yslan commited on
Commit
4c26673
·
verified ·
1 Parent(s): b564f8d

Back up checkpoint lineage metadata cosmos_nano_update_014700_20261107

Browse files
cosmos_nano_full20_21frame_update_014700_20261107/attempt01/update_014700/backup_context.json ADDED
@@ -0,0 +1,17 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "base_model": "nvidia/Cosmos3-Nano@e59a53c25979a090fa8706c9acc0c254a6e89b92",
3
+ "checkpoint_bytes": 2990807208,
4
+ "checkpoint_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",
5
+ "checkpoint_role": "trainable_checkpoint_only",
6
+ "checkpoint_sha256": "ae92b4b891df0a6f44e22027139f48f8f6537a43f78a78f92c54a589853209fb",
7
+ "checkpoint_update": 14700,
8
+ "frozen_base_included": false,
9
+ "optimizer_state_included": false,
10
+ "resolved_config_provenance": "No original resolved Hydra dump was present in the run artifacts. resolved_config.yaml is a derived recipe summary. Exact source module, launcher, restart.env, restart plan, source windows, 11500 parent receipt and evaluation lineage are included.",
11
+ "resolved_hydra_dump_available": false,
12
+ "run_id": "cosmos_nano_full20_21frame_update_014700_20261107",
13
+ "schema": "cosmos_trainable_overlay_hf_backup_context_v1",
14
+ "source_code_commit": "97f885005227cf23bbfb33156f87751db9f14a17",
15
+ "warm_start_sha256": "989dca07ce763dece176e60a6b55f960a69e73ea1791d3cb01869612f1e23727",
16
+ "warm_start_update": 11500
17
+ }
cosmos_nano_full20_21frame_update_014700_20261107/attempt01/update_014700/backup_manifest.json ADDED
@@ -0,0 +1,91 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "checkpoint_bytes": 2990807208,
3
+ "checkpoint_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",
4
+ "checkpoint_role": "trainable_checkpoint_only",
5
+ "checkpoint_sha256": "ae92b4b891df0a6f44e22027139f48f8f6537a43f78a78f92c54a589853209fb",
6
+ "frozen_base_included": false,
7
+ "metadata_files": [
8
+ {
9
+ "bytes": 1204,
10
+ "path": "backup_context.json",
11
+ "role": "lineage_metadata",
12
+ "sha256": "55fe93ca5fa5b9b71196c08899669aad15af8c4b3245371d9d4073462e958e57"
13
+ },
14
+ {
15
+ "bytes": 41,
16
+ "path": "code_commit.txt",
17
+ "role": "lineage_metadata",
18
+ "sha256": "cc953883e6b6b896556689d3d84b55f14af180fbbb66fdc92ccbeec9e8de628b"
19
+ },
20
+ {
21
+ "bytes": 1324,
22
+ "path": "evaluation_lineage.json",
23
+ "role": "lineage_metadata",
24
+ "sha256": "01b40226216646055627cadc4a7b30ae27c9ed97f06c237ae5675823987c7b1e"
25
+ },
26
+ {
27
+ "bytes": 4440,
28
+ "path": "experiment_module.py",
29
+ "role": "lineage_metadata",
30
+ "sha256": "3664abc86b0ba16e7969cd80f339e3abf8f351fd6570288c11ac237ea43e1c79"
31
+ },
32
+ {
33
+ "bytes": 38946,
34
+ "path": "input_step_011500_receipt.json",
35
+ "role": "lineage_metadata",
36
+ "sha256": "a520d263321c55cc2b88230c695c4a696a53cbcfdc2629c84b70c13f52b0cca5"
37
+ },
38
+ {
39
+ "bytes": 1085,
40
+ "path": "launch.sh",
41
+ "role": "lineage_metadata",
42
+ "sha256": "602cdee1bc6a69551ab76b1d4cd2d5c15e4f08465420cc28a80365fff1f17000"
43
+ },
44
+ {
45
+ "bytes": 1490,
46
+ "path": "recipe_summary.yaml",
47
+ "role": "lineage_metadata",
48
+ "sha256": "0aeab55d544ad81d2f859e6f529586c0802d618f99fd6669562b790bb0e6718a"
49
+ },
50
+ {
51
+ "bytes": 1419,
52
+ "path": "restart.env",
53
+ "role": "lineage_metadata",
54
+ "sha256": "8f679cbb9e46da510fb8b9934666d346702e884dc7d7f9e301f45e251936c619"
55
+ },
56
+ {
57
+ "bytes": 1105,
58
+ "path": "restart_plan.json",
59
+ "role": "lineage_metadata",
60
+ "sha256": "d48f0cc440904c2422f6b7909dc363d6b4fe1a21916dd47db4307c6513c72745"
61
+ },
62
+ {
63
+ "bytes": 41,
64
+ "path": "run_source_sha.txt",
65
+ "role": "lineage_metadata",
66
+ "sha256": "cc953883e6b6b896556689d3d84b55f14af180fbbb66fdc92ccbeec9e8de628b"
67
+ },
68
+ {
69
+ "bytes": 42099,
70
+ "path": "source_windows.yaml",
71
+ "role": "lineage_metadata",
72
+ "sha256": "73c90b72adc27e1589f97bff64cb35ace5d03acf7fdac43a3e5a5c86055178d5"
73
+ },
74
+ {
75
+ "bytes": 42099,
76
+ "path": "source_windows_as_recorded.yaml",
77
+ "role": "lineage_metadata",
78
+ "sha256": "73c90b72adc27e1589f97bff64cb35ace5d03acf7fdac43a3e5a5c86055178d5"
79
+ },
80
+ {
81
+ "bytes": 32978,
82
+ "path": "training_launcher_source.sh",
83
+ "role": "lineage_metadata",
84
+ "sha256": "abc48951e0a04d1c5ab48d5d8c32d2b0f54a10df01a4289885e14c3f95280e9e"
85
+ }
86
+ ],
87
+ "optimizer_state_included": false,
88
+ "remote_prefix": "cosmos_nano_full20_21frame_update_014700_20261107/attempt01/update_014700",
89
+ "repository": "yslan/o-intermediate-model",
90
+ "schema": "vgg_trainable_checkpoint_backup_manifest_v1"
91
+ }
cosmos_nano_full20_21frame_update_014700_20261107/attempt01/update_014700/code_commit.txt ADDED
@@ -0,0 +1 @@
 
 
1
+ 97f885005227cf23bbfb33156f87751db9f14a17
cosmos_nano_full20_21frame_update_014700_20261107/attempt01/update_014700/evaluation_lineage.json ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "checkpoint_is_trainable_overlay_only": true,
3
+ "checkpoint_update": 14700,
4
+ "dataset_windows_path": "/scratch/shared/engs2907/logs/vgg_forecast_4d/train/cosmos_next_version_cityscapes_molmo_22_sources_main_process_loader_20261106/attempt02/source_windows.yaml",
5
+ "dataset_windows_sha256": "73c90b72adc27e1589f97bff64cb35ace5d03acf7fdac43a3e5a5c86055178d5",
6
+ "evaluations": [
7
+ {
8
+ "name": "22-source static full cohort update14700",
9
+ "path": "/scratch/shared/engs2907/logs/vgg_forecast_4d/eval/cosmos_nano_static_full_cohort_update14700_20261107/attempt02",
10
+ "scope": "static depth/geometry; see source result files for exact row/status counts"
11
+ },
12
+ {
13
+ "name": "saved-layout temporal and 9-frame evaluations",
14
+ "path": "/scratch/shared/engs2907/logs/vgg_forecast_4d/eval/cosmos_22source_every_overlay_20261106/step_014700",
15
+ "scope": "distinct legacy-caption/empty-caption protocols; do not merge their results"
16
+ }
17
+ ],
18
+ "quality_claim": "none in this backup manifest",
19
+ "schema": "cosmos_checkpoint_evaluation_lineage_v1",
20
+ "training_run": "cosmos_next_version_cityscapes_molmo_22_sources_main_process_loader_20261106/attempt02",
21
+ "training_source_commit": "97f885005227cf23bbfb33156f87751db9f14a17",
22
+ "vgg_data_pin_commit": "f6992c4b67441dc9778642a252447a2687856116"
23
+ }
cosmos_nano_full20_21frame_update_014700_20261107/attempt01/update_014700/experiment_module.py ADDED
@@ -0,0 +1,80 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
2
+ # SPDX-License-Identifier: OpenMDW-1.1
3
+
4
+ """Opt-in B16 continuation of the 21-frame mixture with Cityscapes and eligible MolmoMotion data."""
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_plain_ddp_from_7300 import (
14
+ END_UPDATE,
15
+ OUTPUT as RUNNING_OUTPUT,
16
+ WRITER_CONFIG_IDENTITY as RUNNING_WRITER_CONFIG_IDENTITY,
17
+ recipe as running_recipe,
18
+ )
19
+ from cosmos_framework.inference.omni4d_overlay_config_identities import (
20
+ CITYSCAPES_AND_MOLMO_FROM_OVERLAY_CONFIG_IDENTITY,
21
+ TWENTY_ONE_FRAME_FROM_OVERLAY_LR_2P5E5_CONFIG_IDENTITY,
22
+ )
23
+
24
+ DEFAULT_WINDOWS_YAML = (
25
+ Path(__file__).resolve().parents[4]
26
+ / "configs/omni4d_vgg_stage_one_next_version_all_sources_plus_cityscapes_and_molmo_motion_21_frames_ddp_global32_b16_gas1.yaml"
27
+ )
28
+ WINDOWS_YAML = Path(os.environ.get("COSMOS_VGG_WINDOWS_CONFIG", str(DEFAULT_WINDOWS_YAML)))
29
+ OUTPUT = "/scratch/shared/engs2907/logs/vgg_forecast_4d/train/cosmos_next_version_full20_plus_cityscapes_and_molmo_motion_21_frames_ddp_from_explicit_overlay_20261106/attempt01"
30
+ WRITER_CONFIG_IDENTITY = CITYSCAPES_AND_MOLMO_FROM_OVERLAY_CONFIG_IDENTITY
31
+ if OUTPUT == RUNNING_OUTPUT or WRITER_CONFIG_IDENTITY == RUNNING_WRITER_CONFIG_IDENTITY:
32
+ raise ValueError("the combined real-data variant requires its own output directory and overlay identity")
33
+
34
+ recipe = deepcopy(running_recipe)
35
+ recipe.dataloader_train.dataset.windows_config_path = str(WINDOWS_YAML)
36
+ recipe.dataloader_train.dataset.expected_windows_config_sha256 = hashlib.sha256(
37
+ WINDOWS_YAML.read_bytes()
38
+ ).hexdigest()
39
+ recipe.dataloader_train.dataset.expected_vgg_commit = "f6992c4b67441dc9778642a252447a2687856116"
40
+ recipe.model.trainable_overlay_config_identity = TWENTY_ONE_FRAME_FROM_OVERLAY_LR_2P5E5_CONFIG_IDENTITY
41
+ recipe.job.name = "cosmos3_nano_full20_plus_cityscapes_and_molmo_motion_21_frames_plain_ddp_from_explicit_overlay_20261106"
42
+ recipe.dataloader_train.batch_size = 16
43
+ recipe.trainer.grad_accum_iter = 1
44
+ recipe.trainer.callbacks.omni4d_trainable_overlay.output_dir = f"{OUTPUT}/trainable_overlays"
45
+ recipe.trainer.callbacks.omni4d_trainable_overlay.config_identity = WRITER_CONFIG_IDENTITY
46
+ recipe.trainer.callbacks.omni4d_trainable_overlay.save_updates = list(range(100, END_UPDATE + 1, 100))
47
+
48
+
49
+ def require_launch_ready() -> None:
50
+ """Require both new sources, their pinned readback/caption products and exact mixture arithmetic."""
