8796093e634b759ff582558788dcfe8326f6e8901
675
+ - 3K/f7aaea9ac683bb1945cfeb1daa9abca3375249e0d19281a89a3f575040835464
676
+ - 4K/56452d9cd9e2e034764132a57058517ab085e1352f9664d2d1e10e98741334fc
677
+ - 4K/6d81c5ab0d480fd43d78b75ff372a8113ad38e2c03f1d69627c009883054d4c2
678
+ - 4K/9641a1ed7963ce5ca734cff3e6ccea3dfa8bcb0b0a3ff78f65d32a080de2d71e
679
+ - 4K/d3812aad538261e7f73c75762ff55f23b468bcc76f376d52ac86ca6cf3c44b4b
680
+ - 5K/0a485338bbdaf19ba9090b874bb36ef0599a9c9a475a382c22903cf5981c6ea6
681
+ - 5K/119fd56d3797e2d349ca64ddcc5851463cd13b5974b5b2e4566ed5cf7e02e6c1
682
+ - 5K/c37726ce770ac50a2cf5c0f43022f0268e26da0d777cd8e3a3418c4eed03fd94
683
+ - 5K/dac9796dd69e1c25277e29d7e30af4f21e3b575d62a0a208c2b3d1211e2d5d77
684
+ - 6K/06da796666297fe4c683c231edf56ec00148a6a52ab5bb159fe1be31f53a58df
685
+ - 6K/7a9f97660be8f2421b37dd413c898360532f17c6447c11515e8aded2d80eb99d
686
+ - 6K/7da3db99059a436adf537bc23309a6a4db9e011696d086f3a64037f7497d9df7
687
+ - 6K/e5684b3292bfd77db297839fc37ee4cce7fd59775af1a6a4827e3b4f59c036d3
688
+ - 6K/ec1e44d4dc0f8fa77610866495f9297a7f82158c43e1777668b84fd4b736c7bc
689
+ - 6K/ef59aac437132bfc1dd45a7e1f8e4800978e7bb28bf98c4428d26fb3e1da3e90
690
+ - 6K/f477ffc4b398bed8e0d921f0fba9825ca63f317381c535c84be23be991ae1d7a
691
+ - 7K/0569e83fdc248a51fc0ab082ce5e2baff15755c53c207f545e6d02d91f01d166
692
+ - 7K/073f5a9b983ced6fb28b23051260558b165f328a16b2d33fe20585b7ee4ad561
693
+ - 7K/183dd248f6a86e07c5adf9de8ee2d0abe45b1216331c03678e89634c2e9b1c7f
694
+ - 7K/1ba74c22670ad047981441581d00f26f4a148d1010bcac7468c615adf5fa4d5d
695
+ - 7K/389a460ca1995e0658e85fe8e6b520b4e88b370cd6710dfe728b1564bba31aee
696
+ - 7K/493816813d2d6d248eb3c2b0b77b63e54235266e9a06e270fd0d282f13960493
697
+ - 7K/4ae797d07b6d1644c9db6919c8cc8c0d28d72be45108ac7a3abf8dc21b599d83
698
+ - 7K/50c46cf8b8b22c8d2ffdef8964b05ddbceaef312c9a9ff331d1ecebfd223f72a
699
+ - 7K/565553aa894be621e8b4773cac288e60ad0c2cf7edb621be62b348c9a0f78380
700
+ - 7K/599ca3e04cae3ec83affc426af7d0d7ab36eb91cd8e539edbc13070a4d455792
701
+ - 7K/5c8dafad7d782c76ffad8c14e9e1244ce2b83aa12324c54a3cc10176964acf04
702
+ - 7K/63798f5c6fbfcb4eb686268248b8ecbc8d87d920b2bcce967eeaedfd3b3b6d82
703
+ - 7K/946f49be73928469000baa5ca04d2573137c5ee6a66362bcf8d130354dca8924
704
+ - 7K/9e9a89ae6fed06d6e2f4749b4b0059f35ca97f848cedc4a14345999e746f7884
705
+ - 7K/cd9c981eeb4a9091547af19181b382698e9d9eee0a838c7c9783a8a268af6aee
706
+ - 7K/d4fbeba0168af8fddb2fc695881787aedcd62f477c7dcec9ebca7b8594bbd95b
707
+ scannet:
708
+ stanza: scannet
709
+ caption_windows: true
710
+ pointodyssey_train:
711
+ stanza: pointodyssey_train
712
+ caption_windows: true
713
+ # Its local captions cover 96 native frames. A 49-frame draw must use
714
+ # stride 1 to stay inside that exact caption interval; the inherited
715
+ # [2, 4, 6, 8] choices do not fit any 96-frame caption window.
716
+ caption_compatible_temporal_strides:
717
+ - 1
718
+ robot_sources:
719
+ - name: robotwin
720
+ weight: 10.0
721
+ frames: 49
722
+ stride: 4
723
+ # Simulator ground truth: depth, camera and the arm tracks are generated targets (owner decision 2026-11-05); DINO too.
724
+ condition_only_streams: []
725
+ reader:
726
+ source_split_manifest: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/flexpi_robotwin_robot_arm_track_video_20261023/split_manifest.json
727
+ source_split_sha256: 10467f314d57acf8deba64b94d0842b0dedcc256acf5f055df92203f7bf10f4e
728
+ pretrained_norm_stats: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/robotwin_action_image_clean_train_filtered_20261102/native_stats/dataset_stats.json
729
+ clean_train_split:
730
+ manifest: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/robotwin_action_image_clean_train_filtered_20261102/split_manifest.json
731
+ sha256: 6511ed1a88a1047d74fb80dd11b1304f6d77abce6cbe3f56df4f5cf7b9f9ee7c
732
+ train_episodes: 26288
733
+ held_out_episodes: 150
734
+ excluded_clamped_episodes: 1062
735
+ geometry:
736
+ simulator_ground_truth: true
737
+ log_factor: 0.4
738
+ raymap_camera: {origin_fraction: 0.3333333333333333, origin_range: 1.0, decode_margin: 16, origin_box_grid: 1}
739
+ gauge: window_labelled_median_depth
740
+ release_root: /datasets/engs2907/FlexPi/robotwin_3d
741
+ calibration_root: /datasets/engs2907/FlexPi/robotwin_3d/calibration
742
+ # Two real clean-train windows at 21 frames through reader, DINO and the native batch adapter, on pin f6992c4b.
743
+ readback_receipt:
744
+ path: /scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_nano_two_training_restart_20261107/attempt01/robot49_cpu/results/robotwin_readback.json
745
+ sha256: 785c7dde5f5f69a6a6fe30fe6accf56fe2380ab8903b7c9f2ec625d0501ecf2f
746
+ - name: libero
747
+ weight: 6.0
748
+ frames: 49
749
+ stride: 4
750
+ condition_only_streams: []
751
+ reader:
752
+ hydra_overrides: []
753
+ geometry:
754
+ simulator_ground_truth: true
755
+ readback_receipt:
756
+ path: /scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_nano_two_training_restart_20261107/attempt01/robot49_cpu/results/libero_readback.json
757
+ sha256: c8d49b7d87f64b45f1183ab8babc3c17d632e5345e0f697ed4a2aebc6fac4a0f
758
+ real_video_sources: []
759
+ molmo_motion_sources:
760
+ - name: molmo_motion
761
+ weight: 6.0
762
+ frames: 49
763
+ # The camera is an input only, like every real source's camera; there is no depth. DINO and the tracking video are targets.