51
+ import yaml
52
+
53
+ from cosmos_framework.data.omni4d_next_version_readiness import check_mixture_arithmetic, launch_blockers
54
+
55
+ payload = yaml.safe_load(WINDOWS_YAML.read_text(encoding="utf-8"))
56
+ city = [item for item in payload.get("real_video_sources", []) if item.get("name") == "cityscapes"]
57
+ molmo = payload.get("molmo_motion_sources", [])
58
+ if len(city) != 1 or float(city[0].get("weight", -1)) != 4.0 or int(city[0].get("frames", -1)) != 21:
59
+ raise RuntimeError("the combined recipe needs one 21-frame Cityscapes source at weight 4")
60
+ if len(molmo) != 1 or molmo[0].get("name") != "molmo_motion" or float(molmo[0].get("weight", -1)) != 6.0:
61
+ raise RuntimeError("the combined recipe needs one MolmoMotion source at weight 6")
62
+ reader = molmo[0].get("reader", {})
63
+ if set(reader.get("subsets", [])) != {"egodex", "xperience", "stereo4d"}:
64
+ raise RuntimeError("the combined recipe may use only eligible EgoDex, Xperience and Stereo4D records")
65
+ if set(reader.get("processed_release_only_subsets", [])) != {"egodex"}:
66
+ raise RuntimeError("EgoDex must be restricted to processed release MP4s; raw RGB is not aligned/eligible")
67
+ if float(payload["mixture_policy"]["total_weight"]) != 143.0:
68
+ raise RuntimeError("the 22-source combined recipe must have total weight 143")
69
+ check_mixture_arithmetic(payload)
70
+ blockers = launch_blockers(payload)
71
+ if blockers:
72
+ raise RuntimeError("the combined 21-frame data mix is not ready to launch:\n" + "\n".join(f"- {v}" for v in blockers))
73
+
74
+
75
+ ConfigStore.instance().store(
76
+ group="experiment",
77
+ package="_global_",
78
+ name="omni4d_stage_one_lora_r256_next_version_all_sources_plus_cityscapes_and_molmo_motion_21_frames_plain_ddp_from_explicit_overlay",
79
+ node=recipe,
80
+ )
cosmos_nano_full20_21frame_update_014700_20261107/attempt01/update_014700/input_step_011500_receipt.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "source": "/scratch/shared/engs2907/logs/vgg_forecast_4d/train/cosmos_next_version_full20_21_frames_ddp_b16_gas1_lr_2p5e5_20261105/attempt01/trainable_overlays/step_011500.safetensors",
3
+ "protected_path": "/scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_cityscapes_molmo_production_restart_20261106/attempt01/input_step_011500.safetensors",
4
+ "bytes": 2990807176,
5
+ "sha256": "989dca07ce763dece176e60a6b55f960a69e73ea1791d3cb01869612f1e23727",
6
+ "metadata": {
7
+ "format": "omni4d_trainable_weights_v1",
8
+ "base_identity": "nvidia/Cosmos3-Nano@e59a53c25979a090fa8706c9acc0c254a6e89b92",
9
+ "update": "11500",
10
+ "config_identity": "cosmos3_nano_omni4d_stage_one_lora_r256_plain_ddp_21_frames_temporal_forcing_full20_from_overlay_global32_lr2p5e5",
11
+ "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\"]",
12
+ "exact_training_resume": "false"
13
+ },
14
+ "tensor_count": 530
15
+ }
cosmos_nano_full20_21frame_update_014700_20261107/attempt01/update_014700/launch.sh ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+ set -Eeuo pipefail
3
+ stage=/scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_cityscapes_molmo_production_restart_20261106/attempt02
4
+ source "$stage/restart.env"
5
+ [[ "$(git -C "$COSMOS_SOURCE_ROOT" rev-parse HEAD)" == "$COSMOS_SOURCE_COMMIT" ]]
6
+ git -C "$COSMOS_SOURCE_ROOT" diff --quiet HEAD --
7
+ export COSMOS_ALLOCATION_ID=72961 COSMOS_EXPECTED_NODE=civo-b200-1
8
+ export COSMOS_ALLOCATED_GPU_UUIDS=GPU-7899c40c-79a9-ead3-59d5-f5d7db271f64,GPU-41b7d2ea-0eeb-0e84-4b04-b2baddede118
9
+ export COSMOS_MICROBATCH_PROFILE=b16_gas1 COSMOS_TWENTY_UPDATE_ADMISSION_TEST=0
10
+ export COSMOS_PYTHON=/scratch/shared/engs2907/envs/cosmos3-nano-py313-cu128-001fe183/bin/python COSMOS_NATIVE_STALL_DIAGNOSTICS=1
11
+ export COSMOS_COMPUTE_PYTHON="$COSMOS_PYTHON"
12
+ export COSMOS_VGG_WORKER_STACK_DIR="$COSMOS_ATTEMPT_ROOT/worker_stacks"
13
+ export COSMOS_VGG_WORKER_STACK_SOURCE_IDENTITY="${COSMOS_SOURCE_COMMIT}_windows_${COSMOS_VGG_WINDOWS_CONFIG_SHA256}_overlay_${COSMOS_TRAINABLE_OVERLAY_UPDATE}"
14
+ exec bash "$COSMOS_SOURCE_ROOT/tools/omni4d/cosmos3_nano/run_formal_five_stream_5k_on_b200.sh"
cosmos_nano_full20_21frame_update_014700_20261107/attempt01/update_014700/recipe_summary.yaml ADDED
@@ -0,0 +1,34 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ run_id: cosmos_nano_full20_21frame_update_014700_20261107
2
+ checkpoint: step_014700.safetensors
3
+ 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
4
+ source:
5
+ cosmos_training_commit: 97f885005227cf23bbfb33156f87751db9f14a17
6
+ experiment_module: cosmos_framework.configs.base.experiment.sft.omni4d_stage_one_lora_r256_next_version_all_sources_plus_cityscapes_and_molmo_motion_21_frames_plain_ddp_from_explicit_overlay
7
+ windows_yaml: source_windows.yaml
8
+ windows_sha256: 73c90b72adc27e1589f97bff64cb35ace5d03acf7fdac43a3e5a5c86055178d5
9
+ vgg_data_pin_commit: f6992c4b67441dc9778642a252447a2687856116
10
+ base_model: nvidia/Cosmos3-Nano@e59a53c25979a090fa8706c9acc0c254a6e89b92
11
+ training:
12
+ source_windows: 22
13
+ frames: 21
14
+ streams: [video, pointmap, camera, dino, track_video]
15
+ precision: bf16
16
+ distributed: native_ddp
17
+ world_size: 2
18
+ per_rank_batch_size: 16
19
+ gradient_accumulation_steps: 1
20
+ global_batch_size: 32
21
+ learning_rate: 0.000025
22
+ max_updates: 500000
23
+ warm_start:
24
+ type: trainable_weights_only
25
+ update: 11500
26
+ checkpoint_sha256: 989dca07ce763dece176e60a6b55f960a69e73ea1791d3cb01869612f1e23727
27
+ optimizer_state_restored: false
28
+ rng_state_restored: false
29
+ exact_training_resume: false
30
+ checkpoint:
31
+ contains: trainable_lora_and_projector_parameters_only
32
+ includes_frozen_pretrained_base: false
33
+ includes_optimizer_state: false
34
+ output_update: 14700
cosmos_nano_full20_21frame_update_014700_20261107/attempt01/update_014700/restart.env ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ export COSMOS_SOURCE_ROOT=/scratch/shared/engs2907/cosmos_city_molmo_production_97f885005_20261106
2
+ export COSMOS_SOURCE_COMMIT=97f885005227cf23bbfb33156f87751db9f14a17
3
+ export COSMOS_EXPERIMENT_MODULE=cosmos_framework.configs.base.experiment.sft.omni4d_stage_one_lora_r256_next_version_all_sources_plus_cityscapes_and_molmo_motion_21_frames_plain_ddp_from_explicit_overlay
4
+ export COSMOS_ATTEMPT_ROOT=/scratch/shared/engs2907/logs/vgg_forecast_4d/train/cosmos_next_version_cityscapes_molmo_22_sources_main_process_loader_20261106/attempt02
5
+ export COSMOS_VGG_WINDOWS_CONFIG=/scratch/shared/engs2907/cosmos_city_molmo_production_97f885005_20261106/cosmos_framework/configs/omni4d_vgg_stage_one_next_version_all_sources_plus_cityscapes_and_molmo_motion_21_frames_ddp_global32_b16_gas1.yaml
6
+ export COSMOS_VGG_WINDOWS_CONFIG_SHA256=73c90b72adc27e1589f97bff64cb35ace5d03acf7fdac43a3e5a5c86055178d5
7
+ export COSMOS_TRAINABLE_OVERLAY_PATH=/scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_cityscapes_molmo_production_restart_20261106/attempt01/input_step_011500.safetensors
8
+ export COSMOS_TRAINABLE_OVERLAY_SHA256=989dca07ce763dece176e60a6b55f960a69e73ea1791d3cb01869612f1e23727
9
+ export COSMOS_TRAINABLE_OVERLAY_UPDATE=11500
10
+ export COSMOS_TRAINABLE_OVERLAY_CONFIG_IDENTITY=cosmos3_nano_omni4d_stage_one_lora_r256_plain_ddp_21_frames_temporal_forcing_full20_from_overlay_global32_lr2p5e5
11
+ export COSMOS_TIMING_GATE_UPDATES=50
cosmos_nano_full20_21frame_update_014700_20261107/attempt01/update_014700/restart_plan.json ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "at_utc": "2026-11-06T17:39:02.174967+00:00",
3
+ "source": "/scratch/shared/engs2907/cosmos_city_molmo_production_97f885005_20261106",
4
+ "sha": "97f885005227cf23bbfb33156f87751db9f14a17",
5
+ "input_receipt": "/scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_cityscapes_molmo_production_restart_20261106/attempt01/input_step_011500_receipt.json",
6
+ "supersedes_invalid_step": "72961.15",
7
+ "dataset_constructor": {
8
+ "sample_offset_updates": 11500,
9
+ "batch_size": 16,
10
+ "gradient_accumulation_steps": 1
11
+ },
12
+ "weight_only_fresh_optimizer": true,
13
+ "acceptance_updates": 50,
14
+ "status": "submitted_initializing",
15
+ "test_commit": "8ec3746cf841e5d1e36a2a078152ca9b12541d40",
16
+ "launcher_code_commit": "b20856aba51d1ef9592637dc0747b3e7ebd0ab23",
17
+ "tests": {
18
+ "static_launcher": "1 passed,4 deselected,40.52s",
19
+ "executable_config_composition": "1 passed,39.64s"
20
+ },
21
+ "test_artifacts": "/scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_overlay_data_offset_alignment_20261106/attempt01",
22
+ "slurm_step": "72961.16",
23
+ "submitted_utc": "2026-11-06T17:43:06.667772+00:00"
24
+ }
cosmos_nano_full20_21frame_update_014700_20261107/attempt01/update_014700/run_source_sha.txt ADDED
@@ -0,0 +1 @@
 
 
1
+ 97f885005227cf23bbfb33156f87751db9f14a17
cosmos_nano_full20_21frame_update_014700_20261107/attempt01/update_014700/source_windows.yaml ADDED
@@ -0,0 +1,834 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # B16 continuation of the full 20-source mixture with native-resolution Cityscapes captions.
2
+ # This is a separate model-only run from an explicit five-stream overlay: 22 sources, total weight 143.
3
+ # Cityscapes train captions are converted from the completed native2048 product; val remains evaluation-only.
4
+ # Next Cosmos stage-one version: the owner's weighted mixture at 21 frames (decision 2026-11-05).
5
+ #
6
+ # Weights, not shares. Total 143 over 22 logical sources:
7
+ # synthetic 79 (59.4%) 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 26 (18.2%) ScanNet 8, DL3DV 6, WildRGB-D 4, Waymo 4, Cityscapes 4 (geometry is input only)
11
+ # robot 26 (19.5%) 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 ( 4.2%) PointOdyssey 3, Kubric 3
14
+ # human video 6 ( 4.2%) MolmoMotion; 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 21 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: 21
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: 16
321
+ stream_seed: 42
322
+ epoch_windows: 32000000
323
+ val_set_proportion: 0.0
324
+ resume_state: null
325
+ gradient_accumulation_steps: 1
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
+ robot_sources:
714
+ - name: robotwin
715
+ weight: 10.0
716
+ frames: 21
717
+ stride: 4
718
+ # Simulator ground truth: depth, camera and the arm tracks are generated targets (owner decision 2026-11-05); DINO too.
719
+ condition_only_streams: []
720
+ reader:
721
+ source_split_manifest: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/flexpi_robotwin_robot_arm_track_video_20261023/split_manifest.json
722
+ source_split_sha256: 10467f314d57acf8deba64b94d0842b0dedcc256acf5f055df92203f7bf10f4e
723
+ pretrained_norm_stats: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/robotwin_action_image_clean_train_filtered_20261102/native_stats/dataset_stats.json
724
+ clean_train_split:
725
+ manifest: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/robotwin_action_image_clean_train_filtered_20261102/split_manifest.json
726
+ sha256: 6511ed1a88a1047d74fb80dd11b1304f6d77abce6cbe3f56df4f5cf7b9f9ee7c
727
+ train_episodes: 26288
728
+ held_out_episodes: 150
729
+ excluded_clamped_episodes: 1062
730
+ geometry:
731
+ simulator_ground_truth: true
732
+ log_factor: 0.4
733
+ raymap_camera: {origin_fraction: 0.3333333333333333, origin_range: 1.0, decode_margin: 16, origin_box_grid: 1}
734
+ gauge: window_labelled_median_depth
735
+ release_root: /datasets/engs2907/FlexPi/robotwin_3d
736
+ calibration_root: /datasets/engs2907/FlexPi/robotwin_3d/calibration
737
+ # Two real clean-train windows at 21 frames through reader, DINO and the native batch adapter, on pin f6992c4b.
738
+ readback_receipt:
739
+ path: /scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_city_molmo_pin_migration_20261106/attempt05/robot_sources/robotwin_readback.json
740
+ sha256: 08c29011548725201118a245372365c0fd2430eafa6ee4b2844d2cf2fdca59ff
741
+ - name: libero
742
+ weight: 6.0
743
+ frames: 21
744
+ stride: 4
745
+ condition_only_streams: []
746
+ reader:
747
+ hydra_overrides: []
748
+ geometry:
749
+ simulator_ground_truth: true
750
+ readback_receipt:
751
+ path: /scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_city_molmo_pin_migration_20261106/attempt05/robot_sources/libero_readback.json
752
+ sha256: 5dd74fbd926bc0b7aac7e27da43ae55c1eaa9cb049b9c4656ba27eac71bdfedc
753
+ real_video_sources:
754
+ - name: cityscapes
755
+ weight: 4.0
756
+ frames: 21
757
+ stride: 1
758
+ condition_only_streams:
759
+ - pointmap
760
+ reader:
761
+ data_root: /scratch/shared/engs2907/datasets/cityscapes
762
+ train_manifest: /vgg/yushi/dataset/Cityscapes/prepared/cityscapes_train_driving_sequence_v1.jsonl
763
+ train_manifest_sha256: 9dd2abd964d52c752401b2aef6baa400ec52268143da8b8c623cb059b8f70475
764
+ train_snippets: 2975
765
+ train_frames: 89250
766
+ val_manifest: /vgg/yushi/dataset/Cityscapes/prepared/cityscapes_val_driving_sequence_v1.jsonl
767
+ val_manifest_sha256: 7b6fb5cc52bd4eac0a61874ee14b47988481a6a484ed73f8ca4c959bc1aa5adc
768
+ geometry:
769
+ hole_fill: telea_log_depth
770
+ captions:
771
+ product: cityscapes_gemini_holo4d
772
+ complete_marker: /vgg/yushi/dataset/caption_output/gemini_holo4d/native2048_v1/cityscapes/COMPLETE.json
773
+ texts: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/cosmos_caption_texts_20261106/cityscapes_gemini_holo4d
774
+ molmo_motion_sources:
775
+ - name: molmo_motion
776
+ weight: 6.0
777
+ frames: 21
778
+ # The camera is an input only, like every real source's camera; there is no depth. DINO and the tracking video are targets.
779
+ condition_only_streams: [camera]
780
+ reader:
781
+ config: train/mot/wan_camera_omega_video_depth_vae_and_sparse_future_prediction_molmo_motion_21_frames_512
782
+ stanza: molmo_motion
783
+ release_root: /vgg/yushi/dataset/MolmoMotion-1M
784
+ subsets: [egodex, xperience, stereo4d]
785
+ # molmo-motion's sqrt-frequency mixing (p_i proportional to sqrt N_i, README "Training"), N_i the drawable records of
786
+ # the subset at this window length measured by the readback (census of 2026-11-05).
787
+ subset_weights: {egodex: 133.6, xperience: 432.4, stereo4d: 210.9}
788
+ temporal_stride: {egodex: 2, xperience: 1, stereo4d: 4}
789
+ processed_release_only_subsets: [egodex]
790
+ # Only the already reconstructed release MP4 subset is eligible. Never fall back to newly downloaded raw 30 fps
791
+ # EgoDex videos: their RGB raster has not been validated against the 15 fps track/camera annotations.
792
+ processed_release_only_subsets: [egodex]
793
+ # The pinned config leaves Xperience's `object` records out: their 3D was lifted with the 256-pixel camera matrix.
794
+ excluded_objects: {xperience: [object]}
795
+ max_window_redraws: 64
796
+ exclusions:
797
+ # The registered 120-record test cohort (the VGG held-out spec molmo_motion_test_120_records_droid_egodex_molmospaces_
798
+ # 49_frames_512; tools/omni4d/write_molmo_motion_test_cohort_manifest.py). The release's own test splits (DROID's too) and
799
+ # PointMotionBench (hot3d, worldtrack, davis) are asserted disjoint from the train records in code.
800
+ test_cohort_manifest:
801
+ path: /scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_molmo_motion_20261105/cohort/molmo_motion_test_cohort_120_records.json
802
+ sha256: d1f8e8d8ad8f8f67c8bfd23319cdea026ea8cfcba1194b5f4594366e9f64fd76
803
+ point_motion_bench_root: /scratch/shared/engs2907/PointMotionBench
804
+ readback_receipt:
805
+ path: /scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_city_molmo_pin_migration_20261106/attempt06/molmo_motion/molmo_motion_readback.json
806
+ sha256: 448aa162950993bfbb101d1933e7b8b7692b6a7cd3ddd2abf8faa10b7c6728ba
807
+ launch_blockers: {}
808
+ mixture_policy:
809
+ total_weight: 143.0
810
+ group_weights: {synthetic: 79.0, real_video: 26.0, robot: 26.0, tracks: 6.0, human_video: 6.0}
811
+ group_share_percent: {synthetic: 55.2, real_video: 18.2, robot: 18.2, tracks: 4.2, human_video: 4.2}
812
+ source_groups:
813
+ four_d_pretraining_relaunch_syn4d_sources: synthetic
814
+ four_d_pretraining_relaunch_omniworld_game_source: synthetic
815
+ four_d_pretraining_relaunch_synthverse_source: synthetic
816
+ four_d_pretraining_relaunch_vkitti_source: synthetic
817
+ four_d_pretraining_relaunch_tartanair_sky_then_far_percentile_source: synthetic
818
+ four_d_pretraining_relaunch_dynamic_replica_stored_depth_source: synthetic
819
+ four_d_pretraining_relaunch_spring_stored_depth_source: synthetic
820
+ four_d_pretraining_relaunch_midair_within_200_meters_source: synthetic
821
+ four_d_pretraining_relaunch_mvs_synth_1000_then_far_percentile_source: synthetic
822
+ four_d_pretraining_relaunch_tartanground_sky_then_far_percentile_source: synthetic
823
+ four_d_pretraining_relaunch_paralleldomain4d_sky_then_far_percentile_source: synthetic
824
+ scannet: real_video
825
+ dl3dv: real_video
826
+ wildrgbd: real_video
827
+ waymo: real_video
828
+ cityscapes: real_video
829
+ robotwin: robot
830
+ droid: robot
831
+ libero: robot
832
+ pointodyssey_train: tracks
833
+ kubric_train: tracks
834
+ molmo_motion: human_video
cosmos_nano_full20_21frame_update_014700_20261107/attempt01/update_014700/source_windows_as_recorded.yaml ADDED
@@ -0,0 +1,834 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # B16 continuation of the full 20-source mixture with native-resolution Cityscapes captions.