764
+ condition_only_streams: [camera]
765
+ reader:
766
+ config: train/mot/wan_camera_omega_video_depth_vae_and_sparse_future_prediction_molmo_motion_21_frames_512
767
+ stanza: molmo_motion
768
+ release_root: /vgg/yushi/dataset/MolmoMotion-1M
769
+ subsets: [egodex, xperience, stereo4d, hepic]
770
+ # Retain the prior relative weights for this initial 49-frame launch; these are not a newly measured 49-frame census.
771
+ subset_weights: {egodex: 133.6, xperience: 432.4, stereo4d: 210.9, hepic: 63.1}
772
+ temporal_stride: {egodex: 2, xperience: 1, stereo4d: 4, hepic: 1}
773
+ processed_release_only_subsets: [egodex, hepic]
774
+ # EgoDex and HD-EPIC use their processed release videos; neither raw-video path is used by this recipe.
775
+ # The pinned config leaves Xperience's `object` records out: their 3D was lifted with the 256-pixel camera matrix.
776
+ excluded_objects: {xperience: [object]}
777
+ max_window_redraws: 64
778
+ exclusions:
779
+ # The registered 120-record test cohort (the VGG held-out spec molmo_motion_test_120_records_droid_egodex_molmospaces_
780
+ # 49_frames_512; tools/omni4d/write_molmo_motion_test_cohort_manifest.py). The release's own test splits (DROID's too) and
781
+ # PointMotionBench (hot3d, worldtrack, davis) are asserted disjoint from the train records in code.
782
+ test_cohort_manifest:
783
+ path: /scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_molmo_motion_20261105/cohort/molmo_motion_test_cohort_120_records.json
784
+ sha256: d1f8e8d8ad8f8f67c8bfd23319cdea026ea8cfcba1194b5f4594366e9f64fd76
785
+ point_motion_bench_root: /scratch/shared/engs2907/PointMotionBench
786
+ readback_receipt:
787
+ path: /scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_molmo_49frame_source_readback_20261107/attempt01/native_group_readback/molmo_motion_readback.json
788
+ sha256: ec0577a78e97db76f14cca0bd4d44ab87bc4e588f615fca242c1d675abb8dbb1
789
+ launch_blockers: {}
790
+ mixture_policy:
791
+ total_weight: 139.0
792
+ group_weights: {synthetic: 79.0, real_video: 22.0, robot: 26.0, tracks: 6.0, human_video: 6.0}
793
+ group_share_percent: {synthetic: 56.8, real_video: 15.8, robot: 18.7, tracks: 4.3, human_video: 4.3}
794
+ source_groups:
795
+ four_d_pretraining_relaunch_syn4d_sources: synthetic
796
+ four_d_pretraining_relaunch_omniworld_game_source: synthetic
797
+ four_d_pretraining_relaunch_synthverse_source: synthetic
798
+ four_d_pretraining_relaunch_vkitti_source: synthetic
799
+ four_d_pretraining_relaunch_tartanair_sky_then_far_percentile_source: synthetic
800
+ four_d_pretraining_relaunch_dynamic_replica_stored_depth_source: synthetic
801
+ four_d_pretraining_relaunch_spring_stored_depth_source: synthetic
802
+ four_d_pretraining_relaunch_midair_within_200_meters_source: synthetic
803
+ four_d_pretraining_relaunch_mvs_synth_1000_then_far_percentile_source: synthetic
804
+ four_d_pretraining_relaunch_tartanground_sky_then_far_percentile_source: synthetic
805
+ four_d_pretraining_relaunch_paralleldomain4d_sky_then_far_percentile_source: synthetic
806
+ scannet: real_video
807
+ dl3dv: real_video
808
+ wildrgbd: real_video
809
+ waymo: real_video
810
+ robotwin: robot
811
+ droid: robot
812
+ libero: robot
813
+ pointodyssey_train: tracks
814
+ kubric_train: tracks
815
+ molmo_motion: human_video
cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/source_windows_recipe.yaml ADDED
@@ -0,0 +1,815 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # B4/GAS4 49-frame continuation from the update-14,700 weights-only overlay; no Cityscapes.
2
+ # This is a separate model-only run from an explicit five-stream overlay: 21 sources, total weight 139.
3
+ # The Cityscapes source is deliberately omitted; the remaining real-video sources keep their prior weights.
4
+ # Next Cosmos stage-one version: the owner's weighted mixture at 49 frames (decision 2026-11-07).
5
+ #
6
+ # Weights, not shares. Total 139 over 21 logical sources:
7
+ # synthetic 79 (56.8%) the eleven synthetic sources at the 20,000-update Wan run's weights, unchanged:
8
+ # syn4d 15.2, omniworld_game 8.75, synthverse 11.05, vkitti 6, tartanair 6, dynamic_replica 5,
9
+ # spring 8, midair 3, mvs_synth 3, tartanground 3, paralleldomain4d 10
10
+ # real video 22 (15.8%) ScanNet 8, DL3DV 6, WildRGB-D 4, Waymo 4; Cityscapes excluded.
11
+ # robot 26 (18.7%) RoboTwin 10 (clean-train episodes only), DROID 10 (depth and camera as inputs only),
12
+ # LIBERO 6 (official training demonstrations only); no action stream in stage one
13
+ # tracks 6 ( 4.3%) PointOdyssey 3, Kubric 3
14
+ # human video 6 ( 4.3%) MolmoMotion including HD-EPIC; DINO and sparse tracking video are targets, camera is input-only, depth absent
15
+ # DINO is a generated target on the real-video sources (ScanNet, DL3DV, WildRGB-D, DROID, Waymo): it is computed from the
16
+ # RGB, so it is exact on real data, and the future-feature forecasts the DINO-Foresight and VFMF comparisons make would
17
+ # otherwise never train on real video. This changes the first launch's roles: the owner's rule is that geometry is
18
+ # condition-only on real data, and that rule still holds for depth and camera, which stay inputs only here.
19
+ # Waymo trains on the processed training split only (waymo.training.pickle.gz, SHA-256 6abd3476...: 798 sequences,
20
+ # 158,081 frames; validation never enters training) through the pinned sparse-track reader's waymo stanza: sparse LiDAR
21
+ # depth, the camera and the RGB; no tracking video; strides 1-4 at 21 frames.
22
+ # No segmentation stream, no Cityscapes, no YAM.
23
+ #
24
+ # The VGG window block below is read by the pinned VGG checkout (vgg_commit) through VggStageOneWindowDataset: the
25
+ # synthetic sources are the recorded 500K run's stanzas (omni4d_vgg_sparse_tracking_synthetic_continuation_sources.yaml,
26
+ # SHA-256 ae49ec6d...) with the window length 49 and missing_caption: error. RoboTwin and LIBERO are read through
27
+ # robot_sources (RobotEpisodeStageOneDataset) because the pinned VGG checkout has no Track4D reader for them.
28
+ # The tracking-video query draw keeps the recorded 10-pixel grid (no query_count_range).
29
+ #
30
+ # Train membership of the real-video sources (official split files, 2026-11-05):
31
+ # DL3DV the pinned reader ignores `split` and indexes every processed scene, 85 of which are official
32
+ # DL3DV-Benchmark (test) scenes. The 141 official benchmark scene hashes (Hugging Face
33
+ # DL3DV/DL3DV-Benchmark, dataset revision 9684e838) are in omni4d_dl3dv_official_benchmark_scene_hashes.txt;
34
+ # the 87 of them present in the processed 1K-7K root are added to the stanza's exclude_sequence_ids below
35
+ # (source_reader_overrides), leaving about 5,898 scenes.