2
+ # This is a separate model-only run from an explicit five-stream overlay: 22 sources, total weight 143.
3
+ # Cityscapes train captions are converted from the completed native2048 product; val remains evaluation-only.
4
+ # Next Cosmos stage-one version: the owner's weighted mixture at 21 frames (decision 2026-11-05).
5
+ #
6
+ # Weights, not shares. Total 143 over 22 logical sources:
7
+ # synthetic 79 (59.4%) 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 26 (18.2%) ScanNet 8, DL3DV 6, WildRGB-D 4, Waymo 4, Cityscapes 4 (geometry is input only)
11
+ # robot 26 (19.5%) 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 ( 4.2%) PointOdyssey 3, Kubric 3
14
+ # human video 6 ( 4.2%) MolmoMotion; 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 21 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: 21
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: 16
321
+ stream_seed: 42
322
+ epoch_windows: 32000000
323
+ val_set_proportion: 0.0
324
+ resume_state: null
325
+ gradient_accumulation_steps: 1
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
+ robot_sources:
714
+ - name: robotwin
715
+ weight: 10.0
716
+ frames: 21
717
+ stride: 4
718
+ # Simulator ground truth: depth, camera and the arm tracks are generated targets (owner decision 2026-11-05); DINO too.
719
+ condition_only_streams: []
720
+ reader:
721
+ source_split_manifest: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/flexpi_robotwin_robot_arm_track_video_20261023/split_manifest.json
722
+ source_split_sha256: 10467f314d57acf8deba64b94d0842b0dedcc256acf5f055df92203f7bf10f4e
723
+ pretrained_norm_stats: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/robotwin_action_image_clean_train_filtered_20261102/native_stats/dataset_stats.json
724
+ clean_train_split:
725
+ manifest: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/robotwin_action_image_clean_train_filtered_20261102/split_manifest.json
726
+ sha256: 6511ed1a88a1047d74fb80dd11b1304f6d77abce6cbe3f56df4f5cf7b9f9ee7c
727
+ train_episodes: 26288
728
+ held_out_episodes: 150
729
+ excluded_clamped_episodes: 1062
730
+ geometry:
731
+ simulator_ground_truth: true
732
+ log_factor: 0.4
733
+ raymap_camera: {origin_fraction: 0.3333333333333333, origin_range: 1.0, decode_margin: 16, origin_box_grid: 1}
734
+ gauge: window_labelled_median_depth
735
+ release_root: /datasets/engs2907/FlexPi/robotwin_3d
736
+ calibration_root: /datasets/engs2907/FlexPi/robotwin_3d/calibration
737
+ # Two real clean-train windows at 21 frames through reader, DINO and the native batch adapter, on pin f6992c4b.
738
+ readback_receipt:
739
+ path: /scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_city_molmo_pin_migration_20261106/attempt05/robot_sources/robotwin_readback.json
740
+ sha256: 08c29011548725201118a245372365c0fd2430eafa6ee4b2844d2cf2fdca59ff
741
+ - name: libero
742
+ weight: 6.0
743
+ frames: 21
744
+ stride: 4
745
+ condition_only_streams: []
746
+ reader:
747
+ hydra_overrides: []
748
+ geometry:
749
+ simulator_ground_truth: true
750
+ readback_receipt:
751
+ path: /scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_city_molmo_pin_migration_20261106/attempt05/robot_sources/libero_readback.json
752
+ sha256: 5dd74fbd926bc0b7aac7e27da43ae55c1eaa9cb049b9c4656ba27eac71bdfedc
753
+ real_video_sources:
754
+ - name: cityscapes
755
+ weight: 4.0
756
+ frames: 21
757
+ stride: 1
758
+ condition_only_streams:
759
+ - pointmap
760
+ reader:
761
+ data_root: /scratch/shared/engs2907/datasets/cityscapes
762
+ train_manifest: /vgg/yushi/dataset/Cityscapes/prepared/cityscapes_train_driving_sequence_v1.jsonl
763
+ train_manifest_sha256: 9dd2abd964d52c752401b2aef6baa400ec52268143da8b8c623cb059b8f70475
764
+ train_snippets: 2975
765
+ train_frames: 89250
766
+ val_manifest: /vgg/yushi/dataset/Cityscapes/prepared/cityscapes_val_driving_sequence_v1.jsonl
767
+ val_manifest_sha256: 7b6fb5cc52bd4eac0a61874ee14b47988481a6a484ed73f8ca4c959bc1aa5adc
768
+ geometry:
769
+ hole_fill: telea_log_depth
770
+ captions:
771
+ product: cityscapes_gemini_holo4d
772
+ complete_marker: /vgg/yushi/dataset/caption_output/gemini_holo4d/native2048_v1/cityscapes/COMPLETE.json
773
+ texts: /scratch/shared/engs2907/logs/vgg_forecast_4d/cache/cosmos_caption_texts_20261106/cityscapes_gemini_holo4d
774
+ molmo_motion_sources:
775
+ - name: molmo_motion
776
+ weight: 6.0
777
+ frames: 21
778
+ # The camera is an input only, like every real source's camera; there is no depth. DINO and the tracking video are targets.
779
+ condition_only_streams: [camera]
780
+ reader:
781
+ config: train/mot/wan_camera_omega_video_depth_vae_and_sparse_future_prediction_molmo_motion_21_frames_512
782
+ stanza: molmo_motion
783
+ release_root: /vgg/yushi/dataset/MolmoMotion-1M
784
+ subsets: [egodex, xperience, stereo4d]
785
+ # molmo-motion's sqrt-frequency mixing (p_i proportional to sqrt N_i, README "Training"), N_i the drawable records of
786
+ # the subset at this window length measured by the readback (census of 2026-11-05).
787
+ subset_weights: {egodex: 133.6, xperience: 432.4, stereo4d: 210.9}
788
+ temporal_stride: {egodex: 2, xperience: 1, stereo4d: 4}
789
+ processed_release_only_subsets: [egodex]
790
+ # Only the already reconstructed release MP4 subset is eligible. Never fall back to newly downloaded raw 30 fps
791
+ # EgoDex videos: their RGB raster has not been validated against the 15 fps track/camera annotations.
792
+ processed_release_only_subsets: [egodex]
793
+ # The pinned config leaves Xperience's `object` records out: their 3D was lifted with the 256-pixel camera matrix.
794
+ excluded_objects: {xperience: [object]}
795
+ max_window_redraws: 64
796
+ exclusions:
797
+ # The registered 120-record test cohort (the VGG held-out spec molmo_motion_test_120_records_droid_egodex_molmospaces_
798
+ # 49_frames_512; tools/omni4d/write_molmo_motion_test_cohort_manifest.py). The release's own test splits (DROID's too) and
799
+ # PointMotionBench (hot3d, worldtrack, davis) are asserted disjoint from the train records in code.
800
+ test_cohort_manifest:
801
+ path: /scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_molmo_motion_20261105/cohort/molmo_motion_test_cohort_120_records.json
802
+ sha256: d1f8e8d8ad8f8f67c8bfd23319cdea026ea8cfcba1194b5f4594366e9f64fd76
803
+ point_motion_bench_root: /scratch/shared/engs2907/PointMotionBench
804
+ readback_receipt:
805
+ path: /scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_city_molmo_pin_migration_20261106/attempt06/molmo_motion/molmo_motion_readback.json
806
+ sha256: 448aa162950993bfbb101d1933e7b8b7692b6a7cd3ddd2abf8faa10b7c6728ba
807
+ launch_blockers: {}
808
+ mixture_policy:
809
+ total_weight: 143.0
810
+ group_weights: {synthetic: 79.0, real_video: 26.0, robot: 26.0, tracks: 6.0, human_video: 6.0}
811
+ group_share_percent: {synthetic: 55.2, real_video: 18.2, robot: 18.2, tracks: 4.2, human_video: 4.2}
812
+ source_groups:
813
+ four_d_pretraining_relaunch_syn4d_sources: synthetic
814
+ four_d_pretraining_relaunch_omniworld_game_source: synthetic
815
+ four_d_pretraining_relaunch_synthverse_source: synthetic
816
+ four_d_pretraining_relaunch_vkitti_source: synthetic
817
+ four_d_pretraining_relaunch_tartanair_sky_then_far_percentile_source: synthetic
818
+ four_d_pretraining_relaunch_dynamic_replica_stored_depth_source: synthetic
819
+ four_d_pretraining_relaunch_spring_stored_depth_source: synthetic
820
+ four_d_pretraining_relaunch_midair_within_200_meters_source: synthetic
821
+ four_d_pretraining_relaunch_mvs_synth_1000_then_far_percentile_source: synthetic
822
+ four_d_pretraining_relaunch_tartanground_sky_then_far_percentile_source: synthetic
823
+ four_d_pretraining_relaunch_paralleldomain4d_sky_then_far_percentile_source: synthetic
824
+ scannet: real_video
825
+ dl3dv: real_video
826
+ wildrgbd: real_video
827
+ waymo: real_video
828
+ cityscapes: real_video
829
+ robotwin: robot
830
+ droid: robot
831
+ libero: robot
832
+ pointodyssey_train: tracks
833
+ kubric_train: tracks
834
+ molmo_motion: human_video
cosmos_nano_full20_21frame_update_014700_20261107/attempt01/update_014700/training_launcher_source.sh ADDED
@@ -0,0 +1,528 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+ set -euo pipefail
3
+
4
+ if [[ ${COSMOS_CLEAN_SLURM_ENV:-0} != 1 ]]; then
5
+ export COSMOS_CLEAN_SLURM_ENV=1
6
+ exec python3 - "$0" "$@" <<'PY'
7
+ import os
8
+ import subprocess
9
+ import sys
10
+
11
+ environment = {key: value for key, value in os.environ.items() if not key.startswith("SLURM_")}
12
+ raise SystemExit(subprocess.run(["bash", *sys.argv[1:]], env=environment).returncode)
13
+ PY
14
+ fi
15
+
16
+ repo_root=${COSMOS_SOURCE_ROOT:-$(git rev-parse --show-toplevel)}
17
+ launcher_root=$(cd -- "$(dirname -- "$(readlink -f -- "${BASH_SOURCE[0]}")")/../../.." && pwd)
18
+ allocation=${COSMOS_ALLOCATION_ID:?set COSMOS_ALLOCATION_ID to the owner-approved two-B200 allocation}
19
+ expected_node=${COSMOS_EXPECTED_NODE:?set COSMOS_EXPECTED_NODE to the owner-approved B200 node}
20
+ allocated_gpu_uuids=${COSMOS_ALLOCATED_GPU_UUIDS:?set COSMOS_ALLOCATED_GPU_UUIDS to the two UUIDs verified for this allocation}
21
+ windows_config=${COSMOS_VGG_WINDOWS_CONFIG:?set COSMOS_VGG_WINDOWS_CONFIG to the pinned full-mixture YAML}
22
+ windows_sha256=${COSMOS_VGG_WINDOWS_CONFIG_SHA256:?set COSMOS_VGG_WINDOWS_CONFIG_SHA256 to its SHA-256}
23
+ experiment_module=${COSMOS_EXPERIMENT_MODULE:-cosmos_framework.configs.base.experiment.sft.omni4d_stage_one_formal_5k}
24
+ warmstart_path=${COSMOS_TRAINABLE_OVERLAY_PATH:-}
25
+ next_version_ddp_module=cosmos_framework.configs.base.experiment.sft.omni4d_stage_one_lora_r256_next_version_all_sources_21_frames_temporal_forcing_plain_ddp_from_8200
26
+ next_version_ddp_overlay=/scratch/shared/engs2907/logs/vgg_forecast_4d/tools/cosmos_twenty_one_frame_ddp_handoff_20261105/attempt01/protected_step_008200.safetensors
27
+ next_version_full19_module=cosmos_framework.configs.base.experiment.sft.omni4d_stage_one_lora_r256_next_version_all_sources_21_frames_plain_ddp_from_7300
28
+ next_version_full19_cityscapes_module=cosmos_framework.configs.base.experiment.sft.omni4d_stage_one_lora_r256_next_version_all_sources_plus_cityscapes_21_frames_plain_ddp_from_7300
29
+ next_version_omega_module=cosmos_framework.configs.base.experiment.sft.omni4d_stage_one_lora_r256_next_version_all_sources_21_frames_plain_ddp_from_overlay_with_omega_stream
30
+ next_version_full20_start_identity=cosmos3_nano_omni4d_stage_one_lora_r256_plain_ddp_21_frames_temporal_forcing_full19_from7300_global32
31
+ next_version_full20_writer_identity=cosmos3_nano_omni4d_stage_one_lora_r256_plain_ddp_21_frames_temporal_forcing_full20_from_overlay_global32_lr2p5e5
32
+ next_version_running_writer_identity=$next_version_full20_writer_identity
33
+ next_version_molmo_motion_module=cosmos_framework.configs.base.experiment.sft.omni4d_stage_one_lora_r256_next_version_all_sources_plus_molmo_motion_21_frames_plain_ddp_from_explicit_overlay
34
+ next_version_molmo_motion_windows_sha=3f5503a68154b225f04181e76edd7c6e57dfaddc20d9a754e5adc9a8e190a66d
35
+ next_version_city_molmo_module=cosmos_framework.configs.base.experiment.sft.omni4d_stage_one_lora_r256_next_version_all_sources_plus_cityscapes_and_molmo_motion_21_frames_plain_ddp_from_explicit_overlay
36
+ next_version_city_molmo_windows_sha=73c90b72adc27e1589f97bff64cb35ace5d03acf7fdac43a3e5a5c86055178d5
37
+ next_version_city_molmo_vgg_commit=f6992c4b67441dc9778642a252447a2687856116
38
+ next_version_city_molmo_writer_identity=cosmos3_nano_omni4d_stage_one_lora_r256_plain_ddp_21_frames_temporal_forcing_full20_plus_cityscapes_and_molmo_motion_from_overlay_global32_lr2p5e5
39
+ omega_statistics_path=${COSMOS_OMEGA_STATISTICS_PATH:-}
40
+ omega_statistics_sha256=${COSMOS_OMEGA_STATISTICS_SHA256:-}
41
+ microbatch_profile=${COSMOS_MICROBATCH_PROFILE:-b8_gas2}
42
+ case "$microbatch_profile" in
43
+ b8_gas2)
44
+ expected_batch=8
45
+ expected_accum=2
46
+ expected_windows_sha=4ae8815d1141d7a2215af094232eddde641bb3aff40963896cf2fe1fc00415e3
47
+ ;;
48
+ b16_gas1)
49
+ expected_batch=16
50
+ expected_accum=1
51
+ expected_windows_sha=0b35b990c82717b4d998691e616581573679ca53f30c44dc370577362b31905f
52
+ ;;
53
+ *) echo "COSMOS_MICROBATCH_PROFILE must be b8_gas2 or b16_gas1" >&2; exit 20 ;;
54
+ esac
55
+ if [[ "$experiment_module" == "$next_version_molmo_motion_module" ]]; then
56
+ [[ "$microbatch_profile" == b16_gas1 ]] || {
57
+ echo "the MolmoMotion variant is the B16 profile: set COSMOS_MICROBATCH_PROFILE=b16_gas1" >&2; exit 20
58
+ }
59
+ expected_windows_sha=$next_version_molmo_motion_windows_sha
60
+ fi
61
+ if [[ "$experiment_module" == "$next_version_city_molmo_module" ]]; then
62
+ [[ "$microbatch_profile" == b16_gas1 ]] || {
63
+ echo "the combined City and MolmoMotion recipe is the B16 profile: set COSMOS_MICROBATCH_PROFILE=b16_gas1" >&2; exit 20
64
+ }
65
+ expected_windows_sha=$next_version_city_molmo_windows_sha
66
+ fi
67
+ overlay_sha256=${COSMOS_TRAINABLE_OVERLAY_SHA256:-}
68
+ overlay_update=${COSMOS_TRAINABLE_OVERLAY_UPDATE:-}
69
+ overlay_config_identity=${COSMOS_TRAINABLE_OVERLAY_CONFIG_IDENTITY:-}
70
+ twenty_update_test=${COSMOS_TWENTY_UPDATE_ADMISSION_TEST:-0}
71
+ [[ "$twenty_update_test" == 0 || "$twenty_update_test" == 1 ]] || { echo "COSMOS_TWENTY_UPDATE_ADMISSION_TEST must be 0 or 1" >&2; exit 20; }
72
+ timing_gate_updates=${COSMOS_TIMING_GATE_UPDATES:-0}
73
+ [[ "$timing_gate_updates" =~ ^(0|[1-9][0-9]*)$ ]] || { echo "COSMOS_TIMING_GATE_UPDATES must be 0 or an integer >= 50" >&2; exit 20; }
74
+ [[ "$timing_gate_updates" == 0 || "$timing_gate_updates" -ge 50 ]] || {
75
+ echo "COSMOS_TIMING_GATE_UPDATES must be at least 50 when enabled" >&2; exit 20;
76
+ }
77
+ if [[ "$timing_gate_updates" != 0 && "$experiment_module" != "$next_version_full19_module" \
78
+ && "$experiment_module" != "$next_version_city_molmo_module" ]]; then
79
+ echo "the 50-update timing gate is supported only for the full20 and combined City+Molmo recipes" >&2; exit 20
80
+ fi
81
+ [[ "$timing_gate_updates" == 0 || "$twenty_update_test" == 0 ]] || {
82
+ echo "select either the 20-update admission test or the 50-update timing gate" >&2; exit 20;
83
+ }
84
+ if [[ "$experiment_module" == "$next_version_ddp_module" ]]; then
85
+ warmstart_path=${warmstart_path:-$next_version_ddp_overlay}
86
+ [[ "$warmstart_path" == "$next_version_ddp_overlay" ]] || {
87
+ echo "the DDP 21-frame continuation must use the protected update-8200 overlay: $next_version_ddp_overlay" >&2
88
+ exit 21
89
+ }
90
+ fi
91
+ if [[ "$experiment_module" == "$next_version_full19_module" || "$experiment_module" == "$next_version_molmo_motion_module" \
92
+ || "$experiment_module" == "$next_version_city_molmo_module" ]]; then
93
+ [[ -f "$warmstart_path" ]] || { echo "the 21-frame continuation requires an explicit trainable overlay path" >&2; exit 21; }
94
+ [[ "$overlay_sha256" =~ ^[0-9a-f]{64}$ ]] || { echo "COSMOS_TRAINABLE_OVERLAY_SHA256 must be supplied explicitly" >&2; exit 21; }
95
+ [[ "$overlay_update" =~ ^[1-9][0-9]*$ ]] || { echo "COSMOS_TRAINABLE_OVERLAY_UPDATE must be supplied explicitly" >&2; exit 21; }
96
+ if [[ "$experiment_module" == "$next_version_full19_module" || "$experiment_module" == "$next_version_molmo_motion_module" \
97
+ || "$experiment_module" == "$next_version_city_molmo_module" ]]; then
98
+ [[ "$overlay_config_identity" == "$next_version_full20_start_identity" \
99
+ || "$overlay_config_identity" == "$next_version_running_writer_identity" \
100
+ || ("$experiment_module" == "$next_version_city_molmo_module" && "$overlay_config_identity" == "$next_version_city_molmo_writer_identity") ]] || {
101
+ echo "COSMOS_TRAINABLE_OVERLAY_CONFIG_IDENTITY must be the identity of an overlay of the full-mixture 21-frame run: $next_version_full20_start_identity or $next_version_running_writer_identity" >&2
102
+ exit 21
103
+ }
104
+ else
105
+ [[ "$overlay_config_identity" == "$next_version_full20_start_identity" ]] || {
106
+ echo "COSMOS_TRAINABLE_OVERLAY_CONFIG_IDENTITY must match the reviewed full-mixture 21-frame overlay" >&2
107
+ exit 21
108
+ }
109
+ fi
110
+ fi
111
+ if [[ "$experiment_module" == "$next_version_omega_module" ]]; then
112
+ # The same strict weights-only handoff as the five-stream restart, from a five-stream overlay (the one the restart began
113
+ # from or one the restart wrote); the six Omega tensors are added by the recipe's explicit allowance.