36
+ # WildRGB-D the pinned reader reads <category>/selected_seqs_train.json; those 2,244 processed train scenes are all in
37
+ # the official train-of-both lists (nvs_list.json and camera_eval_list.json of hongchi/wildrgbd, CUT3R's
38
+ # get_set_list rule) and overlap the official NVS and camera-evaluation validation scenes in none.
39
+ #
40
+ # launch_blockers names every evidence gap that stops the launch. A blocked source keeps its weight; composition
41
+ # (omni4d_native_config.make_config) and the window dataset both refuse to start while the list is not empty. A robot
42
+ # source additionally needs a passing real-window readback receipt (readback_receipt). WildRGB-D and Kubric train on
43
+ # caption text converted by tools/omni4d/convert_stage_one_caption_texts.py (text-only products: no umT5 cache exists).
44
+ #
45
+ # VGG pin: the 500K run's f5a59cd51 plus the window-reader updates, processed-release EgoDex support, and the
46
+ # native2048 Cityscapes caption-reader migration at f6992c4b. The approved raw EgoDex window and Cityscapes caption
47
+ # product readbacks used this exact pin. City captions point to the completed 2026-11-06 converted text product below.
48
+ # One makes the DROID and LIBERO robot window length the data node's num_frames: the pinned window class and the LIBERO
49
+ # painter fixed the clock at 33 rows in five places, one a literal inside the painter's constructor; a project adapter
50
+ # could only rebind module globals and fake attributes, so the smallest safe change is in the VGG source, with the
51
+ # recorded 33-row nodes reading as before. Two DROID commits (cherry-picked from the VGG main stack, f5a59cd is their
52
+ # ancestor) keep a window whose FK arm has no visible frame-0 query, with its source motion absent, instead of raising.
53
+ # The fourth adds the stanza key position_windows (src/vggforecast/data/position_windows.py): each listed position range is
54
+ # one window and the stride and start are drawn inside it. ScanNet and PointOdyssey use it (source_reader_overrides,
55
+ # caption_windows: true) with the local caption windows of their products, so every window they draw lies inside one
56
+ # caption window and the strict caption join never redraws. The fifth adds a waymo stanza to the stage-one real-source config
57
+ # (no source uses it yet: every real source leaves it out). The pin is a separate checkout (a git worktree of the earlier pin
58
+ # directory, which a running training still reads at 9cb6a626a), so the earlier pin's files never change under a run.
59
+ vgg_repository: /scratch/shared/engs2907/vgg_pins/vgg_forecast_4d_cosmos_source_20261106_f6992c4
60
+ vgg_commit: f6992c4b67441dc9778642a252447a2687856116
61
+ recipe_source_config: /scratch/shared/engs2907/logs/vgg_forecast_4d/train/flexpi_four_d_pretraining_wan22_5b_separate_streams_full_parameters_video_depth_camera_dino_and_sparse_tracking_video_from_the_15000_update_four_d_run_20261101/attempt02/run/config.yaml
62
+ recipe_source_config_sha256: 72590f4598cae1b329c31b3cb45d059bdaeca23642faa10ca3f31372f9ebc0aa
63
+ window_block:
64
+ _target_: four_d_windows.FourDPretrainingWindows
65
+ sources:
66
+ - name: four_d_pretraining_relaunch_syn4d_sources
67
+ kind: track4d_source_config
68
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_rendered_depth_sources_21_frames_512
69
+ weight: 15.2
70
+ left_out_sources:
71
+ - omniworld_game
72
+ - tartanair
73
+ - mvs_synth
74
+ condition_only_streams: []
75
+ - name: four_d_pretraining_relaunch_omniworld_game_source
76
+ kind: track4d_source_config
77
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_rendered_depth_sources_21_frames_512
78
+ weight: 8.75
79
+ left_out_sources:
80
+ - syn4d_v1
81
+ - syn4d_v2_mixed
82
+ - syn4d_v2_mixed_no_sim
83
+ - tartanair
84
+ - mvs_synth
85
+ condition_only_streams: []
86
+ - name: four_d_pretraining_relaunch_synthverse_source
87
+ kind: track4d_source_config
88
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_synthverse_dynamic_replica_vkitti_midair_spring_sources_without_midair_quality_masks_21_frames_512
89
+ weight: 11.05
90
+ left_out_sources:
91
+ - dynamic_replica
92
+ - vkitti
93
+ - midair
94
+ - spring
95
+ condition_only_streams: []
96
+ - name: four_d_pretraining_relaunch_vkitti_source
97
+ kind: track4d_source_config
98
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_synthverse_dynamic_replica_vkitti_midair_spring_sources_without_midair_quality_masks_21_frames_512
99
+ weight: 6.0
100
+ left_out_sources:
101
+ - synthverse_depth
102
+ - dynamic_replica
103
+ - midair
104
+ - spring
105
+ condition_only_streams: []
106
+ - name: four_d_pretraining_relaunch_tartanair_sky_then_far_percentile_source
107
+ kind: track4d_source_config
108
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_sky_and_far_percentile_depth_validity_rules_21_frames_512
109
+ weight: 6.0
110
+ left_out_sources:
111
+ - dynamic_replica
112
+ - spring
113
+ - midair
114
+ - mvs_synth
115
+ - tartanground
116
+ - paralleldomain4d
117
+ condition_only_streams: []
118
+ - name: four_d_pretraining_relaunch_dynamic_replica_stored_depth_source
119
+ kind: track4d_source_config
120
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_sky_and_far_percentile_depth_validity_rules_21_frames_512
121
+ weight: 5.0
122
+ left_out_sources:
123
+ - tartanair
124
+ - spring
125
+ - midair
126
+ - mvs_synth
127
+ - tartanground
128
+ - paralleldomain4d
129
+ condition_only_streams: []
130
+ - name: four_d_pretraining_relaunch_spring_stored_depth_source
131
+ kind: track4d_source_config
132
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_sky_and_far_percentile_depth_validity_rules_21_frames_512
133
+ weight: 8.0
134
+ left_out_sources:
135
+ - tartanair
136
+ - dynamic_replica
137
+ - midair
138
+ - mvs_synth
139
+ - tartanground
140
+ - paralleldomain4d
141
+ condition_only_streams: []
142
+ - name: four_d_pretraining_relaunch_midair_within_200_meters_source
143
+ kind: track4d_source_config
144
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_sky_and_far_percentile_depth_validity_rules_21_frames_512
145
+ weight: 3.0