114
+ [[ -f "$warmstart_path" ]] || { echo "the Omega-stream run requires an explicit five-stream trainable overlay path" >&2; exit 21; }
115
+ [[ "$overlay_sha256" =~ ^[0-9a-f]{64}$ ]] || { echo "COSMOS_TRAINABLE_OVERLAY_SHA256 must be supplied explicitly" >&2; exit 21; }
116
+ [[ "$overlay_update" =~ ^[1-9][0-9]*$ ]] || { echo "COSMOS_TRAINABLE_OVERLAY_UPDATE must be supplied explicitly" >&2; exit 21; }
117
+ [[ "$overlay_config_identity" == "$next_version_full20_start_identity" || "$overlay_config_identity" == "$next_version_full20_writer_identity" ]] || {
118
+ echo "COSMOS_TRAINABLE_OVERLAY_CONFIG_IDENTITY must be a reviewed five-stream full-mixture 21-frame overlay identity" >&2
119
+ exit 21
120
+ }
121
+ [[ -f "$omega_statistics_path" ]] || { echo "COSMOS_OMEGA_STATISTICS_PATH must name the pinned Omega feature statistics file" >&2; exit 21; }
122
+ [[ "$omega_statistics_sha256" =~ ^[0-9a-f]{64}$ ]] || { echo "COSMOS_OMEGA_STATISTICS_SHA256 must be supplied explicitly" >&2; exit 21; }
123
+ actual_statistics_sha256=$(sha256sum "$omega_statistics_path" | awk '{print $1}')
124
+ [[ "$actual_statistics_sha256" == "$omega_statistics_sha256" ]] || {
125
+ echo "Omega statistics SHA-256 mismatch: expected=$omega_statistics_sha256 got=$actual_statistics_sha256" >&2; exit 22;
126
+ }
127
+ elif [[ -n "$omega_statistics_path$omega_statistics_sha256" ]]; then
128
+ echo "the Omega statistics belong to the Omega-stream module only" >&2; exit 21
129
+ fi
130
+ default_attempt_root=$(bash "$repo_root/tools/omni4d/cosmos3_nano/resolve_native_formal_attempt_root.sh" \
131
+ "$experiment_module" "$warmstart_path")
132
+ attempt_root=${COSMOS_ATTEMPT_ROOT:-$default_attempt_root}
133
+ slurm_category=train
134
+ if [[ "$twenty_update_test" == 1 ]]; then slurm_category=smoke; fi
135
+ step_log_dir="/scratch/shared/engs2907/logs/vgg_forecast_4d/slurm/$slurm_category/$(basename "$(dirname "$attempt_root")")/$(basename "$attempt_root")"
136
+ python_bin=${COSMOS_PYTHON:-/scratch/shared/engs2907/envs/cosmos3-nano-py313-cu128-001fe183/bin/python}
137
+ compute_python=${COSMOS_COMPUTE_PYTHON:-$python_bin}
138
+ [[ "$python_bin" == /scratch/shared/engs2907/envs/cosmos3-nano-py313-cu128-001fe183/bin/python && "$compute_python" == "$python_bin" ]] || { echo "Cosmos uses one persistent Conda environment" >&2; exit 20; }
139
+ native_stall=${COSMOS_NATIVE_STALL_DIAGNOSTICS:-0}
140
+ [[ "$native_stall" == 0 || "$native_stall" == 1 ]] || { echo "COSMOS_NATIVE_STALL_DIAGNOSTICS must be 0 or 1" >&2; exit 20; }
141
+ if [[ "$twenty_update_test" == 1 && "$experiment_module" != "$next_version_full19_module" \
142
+ && "$experiment_module" != "$next_version_omega_module" \
143
+ && "$experiment_module" != "$next_version_molmo_motion_module" \
144
+ && "$experiment_module" != "$next_version_city_molmo_module" ]]; then
145
+ echo "the bounded admission prefix is supported only for the 21-frame full-mixture recipe" >&2; exit 20
146
+ fi
147
+
148
+ if [[ ! -x "$python_bin" || ! -f "$windows_config" ]]; then
149
+ echo "the pinned Cosmos Python environment and full-mixture source config must exist" >&2
150
+ exit 20
151
+ fi
152
+ if [[ ! "$windows_sha256" =~ ^[0-9a-f]{64}$ ]]; then
153
+ echo "invalid source-config SHA-256" >&2
154
+ exit 21
155
+ fi
156
+ actual_sha256=$(sha256sum "$windows_config" | awk '{print $1}')
157
+ if [[ "$actual_sha256" != "$windows_sha256" ]]; then
158
+ echo "source-config hash mismatch: expected=$windows_sha256 got=$actual_sha256" >&2
159
+ exit 22
160
+ fi
161
+ expected_vgg_commit=b9e2059143b4db90a57a4bac7d5dca1a7ee03d48
162
+ if [[ "$experiment_module" == "$next_version_city_molmo_module" ]]; then
163
+ expected_vgg_commit=$next_version_city_molmo_vgg_commit
164
+ fi
165
+ if [[ "$experiment_module" == "$next_version_ddp_module" || "$experiment_module" == "$next_version_full19_module" \
166
+ || "$experiment_module" == "$next_version_full19_cityscapes_module" || "$experiment_module" == "$next_version_omega_module" \
167
+ || "$experiment_module" == "$next_version_molmo_motion_module" || "$experiment_module" == "$next_version_city_molmo_module" ]]; then
168
+ expected_sources=20
169
+ expected_total_weight=133.0
170
+ expected_molmo_motion_sources=0
171
+ if [[ "$experiment_module" == "$next_version_molmo_motion_module" ]]; then
172
+ expected_sources=21
173
+ expected_total_weight=139.0
174
+ expected_molmo_motion_sources=1
175
+ elif [[ "$experiment_module" == "$next_version_full19_cityscapes_module" ]]; then
176
+ expected_sources=21
177
+ expected_total_weight=137.0
178
+ elif [[ "$experiment_module" == "$next_version_city_molmo_module" ]]; then
179
+ expected_sources=22
180
+ expected_total_weight=143.0
181
+ expected_molmo_motion_sources=1
182
+ fi
183
+ "$python_bin" - "$windows_config" "$expected_batch" "$expected_accum" "$expected_sources" "$expected_total_weight" \
184
+ "$expected_molmo_motion_sources" "$expected_vgg_commit" <<'PY'
185
+
186
+ import sys
187
+ import yaml
188
+
189
+ with open(sys.argv[1], encoding="utf-8") as stream:
190
+ payload = yaml.safe_load(stream)
191
+ window = payload["window_block"]
192
+ checks = {
193
+ "frames": (window["frames"], 21),
194
+ "per-rank batch size": (window["batch_size"], int(sys.argv[2])),
195
+ "gradient accumulation": (window["gradient_accumulation_steps"], int(sys.argv[3])),
196
+ "loader worker count": (window["loader_workers"], 12),
197
+ }
198
+ if payload.get("vgg_commit") != sys.argv[7]:
199
+ raise SystemExit(f"the selected full-mixture YAML must pin the reviewed VGG source {sys.argv[7]}")
200
+ molmo_motion = payload.get("molmo_motion_sources", [])
201
+ if len(molmo_motion) != int(sys.argv[6]):
202
+ raise SystemExit("the selected YAML lists a different number of MolmoMotion sources than the selected recipe")
203
+ if len(window.get("sources", [])) + len(payload.get("robot_sources", [])) + \
204
+ len(payload.get("real_video_sources", [])) + len(molmo_motion) != int(sys.argv[4]):
205
+ raise SystemExit("the selected YAML must retain all approved source stanzas")
206
+ if payload.get("mixture_policy", {}).get("total_weight") != float(sys.argv[5]):
207
+ raise SystemExit("the selected YAML has a different full-mixture source weighting")
208
+ failed = [f"{name}: expected {expected}, got {actual}" for name, (actual, expected) in checks.items()
209
+ if actual != expected]
210
+ if failed:
211
+ raise SystemExit("DDP 21-frame windows config does not match the selected microbatch profile:\n" + "\n".join(failed))
212
+ PY
213
+ [[ "$windows_sha256" == "$expected_windows_sha" ]] || {
214
+ echo "windows YAML SHA-256 does not match the selected microbatch profile" >&2; exit 22;
215
+ }
216
+ fi
217
+ if [[ -e "$attempt_root" ]]; then
218
+ echo "refusing to overwrite existing attempt: $attempt_root" >&2
219
+ exit 23
220
+ fi
221
+ mkdir -p "$step_log_dir"
222
+ allocation_state=$(squeue -h -j "$allocation" -o '%T|%N|%b')
223
+ allocation_state_name=${allocation_state%%|*}
224
+ allocation_rest=${allocation_state#*|}
225
+ allocation_nodes=${allocation_rest%%|*}
226
+ allocation_gres=${allocation_rest#*|}
227
+ if [[ "$allocation_state_name" != RUNNING || "$allocation_nodes" != *"$expected_node"* || "$allocation_gres" != *gpu*2* ]]; then
228
+ echo "approved allocation is not RUNNING with two GPUs on the expected node: $allocation_state" >&2
229
+ exit 24
230
+ fi
231
+ mkdir -p "$(dirname "$attempt_root")"
232
+ if ! mkdir "$attempt_root"; then
233
+ echo "another launcher reserved this attempt concurrently: $attempt_root" >&2
234
+ exit 23
235
+ fi
236
+ git -C "$repo_root" diff --quiet HEAD -- || { echo "training source has tracked edits" >&2; exit 25; }
237
+ source_sha=$(git -C "$repo_root" rev-parse HEAD)
238
+ git -C "$repo_root" archive --format=tar "$source_sha" > "$attempt_root/source.tar"
239
+ printf '%s\n' "$source_sha" > "$attempt_root/source_sha.txt"
240
+ cp "$windows_config" "$attempt_root/source_windows.yaml"
241
+ windows_config="$attempt_root/source_windows.yaml"
242
+ if [[ "$timing_gate_updates" != 0 ]]; then
243
+ cp "$launcher_root/tools/omni4d/cosmos3_nano/sample_training_host_memory.py" \
244
+ "$attempt_root/sample_training_host_memory.py"
245
+ fi
246
+ if [[ "$experiment_module" == "$next_version_omega_module" ]]; then
247
+ cp "$omega_statistics_path" "$attempt_root/omega_feature_statistics.safetensors"
248
+ omega_statistics_path="$attempt_root/omega_feature_statistics.safetensors"
249
+ fi
250
+
251
+ srun --jobid="$allocation" --kill-on-bad-exit=1 --overlap --exact --nodes=1 --ntasks=1 --cpus-per-task=55 --mem=644000M \
252
+ --gres=gpu:2 --gpu-bind=none --cpu-bind=none --chdir=/tmp \
253
+ --output="$step_log_dir/slurm-%j-%s.out" \
254