146
+ left_out_sources:
147
+ - tartanair
148
+ - dynamic_replica
149
+ - spring
150
+ - mvs_synth
151
+ - tartanground
152
+ - paralleldomain4d
153
+ condition_only_streams: []
154
+ - name: four_d_pretraining_relaunch_mvs_synth_1000_then_far_percentile_source
155
+ kind: track4d_source_config
156
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_sky_and_far_percentile_depth_validity_rules_21_frames_512
157
+ weight: 3.0
158
+ left_out_sources:
159
+ - tartanair
160
+ - dynamic_replica
161
+ - spring
162
+ - midair
163
+ - tartanground
164
+ - paralleldomain4d
165
+ condition_only_streams: []
166
+ - name: four_d_pretraining_relaunch_tartanground_sky_then_far_percentile_source
167
+ kind: track4d_source_config
168
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_sky_and_far_percentile_depth_validity_rules_21_frames_512
169
+ weight: 3.0
170
+ left_out_sources:
171
+ - tartanair
172
+ - dynamic_replica
173
+ - spring
174
+ - midair
175
+ - mvs_synth
176
+ - paralleldomain4d
177
+ condition_only_streams: []
178
+ - name: four_d_pretraining_relaunch_paralleldomain4d_sky_then_far_percentile_source
179
+ kind: track4d_source_config
180
+ config: train/mot/wan_video_depth_vae_from_video_four_d_pretraining_relaunch_sky_and_far_percentile_depth_validity_rules_21_frames_512
181
+ weight: 10.0
182
+ left_out_sources:
183
+ - tartanair
184
+ - dynamic_replica
185
+ - spring
186
+ - midair
187
+ - mvs_synth
188
+ - tartanground
189
+ condition_only_streams: []
190
+ - name: scannet
191
+ kind: track4d_source_config
192
+ config: train/mot/stage_one_real_and_native_track_source_windows_9_frames_288_512
193
+ weight: 8.0
194
+ left_out_sources:
195
+ - dl3dv
196
+ - wildrgbd
197
+ - pointodyssey_train
198
+ - kubric_train
199
+ - droid
200
+ - waymo
201
+ condition_only_streams:
202
+ - pointmap
203
+ - camera
204
+ - name: dl3dv
205
+ kind: track4d_source_config
206
+ config: train/mot/stage_one_real_and_native_track_source_windows_9_frames_288_512
207
+ weight: 6.0
208
+ left_out_sources:
209
+ - scannet
210
+ - wildrgbd
211
+ - pointodyssey_train
212
+ - kubric_train
213
+ - droid
214
+ - waymo
215
+ condition_only_streams:
216
+ - pointmap
217
+ - camera
218
+ - name: wildrgbd
219
+ kind: track4d_source_config
220
+ config: train/mot/stage_one_real_and_native_track_source_windows_9_frames_288_512
221
+ weight: 4.0
222
+ left_out_sources:
223
+ - dl3dv
224
+ - scannet
225
+ - pointodyssey_train
226
+ - kubric_train
227
+ - droid
228
+ - waymo
229
+ condition_only_streams:
230
+ - pointmap
231
+ - camera
232
+ - name: waymo
233
+ kind: track4d_source_config
234
+ config: train/mot/stage_one_real_and_native_track_source_windows_9_frames_288_512
235
+ weight: 4.0
236
+ left_out_sources:
237
+ - dl3dv
238
+ - scannet
239
+ - wildrgbd
240
+ - pointodyssey_train
241
+ - kubric_train
242
+ - droid
243
+ condition_only_streams:
244
+ - pointmap
245
+ - camera
246
+ - name: droid
247
+ kind: track4d_source_config
248
+ config: train/mot/stage_one_real_and_native_track_source_windows_9_frames_288_512
249
+ weight: 10.0
250
+ left_out_sources:
251
+ - dl3dv
252
+ - scannet
253
+ - wildrgbd
254
+ - pointodyssey_train
255
+ - kubric_train
256
+ - waymo
257
+ condition_only_streams:
258
+ - pointmap
259
+ - camera
260
+ - name: pointodyssey_train
261
+ kind: track4d_source_config
262
+ config: train/mot/stage_one_real_and_native_track_source_windows_9_frames_288_512
263
+ weight: 3.0
264
+ left_out_sources:
265
+ - dl3dv
266
+ - scannet
267
+ - wildrgbd
268
+ - kubric_train
269
+ - droid
270
+ - waymo
271
+ condition_only_streams: []
272
+ - name: kubric_train
273
+ kind: track4d_source_config
274
+ config: train/mot/stage_one_real_and_native_track_source_windows_9_frames_288_512
275
+ weight: 3.0
276
+ left_out_sources:
277
+ - dl3dv
278
+ - scannet
279
+ - wildrgbd
280
+ - pointodyssey_train
281
+ - droid
282
+ - waymo
283
+ condition_only_streams:
284
+ - pointmap
285
+ frames: 49
286
+ raster_hw:
287
+ - 288
288
+ - 512
289
+ raymap_camera:
290
+ origin_fraction: 0.3333333333333333
291
+ origin_range: 1.0
292
+ decode_margin: 16
293
+ origin_box_grid: 1
294
+ pointmap_log_factor: 0.4
295
+ pointmap_hole_fill:
296
+ synthverse_depth: far_clip
297
+ syn4d: far_clip
298
+ omniworld_game: far_clip
299
+ dynamic_replica: nearest_labelled_depth_of_3x3_neighbours
300
+ tartanair: far_clip
301
+ dl3dv: far_clip
302
+ scannet: telea_log_depth
303
+ virtual_kitti2: far_clip
304
+ midair: far_clip
305
+ pointodyssey: far_clip
306
+ mvs_synth: far_clip
307
+ wildrgbd: telea_log_depth
308
+ hypersim: far_clip
309
+ spring: far_clip
310
+ kubric49: far_clip
311
+ kubric81: far_clip
312
+ droid_stereo_depth: telea_log_depth
313
+ tartanground: far_clip
314
+ paralleldomain4d: far_clip
315
+ syn4d_v2: far_clip
316
+ waymo: far_clip
317
+ pointmap_gauge: window_labelled_median_depth
318
+ geometry_image: genception_log_depth
319
+ context_len: 512
320
+ batch_size: 4
321
+ stream_seed: 42
322
+ epoch_windows: 32000000
323
+ val_set_proportion: 0.0
324
+ resume_state: null
325
+ gradient_accumulation_steps: 4
326
+ loader_workers: 12
327
+ loader_opencv_threads: 1
328
+ window_captions:
329
+ caption_dropout: 0.1
330
+ missing_caption: error
331
+ caches:
332
+ syn4d_v1_gemini_holo4d:
333
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/syn4d_v1_gemini_holo4d_local_and_global_with_repaired_globals_20261031/captions_t5_emb.safetensors
334
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/syn4d_v1_gemini_holo4d_local_and_global_with_repaired_globals_20261031/captions_t5_emb_meta.json
335
+ syn4d_v2_gemini_holo4d:
336
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/syn4d_v2_gemini_holo4d_20261030/captions_t5_emb.safetensors
337
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/syn4d_v2_gemini_holo4d_20261030/captions_t5_emb_meta.json
338
+ tartanair_gemini_holo4d:
339
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/tartanair_gemini_holo4d/captions_t5_emb.safetensors
340
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/tartanair_gemini_holo4d/captions_t5_emb_meta.json
341
+ mvs_synth_gemini_holo4d:
342
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/mvs_synth_gemini_holo4d/captions_t5_emb.safetensors
343
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/mvs_synth_gemini_holo4d/captions_t5_emb_meta.json
344
+ omniworld_game_gemini_local_official_text_chunks_and_gemini_global:
345
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/omniworld_game_gemini_local_official_text_chunks_and_gemini_global_20261030b/captions_t5_emb.safetensors
346
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/omniworld_game_gemini_local_official_text_chunks_and_gemini_global_20261030b/captions_t5_emb_meta.json
347
+ synthverse_gemini_holo4d:
348
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/synthverse_gemini_holo4d_local96_and_global_20261031/captions_t5_emb.safetensors
349
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/synthverse_gemini_holo4d_local96_and_global_20261031/captions_t5_emb_meta.json
350
+ dynamic_replica_gemini_holo4d:
351
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/dynamic_replica_gemini_holo4d/captions_t5_emb.safetensors
352
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/dynamic_replica_gemini_holo4d/captions_t5_emb_meta.json
353
+ vkitti_gemini_holo4d:
354
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/vkitti_gemini_holo4d/captions_t5_emb.safetensors
355
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/vkitti_gemini_holo4d/captions_t5_emb_meta.json
356
+ midair_gemini_holo4d:
357
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/midair_gemini_holo4d_20261030/captions_t5_emb.safetensors
358
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/midair_gemini_holo4d_20261030/captions_t5_emb_meta.json
359
+ spring_gemini_holo4d:
360
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/spring_gemini_holo4d/captions_t5_emb.safetensors
361
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/spring_gemini_holo4d/captions_t5_emb_meta.json
362
+ tartanground_gemini_holo4d:
363
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/tartanground_gemini_holo4d_20261030/captions_t5_emb.safetensors
364
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/tartanground_gemini_holo4d_20261030/captions_t5_emb_meta.json
365
+ paralleldomain4d_gemini_holo4d:
366
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/paralleldomain4d_gemini_holo4d_20261030/captions_t5_emb.safetensors
367
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/paralleldomain4d_gemini_holo4d_20261030/captions_t5_emb_meta.json
368
+ scannet_gemini_holo4d:
369
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/.scannet_gemini_holo4d.version-tovvjwai/captions_t5_emb.safetensors
370
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/.scannet_gemini_holo4d.version-tovvjwai/captions_t5_emb_meta.json
371
+ pointodyssey_gemini_holo4d:
372
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/.pointodyssey_gemini_holo4d.version-6vfyvsbi/captions_t5_emb.safetensors
373
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/.pointodyssey_gemini_holo4d.version-6vfyvsbi/captions_t5_emb_meta.json
374
+ droid_vp_gemini_holo4d:
375
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/.droid_vp_gemini_holo4d.version-e51r5urf/captions_t5_emb.safetensors
376
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/.droid_vp_gemini_holo4d.version-e51r5urf/captions_t5_emb_meta.json
377
+ dl3dv_gemini_holo4d:
378
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/dl3dv_gemini_holo4d_20261031/captions_t5_emb.safetensors
379
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/wan_umt5_captions/four_d_window_captions_512/dl3dv_gemini_holo4d_20261031/captions_t5_emb_meta.json
380
+ wildrgbd_text_captions:
381
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/cosmos_caption_texts_20261105/wildrgbd_text_captions/captions_t5_emb.safetensors
382
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/cosmos_caption_texts_20261105/wildrgbd_text_captions/captions_t5_emb_meta.json
383
+ kubric49_text_captions:
384
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/cosmos_caption_texts_20261105/kubric49_text_captions/captions_t5_emb.safetensors
385
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/cosmos_caption_texts_20261105/kubric49_text_captions/captions_t5_emb_meta.json
386
+ waymo_gemini_holo4d:
387
+ embeddings: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/cosmos_caption_texts_20261105/waymo_gemini_holo4d/captions_t5_emb.safetensors
388
+ meta: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/cosmos_caption_texts_20261105/waymo_gemini_holo4d/captions_t5_emb_meta.json
389
+ stanzas:
390
+ syn4d_v1:
391
+ product: syn4d_v1_gemini_holo4d
392
+ join: local_covering_then_global
393
+ syn4d_v2_mixed:
394
+ product: syn4d_v2_gemini_holo4d
395
+ join: local_covering_then_global
396
+ syn4d_v2_mixed_no_sim:
397
+ product: syn4d_v2_gemini_holo4d
398
+ join: local_covering_then_global
399
+ tartanair:
400
+ product: tartanair_gemini_holo4d
401
+ join: local_covering
402
+ mvs_synth:
403
+ product: mvs_synth_gemini_holo4d
404
+ join: local_covering
405
+ omniworld_game:
406
+ product: omniworld_game_gemini_local_official_text_chunks_and_gemini_global
407
+ join: local_covering_then_second_local_covering_then_global
408
+ pointodyssey_test: none
409
+ kubric49_validation: none
410
+ wildrgbd_test: none
411
+ vkitti_test:
412
+ product: vkitti_gemini_holo4d
413
+ join: local_covering_then_global
414
+ droid_held_out: none
415
+ synthverse_depth:
416
+ product: synthverse_gemini_holo4d
417
+ join: local_covering_then_global
418
+ dynamic_replica:
419
+ product: dynamic_replica_gemini_holo4d
420
+ join: global
421
+ vkitti:
422
+ product: vkitti_gemini_holo4d
423
+ join: local_covering_then_global
424
+ midair:
425
+ product: midair_gemini_holo4d
426
+ join: local_covering_then_global
427
+ spring:
428
+ product: spring_gemini_holo4d
429
+ join: global
430
+ tartanground:
431
+ product: tartanground_gemini_holo4d
432
+ join: local_covering_then_global
433
+ paralleldomain4d:
434
+ product: paralleldomain4d_gemini_holo4d
435
+ join: global
436
+ paralleldomain4d_validation:
437
+ product: paralleldomain4d_gemini_holo4d
438
+ join: global
439
+ scannet:
440
+ product: scannet_gemini_holo4d
441
+ join: local_containing_then_global
442
+ pointodyssey_train:
443
+ product: pointodyssey_gemini_holo4d
444
+ join: local_containing_then_global
445
+ droid:
446
+ product: droid_vp_gemini_holo4d
447
+ join: global
448
+ dl3dv:
449
+ product: dl3dv_gemini_holo4d
450
+ join: local_containing_then_global
451
+ wildrgbd:
452
+ product: wildrgbd_text_captions
453
+ join: local_containing_then_global
454
+ kubric_train:
455
+ product: kubric49_text_captions
456
+ join: global_required
457
+ waymo:
458
+ product: waymo_gemini_holo4d
459
+ join: local_containing_then_global
460
+ training_scene_exclusions:
461
+ reason: rendered depth passes through glass (raw EXR depth through glass panes at 190-260 m), owner 10-29; evidence /scratch/shared/engs2907/logs/vgg_forecast_4d/tools/syn4d_glass_depth_check_20261029;
462
+ scifiroom_bald added 10-30 from /scratch/shared/engs2907/logs/vgg_forecast_4d/tools/depth_clipping_audit_20261030
463
+ stanzas:
464
+ syn4d_v1:
465
+ - bigoffice_v1
466
+ - lab_bald
467
+ - office_bald
468
+ - scifiroom_bald
469
+ syn4d_v2_mixed:
470
+ - citynight
471
+ - mall
472
+ - metro
473
+ syn4d_v2_mixed_no_sim:
474
+ - citynight
475
+ - mall
476
+ - metro
477
+ tracking_video:
478
+ query_grid_px: 10
479
+ disc_radius_px: 2
480
+ occluded_style: ring
481
+ ring_radius_px: 3
482
+ ring_width_px: 1
483
+ xyz_normalization: quantile_0.02_0.98
484
+ colour_floor: 0.2
485
+ min_frame_zero_points: 16
486
+ surface_test:
487
+ relative_tolerance: 0.1
488
+ absolute_tolerance: 0.05
489
+ stanza_rules:
490
+ tartanair: static_scene_points
491
+ midair: static_scene_points
492
+ mvs_synth: static_scene_points
493
+ tartanground: static_scene_points
494
+ syn4d_v1: tracked_points
495
+ syn4d_v2_mixed: tracked_points
496
+ syn4d_v2_mixed_no_sim: tracked_points
497
+ omniworld_game: no_tracking_video
498
+ synthverse_depth: no_tracking_video
499
+ dynamic_replica: tracked_points
500
+ vkitti: no_tracking_video
501
+ spring: no_tracking_video
502
+ paralleldomain4d: no_tracking_video
503
+ scannet: static_scene_points
504
+ wildrgbd_test: static_scene_points
505
+ pointodyssey_test: tracked_points
506
+ kubric49_validation: no_tracking_video
507
+ vkitti_test: no_tracking_video
508
+ droid_held_out: no_tracking_video
509
+ paralleldomain4d_validation: no_tracking_video
510
+ dynamic_replica_test: tracked_points
511
+ pointodyssey_train: tracked_points
512
+ kubric_train: tracked_points
513
+ droid: tracked_points
514
+ dl3dv: static_scene_points
515
+ wildrgbd: static_scene_points
516
+ waymo: no_tracking_video
517
+ stanza_track_reading:
518
+ syn4d_v1:
519
+ stanza:
520
+ capability: dynamic_cross_time
521
+ supervision:
522
+ - pointmap
523
+ - source_motion
524
+ annotation_provenance: syn4d_mesh_surface_forward_track_from_frame0
525
+ allow_missing_tracks: false
526
+ kwargs:
527
+ return_tracks: true
528
+ depth_only: false
529
+ metadata_root: /vgg/chenzhao/syn4d_metadata
530
+ wrapped_face_distance_thresh: 0.1
531
+ query_inconsistent_distance_thresh: 0.1
532
+ pointodyssey_test:
533
+ stanza:
534
+ supervision:
535
+ - pointmap
536
+ - source_motion
537
+ kwargs:
538
+ return_tracks: true
539
+ dynamic_replica_test:
540
+ stanza:
541
+ supervision:
542
+ - pointmap
543
+ - source_motion
544
+ kwargs:
545
+ return_tracks: true
546
+ syn4d_v2_mixed:
547
+ loader: syn4d_v2
548
+ stanza:
549
+ capability: dynamic_cross_time
550
+ supervision:
551
+ - pointmap
552
+ - source_motion
553
+ annotation_provenance: syn4d_v2_mesh_surface_forward_track_from_frame0
554
+ allow_missing_tracks: false
555
+ kwargs:
556
+ return_tracks: true
557
+ depth_only: false
558
+ metadata_root: /vgg/chenzhao/syn4d_metadata
559
+ tracking_format: safetensor
560
+ mask_bg: false
561
+ trusted_background_depth_max: 80.0
562
+ wrapped_face_distance_thresh: 0.1
563
+ query_inconsistent_distance_thresh: 0.1
564
+ syn4d_v2_mixed_no_sim:
565
+ loader: syn4d_v2
566
+ stanza:
567
+ capability: dynamic_cross_time
568
+ supervision:
569
+ - pointmap
570
+ - source_motion
571
+ annotation_provenance: syn4d_v2_mesh_surface_forward_track_from_frame0
572
+ allow_missing_tracks: false
573
+ kwargs:
574
+ return_tracks: true
575
+ depth_only: false
576
+ metadata_root: /vgg/chenzhao/syn4d_metadata
577
+ tracking_format: safetensor
578
+ mask_bg: false
579
+ trusted_background_depth_max: 80.0
580
+ wrapped_face_distance_thresh: 0.1
581
+ query_inconsistent_distance_thresh: 0.1
582
+ dynamic_replica:
583
+ stanza:
584
+ supervision:
585
+ - pointmap
586
+ - source_motion
587
+ kwargs:
588
+ return_tracks: true
589
+ pointodyssey_train:
590
+ stanza:
591
+ allow_missing_tracks: false
592
+ capability: dynamic_cross_time
593
+ supervision:
594
+ - pointmap
595
+ - source_motion
596
+ kwargs:
597
+ return_tracks: true
598
+ kubric_train:
599
+ stanza:
600
+ allow_missing_tracks: false
601
+ capability: dynamic_cross_time
602
+ supervision:
603
+ - pointmap
604
+ - source_motion
605
+ kwargs:
606
+ return_tracks: true
607
+ droid:
608
+ stanza:
609
+ allow_missing_tracks: false
610
+ capability: dynamic_cross_time
611
+ supervision:
612
+ - pointmap
613
+ - source_motion
614
+ kwargs:
615
+ return_tracks: true
616
+ source_reader_overrides:
617
+ dl3dv:
618
+ stanza: dl3dv
619
+ exclude_sequence_ids:
620
+ - 1K/0a1b7c20a92c43c6b8954b1ac909fb2f0fa8b2997b80604bc8bbec80a1cb2da3
621
+ - 1K/14eb48a50e37df548894ab6d8cd628a21dae14bbe6c462e894616fc5962e6c49
622
+ - 1K/1da888bdedfc9629c0fa9f82cf3f5d96f8103baee0ff64d9311aea1224a9f2ae
623
+ - 1K/2991a75d1f0fb2b7a891d424fc2459bb954b61599b52e74a056a7acfb2998b77
624
+ - 1K/2cbfe28643b6636f9c70813cae7625aa858a352109493ac70fb429ce94dd55b3
625
+ - 1K/5a27c00f525f5484b7aa0d0107051253bf7bb395bde0b3a4486ff8c52fb4965e
626
+ - 1K/6d221625614d0515fa44d59a8797a49b64f86bc775e66be22f07a66aa0a0cb3a
627
+ - 1K/917e9c8985d353b0ee4c281f11fa84eb8550814562670cbb82cd0ec9c1194fe7
628
+ - 1K/918c8dad730c3b804306c5da8486124be4aa0612e85fb825338fd350c912e1b0
629
+ - 1K/a17a984ca90a9b5840fdf85b15104b0d18e25975981c1aa90fcdfd6eeeb285f3
630
+ - 1K/a726c1112ad026efc92f59f196c0698e67abd12a04135a7ff6349b1d31b089eb
631