+ bash -c '
255
+ set -euo pipefail
256
+ ulimit -Sn 65536
257
+ printf "trainer_open_file_limit=%s\n" "$(ulimit -Sn)"
258
+ repo_root=$1
259
+ expected_node=$2
260
+ windows_config=$3
261
+ windows_sha256=$4
262
+ attempt_root=$5
263
+ python_bin=$6
264
+ source_sha=$7
265
+ allocated_gpu_uuids=$8
266
+ experiment_module=$9
267
+ warmstart_path=${10}
268
+ native_stall=${11}
269
+ next_version_ddp_module=${12}
270
+ next_version_full19_module=${13}
271
+ next_version_full19_cityscapes_module=${14}
272
+ microbatch_profile=${15}
273
+ overlay_sha256=${16}
274
+ overlay_update=${17}
275
+ overlay_config_identity=${18}
276
+ twenty_update_test=${19}
277
+ next_version_omega_module=${20}
278
+ omega_statistics_path=${21}
279
+ omega_statistics_sha256=${22}
280
+ next_version_full20_start_identity=${23}
281
+ next_version_full20_writer_identity=${24}
282
+ next_version_molmo_motion_module=${25}
283
+ next_version_city_molmo_module=${26}
284
+ next_version_city_molmo_writer_identity=${27}
285
+ timing_gate_updates=${28}
286
+ case "$microbatch_profile" in
287
+ b8_gas2) expected_batch=8; expected_accum=2 ;;
288
+ b16_gas1) expected_batch=16; expected_accum=1 ;;
289
+ *) echo "invalid microbatch profile in compute step: $microbatch_profile" >&2; exit 20 ;;
290
+ esac
291
+ [[ -x "$python_bin" ]] || { echo "compute-node Python is missing: $python_bin" >&2; exit 20; }
292
+ printf "compute_python=%s\n" "$python_bin"
293
+ [[ "$allocated_gpu_uuids" == GPU-*,GPU-* && "$allocated_gpu_uuids" != *,*,* ]] || { echo "expected exactly two explicit allocated GPU UUIDs" >&2; exit 32; }
294
+ printf "slurm_step_gpus=%s inherited_cuda_visible=%s explicit_gpu_uuids=%s\n" "${SLURM_STEP_GPUS:-unset}" "${CUDA_VISIBLE_DEVICES:-unset}" "$allocated_gpu_uuids"
295
+ export CUDA_VISIBLE_DEVICES="$allocated_gpu_uuids"
296
+ slurm_tasks=${SLURM_NTASKS:?Slurm task count is required}
297
+ [[ "$slurm_tasks" == 1 ]] || { echo "expected one Slurm task owning both GPUs, got $slurm_tasks" >&2; exit 30; }
298
+ [[ "$(hostname -s)" == "$expected_node" ]] || { echo "unexpected compute node $(hostname -s)" >&2; exit 31; }
299
+ [[ -n "${CUDA_VISIBLE_DEVICES:-}" ]] || {
300
+ echo "the single Slurm task must see both allocated GPUs; CUDA_VISIBLE_DEVICES is unset" >&2; exit 32;
301
+ }
302
+ mapfile -t gpu_uuids < <("$python_bin" -c "import torch; n=torch.cuda.device_count(); assert n == 2, n; print(chr(10).join(str(torch.cuda.get_device_properties(i).uuid) for i in range(n)))")
303
+ [[ ${#gpu_uuids[@]} == 2 ]] || { echo "could not enumerate two GPUs from CVD=$CUDA_VISIBLE_DEVICES" >&2; exit 33; }
304
+ uuid0=${gpu_uuids[0]}; uuid1=${gpu_uuids[1]}
305
+ [[ "$uuid0" == GPU-* ]] || uuid0="GPU-$uuid0"
306
+ [[ "$uuid1" == GPU-* ]] || uuid1="GPU-$uuid1"
307
+ [[ "$uuid0" == GPU-* && "$uuid1" == GPU-* && "$uuid0" != "$uuid1" ]] || {
308
+ echo "could not resolve two distinct physical UUIDs from CVD=$CUDA_VISIBLE_DEVICES: $uuid0 $uuid1" >&2; exit 33;
309
+ }
310
+ mkdir -p "$attempt_root/preflight"
311
+ printf "%s\n%s\n" "$uuid0" "$uuid1" > "$attempt_root/preflight/gpu_uuids.txt.tmp"
312
+ mv "$attempt_root/preflight/gpu_uuids.txt.tmp" "$attempt_root/preflight/gpu_uuids.txt"
313
+ for uuid in "$uuid0" "$uuid1"; do
314
+ foreign_pids=$(nvidia-smi --id="$uuid" --query-compute-apps=pid --format=csv,noheader | grep -v -i "^No running processes found$" || true)
315
+ [[ -z "$foreign_pids" ]] || { echo "GPU $uuid is already occupied by: $foreign_pids" >&2; exit 36; }
316
+ done
317
+ actual_sha=$(sha256sum "$windows_config" | awk "{print \$1}")
318
+ [[ "$actual_sha" == "$windows_sha256" ]] || { echo "source-config changed after launch preflight" >&2; exit 37; }
319
+ node_source="/tmp/${USER}_cosmos3_formal_${SLURM_JOB_ID}_${SLURM_STEP_ID}_${source_sha}"
320
+ mkdir "$node_source"
321
+ tar -xf "$attempt_root/source.tar" -C "$node_source"
322
+ printf "%s\n" "$source_sha" > "$node_source/.source_ready"
323
+ [[ "$(cat "$node_source/.source_ready")" == "$source_sha" ]] || { echo "immutable node-local source is not ready" >&2; exit 38; }
324
+ repo_root="$node_source"
325
+ export COSMOS_BOUND_GPU_UUIDS="$uuid0,$uuid1"
326
+ export COSMOS_TRAINING=1 WANDB_MODE=offline WANDB_DIR="$attempt_root/wandb"
327
+ export PYTHONNOUSERSITE=1 HF_HUB_OFFLINE=1 HF_DATASETS_OFFLINE=1 TRANSFORMERS_OFFLINE=1
328
+ export OMP_NUM_THREADS=1 MKL_NUM_THREADS=1 SYN4D_NPZ_CACHE_SIZE=64
329
+ export COSMOS_NODE_LOCAL_COMPILER_CACHE_ROOT="/tmp/${USER}_cosmos3_compiler_${SLURM_JOB_ID}_${SLURM_STEP_ID}"
330
+ export IMAGINAIRE_OUTPUT_ROOT=/scratch/shared/engs2907/logs/vgg_forecast_4d/train
331
+ export PYTHONPATH="$repo_root"
332
+ export COSMOS_VGG_WINDOWS_CONFIG="$windows_config"
333
+ export PATH="$(dirname "$python_bin"):$PATH"
334
+ torchcodec_library_path=$("$python_bin" "$repo_root/tools/omni4d/cosmos3_nano/prepare_torchcodec_runtime.py" --print-library-path)
335
+ export LD_LIBRARY_PATH="$torchcodec_library_path${LD_LIBRARY_PATH:+:$LD_LIBRARY_PATH}"
336
+ printf "torchcodec_ld_library_path=%s\n" "$torchcodec_library_path"
337
+ printf "slurm_tasks=1 host=%s cuda_visible=%s gpu_uuids=%s,%s\n" \
338
+ "$(hostname -s)" "$CUDA_VISIBLE_DEVICES" "$uuid0" "$uuid1"
339
+ cd "$repo_root"
340
+ "$python_bin" tools/omni4d/cosmos3_nano/check_shared_runtime.py --output "$attempt_root/runtime_import_receipt.json"
341
+ warmstart_args=()
342
+ recipe_overrides=()
343
+ if [[ -n "$warmstart_path" ]]; then
344
+ warmstart_args+=("model.trainable_overlay_path=$warmstart_path")
345
+ fi
346
+ if [[ "$experiment_module" == "$next_version_ddp_module" ]]; then
347
+ recipe_overrides+=(
348
+ "dataloader_train.batch_size=8"
349
+ "dataloader_train.num_workers=12"
350
+ "trainer.grad_accum_iter=2"
351
+ )
352
+ overlay_clock=$("$python_bin" -c "
353
+ import json
354
+ import math
355
+ import sys
356
+ from safetensors import safe_open
357
+
358
+ expected = {
359
+ \"format\": \"omni4d_trainable_weights_v1\",
360
+ \"exact_training_resume\": \"false\",
361
+ \"update\": \"8200\",
362
+ \"base_identity\": \"nvidia/Cosmos3-Nano@e59a53c25979a090fa8706c9acc0c254a6e89b92\",
363
+ \"config_identity\": \"cosmos3_nano_omni4d_stage_one_lora_r256_native_layer_masks_fp32_rope_full_checkpoint_retained_forward_parameters_global32\",
364
+ }
365
+ with safe_open(sys.argv[1], framework=\"pt\", device=\"cpu\") as stream:
366
+ metadata = stream.metadata() or {}
367
+ keys = stream.keys()
368
+ tensor_headers = [(stream.get_slice(key).get_dtype(), stream.get_slice(key).get_shape()) for key in keys]
369
+ if any(metadata.get(key) != value for key, value in expected.items()):
370
+ raise ValueError(\"the DDP 21-frame run requires the strict update-8200 native trainable overlay\")
371
+ names = json.loads(metadata.get(\"parameter_names\", \"null\"))
372
+ if not isinstance(names, list) or len(keys) != 530 or len(names) != 530 or set(names) != set(keys):
373
+ raise ValueError(f\"update-8200 overlay must contain the exact 530 trainable tensors, got {len(keys)}\")
374
+ if any(dtype != \"F32\" for dtype, _ in tensor_headers):
375
+ raise ValueError(\"update-8200 trainable overlay tensors must all be FP32\")
376
+ parameters = sum(math.prod(shape) for _, shape in tensor_headers)
377
+ if parameters != 747673344:
378
+ raise ValueError(f\"update-8200 overlay parameter count is {parameters}, expected 747673344\")
379
+ print(8200)
380
+ " "$warmstart_path")
381
+ warmstart_args+=("checkpoint.model_only_overlay_start_iteration=$overlay_clock")
382
+ printf "model_only_start_iteration=%s optimizer_state=fresh exact_training_resume=false\n" "$overlay_clock"
383
+ elif [[ "$experiment_module" == "$next_version_full19_module" || "$experiment_module" == "$next_version_omega_module" \
384
+ || "$experiment_module" == "$next_version_molmo_motion_module" || "$experiment_module" == "$next_version_city_molmo_module" ]]; then
385
+ if [[ "$experiment_module" == "$next_version_omega_module" ]]; then
386
+ [[ "$overlay_config_identity" == "$next_version_full20_start_identity" || "$overlay_config_identity" == "$next_version_full20_writer_identity" ]] || {
387
+ echo "the Omega-stream run starts from a reviewed five-stream overlay identity" >&2; exit 39;
388
+ }
389
+ actual_statistics_sha=$(sha256sum "$omega_statistics_path")
390
+ actual_statistics_sha=${actual_statistics_sha%% *}
391
+ [[ "$actual_statistics_sha" == "$omega_statistics_sha256" ]] || {
392
+ echo "Omega statistics SHA-256 mismatch: expected=$omega_statistics_sha256 got=$actual_statistics_sha" >&2; exit 39;
393
+ }
394
+ recipe_overrides+=(
395
+ "dataloader_train.main_process_collator.omega.statistics=$omega_statistics_path"
396
+ "dataloader_train.main_process_collator.omega.expected_statistics_sha256=$omega_statistics_sha256"