+ - 1K/adb95f29c15b8a89e40c17d10596679be09d7719c6f08a453facd506d056ea8b
632
+ - 1K/d1b3a0b37acc72207295b05e8c359f8b43bf879eb882c60747cfc6d3c7d8efa3
633
+ - 1K/eb4cf52988f805e6fce11d1b239fa9de32eb157364cff06ebac0aa50e0a46567
634
+ - 2K/032dee9fb0a8bc1b90871dc5fe950080d0bcd3caf166447f44e60ca50ac04ec7
635
+ - 2K/0853979305f7ecb80bd8fc2c8df916410d471ef04ed5f1a64e9651baa41d7695
636
+ - 2K/093ef327b4e4f9d4ee52c02a354a53558a8652157fb0d58f3b4a708734afb334
637
+ - 2K/0f8ac521439691fe429e9efbed8d5ded1cee35bcf52d47731fb60d5a2ff661a7
638
+ - 2K/1d6a9ed47cce39fd1c4d18f776bcc97e507b81bde921ca596bd91b0b02b5e414
639
+ - 2K/26fd23358fa11fff0fb3180ef0b65591b486e20dcf753ce4a7aae49a37e370c7
640
+ - 2K/280abf7bd93b81b077af1db638229dbb09869052fa0b7b57c81c94d2db893829
641
+ - 2K/2beaca318994c25409dcbb6d0bdd96c3620f2f18aec44ea3f20edd302f18ca78
642
+ - 2K/341b4ff3dfd3d377d7167bd81f443bedafbff003bf04881b99760fc0aeb69510
643
+ - 2K/374ffd0c5ff8e544f588f1da3ce0d5b69dc39ff3af95d7ccec480fe5e533c45e
644
+ - 2K/41036716da7efda334c1d434c4141d15642e0e02f881a01b6c8c36f8bea64c45
645
+ - 2K/484c0aca40c1087f196d2caf3c8cb463864c08e60b98326ca02eb6dae8ce4984
646
+ - 2K/4ff8650b5c1140e98d957fbd1abbe2cfbbb416b37cb9f06f7f376956703801bb
647
+ - 2K/513e4ea2e8477b06f2b32417d5c243aec71d491cb0596a60e9fe304c635f20a1
648
+ - 2K/8b9fb9d9f10e8c64d5034be69809465753b8ba88ef12da82afd47d38ee789934
649
+ - 2K/91afb9910b042f7185c2b8e4b6b24b5785dae4542617b3f8005b5492f6d123f7
650
+ - 2K/ba55c875d20c34ee85ffc72264c4d77710852e5fb7d9ce4b9c26a8442850e98f
651
+ - 2K/c076929db6501cf7ebe386c70e6d77ea3af844a745e794f2ec17c981c465a69b
652
+ - 2K/ceb252f5d4419510655cf9ed7afbf3e8e688825f798d80414c7715dc8ace153a
653
+ - 2K/df29c225863d0173e060b1f3fde4e6eb5a57a4e4edabd397026ba1a028aeb39e
654
+ - 2K/ff592398657b3dfe94153332861985194a3e3c9d199c4a3a27a0ce4038e81ade
655
+ - 3K/35872363e17af5d173b6a0b09fcf5de94627ad5dc5f8a9ad4c579f3e70b4797a
656
+ - 3K/387eeb925ba47af5db9ce6102aaf6b07e6d3573b7e7568dda4cef22d5ab1342f
657
+ - 3K/3bb894d1933f3081134ad2d40e54de5f0636bd8b502b0a8561873bb63b0dce85
658
+ - 3K/5c3af581028068a3c402c7cbe16ecf9471ddf2897c34ab634b7b1b6cf81aba00
659
+ - 3K/669c36225b273e22ce08364f166dce910b708d7abd89e58463a15d2790b534d5
660
+ - 3K/66fd66cbed40a2d11ecd17ae6d6d60cd60d476616864ecdd351fe69e644f30a1
661
+ - 3K/70eac6ff18a1daae4eeccc5eb18723eaf7e029a77e79c79d9573f7bac59ba92b
662
+ - 3K/75fbbe467353cfdeb93bc1872144855576797be83a0c71c51155ee6b355d6d5e
663
+ - 3K/800cf88687a79c232700454c2f5a402e771bc7157037a17b0ce93da365b25575
664
+ - 3K/8cb2e97d26a639f05a571476240a8fa86988e6853f0f13cc05830d1578002aad
665
+ - 3K/9cbc5548643ca9cfe797f871bac58bccd02f0510c49b7cb3697b38dfd780a508
666
+ - 3K/a62f9a1c6362e7c8a3440261ea0921a0a4b60cf5822bd19579836694f26b3abd
667
+ - 3K/adf35184a12d4cfa3f4248b87aa5adb4f39f179df460d6d76136e13d37299a2a
668
+ - 3K/af0d7039e64af3959210c235e59425896ac064f4e0f721c437bbd2cf1d8ea910
669
+ - 3K/b3bf9079b442c681749f6bd95cfb0659c6d844f2228b2f4fa4b602a0e3627b9a
670
+ - 3K/b6d1134cb03c855d246f916beaa2d3d01290cf47937a5176011ff39b850ec615
671
+ - 3K/d8de66037bd03dd0d39d54f9978bac3318d912e30e22c21e1ada82a98ed48c53
672
+ - 3K/ddfdcfdf02d53dd89a055cd9e1ed30272ab3f57c7560ac2c3761bb01d76a7882
673
+ - 3K/e8ce51b6abfe05bf8dca47e29c8be6c1e6de27a8c9fece7a121400b931b2ca0f
674
+ - 3K/e9360e7a89bee835dc847cf8796093e634b759ff582558788dcfe8326f6e8901
675
+ - 3K/f7aaea9ac683bb1945cfeb1daa9abca3375249e0d19281a89a3f575040835464
676
+ - 4K/56452d9cd9e2e034764132a57058517ab085e1352f9664d2d1e10e98741334fc
677
+ - 4K/6d81c5ab0d480fd43d78b75ff372a8113ad38e2c03f1d69627c009883054d4c2
678
+ - 4K/9641a1ed7963ce5ca734cff3e6ccea3dfa8bcb0b0a3ff78f65d32a080de2d71e
679
+ - 4K/d3812aad538261e7f73c75762ff55f23b468bcc76f376d52ac86ca6cf3c44b4b
680
+ - 5K/0a485338bbdaf19ba9090b874bb36ef0599a9c9a475a382c22903cf5981c6ea6
681
+ - 5K/119fd56d3797e2d349ca64ddcc5851463cd13b5974b5b2e4566ed5cf7e02e6c1
682
+ - 5K/c37726ce770ac50a2cf5c0f43022f0268e26da0d777cd8e3a3418c4eed03fd94
683
+ - 5K/dac9796dd69e1c25277e29d7e30af4f21e3b575d62a0a208c2b3d1211e2d5d77
684
+ - 6K/06da796666297fe4c683c231edf56ec00148a6a52ab5bb159fe1be31f53a58df
685
+ - 6K/7a9f97660be8f2421b37dd413c898360532f17c6447c11515e8aded2d80eb99d
686
+ - 6K/7da3db99059a436adf537bc23309a6a4db9e011696d086f3a64037f7497d9df7
687
+ - 6K/e5684b3292bfd77db297839fc37ee4cce7fd59775af1a6a4827e3b4f59c036d3
688
+ - 6K/ec1e44d4dc0f8fa77610866495f9297a7f82158c43e1777668b84fd4b736c7bc
689
+ - 6K/ef59aac437132bfc1dd45a7e1f8e4800978e7bb28bf98c4428d26fb3e1da3e90
690
+ - 6K/f477ffc4b398bed8e0d921f0fba9825ca63f317381c535c84be23be991ae1d7a
691
+ - 7K/0569e83fdc248a51fc0ab082ce5e2baff15755c53c207f545e6d02d91f01d166
692
+ - 7K/073f5a9b983ced6fb28b23051260558b165f328a16b2d33fe20585b7ee4ad561
693
+ - 7K/183dd248f6a86e07c5adf9de8ee2d0abe45b1216331c03678e89634c2e9b1c7f
694
+ - 7K/1ba74c22670ad047981441581d00f26f4a148d1010bcac7468c615adf5fa4d5d
695
+ - 7K/389a460ca1995e0658e85fe8e6b520b4e88b370cd6710dfe728b1564bba31aee
696
+ - 7K/493816813d2d6d248eb3c2b0b77b63e54235266e9a06e270fd0d282f13960493
697
+ - 7K/4ae797d07b6d1644c9db6919c8cc8c0d28d72be45108ac7a3abf8dc21b599d83
698
+ - 7K/50c46cf8b8b22c8d2ffdef8964b05ddbceaef312c9a9ff331d1ecebfd223f72a
699
+ - 7K/565553aa894be621e8b4773cac288e60ad0c2cf7edb621be62b348c9a0f78380
700
+ - 7K/599ca3e04cae3ec83affc426af7d0d7ab36eb91cd8e539edbc13070a4d455792
701
+ - 7K/5c8dafad7d782c76ffad8c14e9e1244ce2b83aa12324c54a3cc10176964acf04
702
+ - 7K/63798f5c6fbfcb4eb686268248b8ecbc8d87d920b2bcce967eeaedfd3b3b6d82
703
+ - 7K/946f49be73928469000baa5ca04d2573137c5ee6a66362bcf8d130354dca8924
704
+ - 7K/9e9a89ae6fed06d6e2f4749b4b0059f35ca97f848cedc4a14345999e746f7884
705
+ - 7K/cd9c981eeb4a9091547af19181b382698e9d9eee0a838c7c9783a8a268af6aee
706
+ - 7K/d4fbeba0168af8fddb2fc695881787aedcd62f477c7dcec9ebca7b8594bbd95b
707
+ scannet:
708
+ stanza: scannet
709
+ caption_windows: true
710
+ pointodyssey_train:
711
+ stanza: pointodyssey_train
712
+ caption_windows: true
713
+ # Its local captions cover 96 native frames. A 49-frame draw must use
714
+ # stride 1 to stay inside that exact caption interval; the inherited
715
+ # [2, 4, 6, 8] choices do not fit any 96-frame caption window.
716
+ caption_compatible_temporal_strides:
717
+ - 1
718
+ robot_sources:
719
+ - name: robotwin
720
+ weight: 10.0
721
+ frames: 49
722
+ stride: 4
723
+ # Simulator ground truth: depth, camera and the arm tracks are generated targets (owner decision 2026-11-05); DINO too.
724
+ condition_only_streams: []
725
+ reader:
726
+ source_split_manifest: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/flexpi_robotwin_robot_arm_track_video_20261023/split_manifest.json
727
+ source_split_sha256: 10467f314d57acf8deba64b94d0842b0dedcc256acf5f055df92203f7bf10f4e
728
+ pretrained_norm_stats: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/robotwin_action_image_clean_train_filtered_20261102/native_stats/dataset_stats.json
729
+ clean_train_split:
730
+ manifest: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/robotwin_action_image_clean_train_filtered_20261102/split_manifest.json
731
+ sha256: 6511ed1a88a1047d74fb80dd11b1304f6d77abce6cbe3f56df4f5cf7b9f9ee7c
732
+ train_episodes: 26288
733
+ held_out_episodes: 150
734
+ excluded_clamped_episodes: 1062
735
+ geometry:
736
+ simulator_ground_truth: true
737
+ log_factor: 0.4
738
+ raymap_camera: {origin_fraction: 0.3333333333333333, origin_range: 1.0, decode_margin: 16, origin_box_grid: 1}
739
+ gauge: window_labelled_median_depth
740
+ release_root: /datasets/engs2907/FlexPi/robotwin_3d
741
+ calibration_root: /datasets/engs2907/FlexPi/robotwin_3d/calibration
742
+ # Two real clean-train windows at 21 frames through reader, DINO and the native batch adapter, on pin f6992c4b.
743
+ readback_receipt:
744
+ path: /scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_nano_two_training_restart_20261107/attempt01/robot49_cpu/results/robotwin_readback.json
745
+ sha256: 785c7dde5f5f69a6a6fe30fe6accf56fe2380ab8903b7c9f2ec625d0501ecf2f
746
+ - name: libero
747
+ weight: 6.0
748
+ frames: 49
749
+ stride: 4
750
+ condition_only_streams: []
751
+ reader:
752
+ hydra_overrides: []
753
+ geometry:
754
+ simulator_ground_truth: true
755
+ readback_receipt:
756
+ path: /scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_nano_two_training_restart_20261107/attempt01/robot49_cpu/results/libero_readback.json
757
+ sha256: c8d49b7d87f64b45f1183ab8babc3c17d632e5345e0f697ed4a2aebc6fac4a0f
758
+ real_video_sources: []
759
+ molmo_motion_sources:
760
+ - name: molmo_motion
761
+ weight: 6.0
762
+ frames: 49
763
+ # The camera is an input only, like every real source's camera; there is no depth. DINO and the tracking video are targets.
764
+ condition_only_streams: [camera]
765
+ reader:
766
+ config: train/mot/wan_camera_omega_video_depth_vae_and_sparse_future_prediction_molmo_motion_21_frames_512
767
+ stanza: molmo_motion
768
+ release_root: /vgg/yushi/dataset/MolmoMotion-1M
769
+ subsets: [egodex, xperience, stereo4d, hepic]
770
+ # Retain the prior relative weights for this initial 49-frame launch; these are not a newly measured 49-frame census.
771
+ subset_weights: {egodex: 133.6, xperience: 432.4, stereo4d: 210.9, hepic: 63.1}
772
+ temporal_stride: {egodex: 2, xperience: 1, stereo4d: 4, hepic: 1}
773
+ processed_release_only_subsets: [egodex, hepic]
774
+ # EgoDex and HD-EPIC use their processed release videos; neither raw-video path is used by this recipe.
775
+ # The pinned config leaves Xperience's `object` records out: their 3D was lifted with the 256-pixel camera matrix.
776
+ excluded_objects: {xperience: [object]}
777
+ max_window_redraws: 64
778
+ exclusions:
779
+ # The registered 120-record test cohort (the VGG held-out spec molmo_motion_test_120_records_droid_egodex_molmospaces_
780
+ # 49_frames_512; tools/omni4d/write_molmo_motion_test_cohort_manifest.py). The release's own test splits (DROID's too) and
781
+ # PointMotionBench (hot3d, worldtrack, davis) are asserted disjoint from the train records in code.
782
+ test_cohort_manifest:
783
+ path: /scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_molmo_motion_20261105/cohort/molmo_motion_test_cohort_120_records.json
784
+ sha256: d1f8e8d8ad8f8f67c8bfd23319cdea026ea8cfcba1194b5f4594366e9f64fd76
785
+ point_motion_bench_root: /scratch/shared/engs2907/PointMotionBench
786
+ readback_receipt:
787
+ path: /scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_molmo_49frame_source_readback_20261107/attempt01/native_group_readback/molmo_motion_readback.json
788
+ sha256: ec0577a78e97db76f14cca0bd4d44ab87bc4e588f615fca242c1d675abb8dbb1
789
+ launch_blockers: {}
790
+ mixture_policy:
791
+ total_weight: 139.0
792
+ group_weights: {synthetic: 79.0, real_video: 22.0, robot: 26.0, tracks: 6.0, human_video: 6.0}
793
+ group_share_percent: {synthetic: 56.8, real_video: 15.8, robot: 18.7, tracks: 4.3, human_video: 4.3}
794
+ source_groups:
795
+ four_d_pretraining_relaunch_syn4d_sources: synthetic
796
+ four_d_pretraining_relaunch_omniworld_game_source: synthetic
797
+ four_d_pretraining_relaunch_synthverse_source: synthetic
798
+ four_d_pretraining_relaunch_vkitti_source: synthetic
799
+ four_d_pretraining_relaunch_tartanair_sky_then_far_percentile_source: synthetic
800
+ four_d_pretraining_relaunch_dynamic_replica_stored_depth_source: synthetic
801
+ four_d_pretraining_relaunch_spring_stored_depth_source: synthetic
802
+ four_d_pretraining_relaunch_midair_within_200_meters_source: synthetic
803
+ four_d_pretraining_relaunch_mvs_synth_1000_then_far_percentile_source: synthetic
804
+ four_d_pretraining_relaunch_tartanground_sky_then_far_percentile_source: synthetic
805
+ four_d_pretraining_relaunch_paralleldomain4d_sky_then_far_percentile_source: synthetic
806
+ scannet: real_video
807
+ dl3dv: real_video
808
+ wildrgbd: real_video
809
+ waymo: real_video
810
+ robotwin: robot
811
+ droid: robot
812
+ libero: robot
813
+ pointodyssey_train: tracks
814
+ kubric_train: tracks
815
+ molmo_motion: human_video
cosmos_nano_all_sources_hdepic_49frame_update_014720_20261108/attempt01/update_014720/syn4d_readback.json ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "blockers": [],
3
+ "failures": {},
4
+ "model_forward_or_training": false,
5
+ "sources_read": [
6
+ "four_d_pretraining_relaunch_syn4d_sources"
7
+ ],
8
+ "status": "passed",
9
+ "windows_yaml": "/scratch/shared/engs2907/logs/vgg_forecast_4d/train/cosmos_nano_49_frame_midtraining_all_sources_hdepic_without_cityscapes_from14700_20261107/attempt03/source_windows.yaml",
10
+ "windows_yaml_sha256": "6f4010e7495eafbbbd6a6c74db5433888f3307f8a7381ee863894eeed00bf87e"
11
+ }