397
+ )
398
+ printf "omega_statistics_sha256=%s trainable_overlay_new_tensors=6 start_overlay_tensors=530\n" "$omega_statistics_sha256"
399
+ fi
400
+ actual_overlay_sha=$(sha256sum "$warmstart_path")
401
+ actual_overlay_sha=${actual_overlay_sha%% *}
402
+ [[ "$actual_overlay_sha" == "$overlay_sha256" ]] || {
403
+ echo "trainable overlay SHA-256 mismatch: expected=$overlay_sha256 got=$actual_overlay_sha" >&2; exit 39;
404
+ }
405
+ overlay_clock=$("$python_bin" -c "
406
+ import json, math, sys
407
+ from safetensors import safe_open
408
+ path, expected_update, expected_identity = sys.argv[1], int(sys.argv[2]), sys.argv[3]
409
+ expected = {
410
+ \"format\": \"omni4d_trainable_weights_v1\",
411
+ \"exact_training_resume\": \"false\",
412
+ \"update\": str(expected_update),
413
+ \"base_identity\": \"nvidia/Cosmos3-Nano@e59a53c25979a090fa8706c9acc0c254a6e89b92\",
414
+ \"config_identity\": expected_identity,
415
+ }
416
+ with safe_open(path, framework=\"pt\", device=\"cpu\") as stream:
417
+ metadata = stream.metadata() or {}
418
+ keys = stream.keys()
419
+ headers = [(stream.get_slice(key).get_dtype(), stream.get_slice(key).get_shape()) for key in keys]
420
+ if any(metadata.get(key) != value for key, value in expected.items()):
421
+ raise ValueError(f\"trainable overlay metadata mismatch for requested update {expected_update}\")
422
+ names = json.loads(metadata.get(\"parameter_names\", \"null\"))
423
+ if not isinstance(names, list) or len(keys) != 530 or len(names) != 530 or set(names) != set(keys):
424
+ raise ValueError(f\"trainable overlay must contain the exact 530 trainable tensors, got {len(keys)}\")
425
+ if any(dtype != \"F32\" for dtype, _ in headers):
426
+ raise ValueError(\"trainable overlay tensors must all be FP32\")
427
+ if sum(math.prod(shape) for _, shape in headers) != 747673344:
428
+ raise ValueError(\"trainable overlay parameter count differs from the native LoRA scope\")
429
+ print(expected_update)
430
+ " "$warmstart_path" "$overlay_update" "$overlay_config_identity")
431
+ warmstart_args+=("checkpoint.model_only_overlay_start_iteration=$overlay_clock")
432
+ warmstart_args+=("model.trainable_overlay_config_identity=$overlay_config_identity")
433
+ recipe_overrides+=(
434
+ "dataloader_train.batch_size=$expected_batch"
435
+ "dataloader_train.num_workers=12"
436
+ "trainer.grad_accum_iter=$expected_accum"
437
+ "dataloader_train.dataset.sample_offset_updates=$overlay_clock"
438
+ "dataloader_train.dataset.batch_size=$expected_batch"
439
+ "dataloader_train.dataset.gradient_accumulation_steps=$expected_accum"
440
+ "optimizer.lr=2.5e-5"
441
+ )
442
+ timing_prefetch_factor=2
443
+ if [[ "$microbatch_profile" == b16_gas1 ]]; then
444
+ recipe_overrides+=("dataloader_train.prefetch_factor=1")
445
+ timing_prefetch_factor=1
446
+ fi
447
+ if [[ "$twenty_update_test" == 1 ]]; then
448
+ test_end_update=$((overlay_clock + 20))
449
+ recipe_overrides+=("trainer.max_iter=$test_end_update" "trainer.callbacks.omni4d_trainable_overlay.save_updates=[$test_end_update]")
450
+ fi
451
+ if [[ "$timing_gate_updates" != 0 ]]; then
452
+ recipe_overrides+=(
453
+ "trainer.callbacks.iter_speed.native_phase_timing.output_dir=$attempt_root/native_phase_timings"
454
+ "trainer.callbacks.iter_speed.native_phase_timing.start_after_updates=1"
455
+ "trainer.callbacks.iter_speed.native_phase_timing.num_updates=$timing_gate_updates"
456
+ )
457
+ printf "timing_observation_updates=%s workers=12 prefetch_factor=%s host_memory_sample_seconds=10 training_continues=true\n" \
458
+ "$timing_gate_updates" "$timing_prefetch_factor"
459
+ fi
460
+ printf "model_only_start_iteration=%s optimizer_state=fresh exact_training_resume=false overlay_sha256=%s profile=%s lr=2.5e-5\n" "$overlay_clock" "$actual_overlay_sha" "$microbatch_profile"
461
+ if [[ "$twenty_update_test" == 1 ]]; then
462
+ printf "bounded_admission_test_updates=20 checkpoint_update=%s\n" "$test_end_update"
463
+ fi
464
+ elif [[ "$experiment_module" == cosmos_framework.configs.base.experiment.sft.omni4d_stage_one_lora_r256_replicate_only_500k \
465
+ || "$experiment_module" == cosmos_framework.configs.base.experiment.sft.omni4d_stage_one_lora_r256_replicate_only_500k_temporal_forcing \
466
+ || "$experiment_module" == cosmos_framework.configs.base.experiment.sft.omni4d_stage_one_lora_r256_plain_ddp_500k \
467
+ || "$experiment_module" == cosmos_framework.configs.base.experiment.sft.omni4d_stage_one_lora_r256_next_version_all_sources_21_frames_temporal_forcing_from_7300 ]]; then
468
+ overlay_clock=$("$python_bin" -c "
469
+ import sys
470
+ from safetensors import safe_open
471
+ with safe_open(sys.argv[1], framework=\"pt\", device=\"cpu\") as stream:
472
+ metadata = stream.metadata()
473
+ if metadata.get(\"format\") != \"omni4d_trainable_weights_v1\" or metadata.get(\"exact_training_resume\") != \"false\":
474
+ raise ValueError(\"500K handoff requires a trainable-only model-weight overlay\")
475
+ update = int(metadata[\"update\"])
476
+ if not 0 < update < 500000:
477
+ raise ValueError(\"overlay update must leave work in the configured 500K horizon\")
478
+ print(update)
479
+ " "$warmstart_path")
480
+ warmstart_args+=("checkpoint.model_only_overlay_start_iteration=$overlay_clock")
481
+ printf "model_only_start_iteration=%s optimizer_state=fresh exact_training_resume=false\n" "$overlay_clock"
482
+ fi
483
+ nvidia-smi --id="$uuid0,$uuid1" --query-gpu=timestamp,uuid,utilization.gpu,memory.used,power.draw --format=csv,noheader -l 1 > "$attempt_root/gpu_utilization.csv" 2> "$attempt_root/gpu_utilization.err" &
484
+ metrics_pid=$!
485
+ nvidia-smi --id="$uuid0,$uuid1" --query-compute-apps=gpu_uuid,pid,used_gpu_memory --format=csv,noheader -l 10 > "$attempt_root/gpu_processes.csv" 2> "$attempt_root/gpu_processes.err" &
486
+ process_metrics_pid=$!
487
+ host_memory_pid=""
488
+ if [[ "$timing_gate_updates" != 0 ]]; then
489
+ "$python_bin" "$attempt_root/sample_training_host_memory.py" \
490
+ --root-pid "$$" --output "$attempt_root/host_memory.csv" --interval-seconds 10 &
491
+ host_memory_pid=$!
492
+ fi
493
+ diagnostics_pid=""
494
+ if [[ "$native_stall" == 1 ]]; then
495
+ native_stall_options="output_dir:$attempt_root/native_stall,source_identity:${source_sha}_windows_${windows_sha256}_overlay_$(basename "$warmstart_path"),signal:SIGUSR2,progress_timeout_seconds:300,max_updates:500000"
496
+ if [[ "$twenty_update_test" == 1 ]]; then
497
+ native_stall_options+=",capture_batch_details_updates:20"
498
+ elif [[ "$timing_gate_updates" != 0 ]]; then
499
+ native_stall_options+=",capture_batch_details_updates:$timing_gate_updates"
500
+ fi
501
+ warmstart_args+=("trainer.callbacks.iter_speed.native_stall_diagnostics={$native_stall_options}")
502
+ (
503
+ while true; do
504
+ if compgen -G "$attempt_root/native_stall/rank_*_pid_*.handler.json" >/dev/null; then
505
+ "$python_bin" tools/omni4d/watch_native_training_stall.py "$attempt_root/native_stall" --send-signal >> "$attempt_root/native_stall_watchdog.log" 2>&1 || true
506
+ fi
507
+ sleep 30
508
+ done
509
+ ) &
510
+ diagnostics_pid=$!
511
+ fi
512
+ trap "kill $metrics_pid $process_metrics_pid ${diagnostics_pid:-} ${host_memory_pid:-} 2>/dev/null || true" EXIT
513
+ "$python_bin" -m torch.distributed.run --standalone --nnodes=1 --nproc-per-node=2 --max-restarts=0 \
514
+ -m cosmos_framework.scripts.train_experiment \
515
+ --config=cosmos_framework/configs/omni4d_native_config.py \
516
+ --experiment-module="$experiment_module" \
517
+ -- \
518
+ experiment="${experiment_module##*.}" \
519
+ job.name="$(basename "$(dirname "$attempt_root")")_$(basename "$attempt_root")" \
520
+ dataloader_train.dataset.windows_config_path="$windows_config" \
521
+ dataloader_train.dataset.expected_windows_config_sha256="$windows_sha256" \
522
+ trainer.callbacks.omni4d_trainable_overlay.output_dir="$attempt_root/trainable_overlays" \
523
+ "${warmstart_args[@]}" "${recipe_overrides[@]}"
524
+ ' bash "$repo_root" "$expected_node" "$windows_config" "$windows_sha256" "$attempt_root" "$python_bin" \
525
+ "$source_sha" "$allocated_gpu_uuids" "$experiment_module" "$warmstart_path" "$native_stall" "$next_version_ddp_module" "$next_version_full19_module" \
526
+ "$next_version_full19_cityscapes_module" "$microbatch_profile" "$overlay_sha256" "$overlay_update" "$overlay_config_identity" "$twenty_update_test" \
527
+ "$next_version_omega_module" "$omega_statistics_path" "$omega_statistics_sha256" "$next_version_full20_start_identity" "$next_version_full20_writer_identity" \
528
+ "$next_version_molmo_motion_module" "$next_version_city_molmo_module" "$next_version_city_molmo_writer_identity" "$timing_gate_updates"