Add immutable checkpoint backup cosmos3_nano_stage_one_lora_r256_replicate_only_500k_step_003000_20261104 run_execution_metadata

cosmos3_nano_stage_one_lora_r256_replicate_only_500k_20261103/attempt01/step_003000/backup_context.json

@@ -0,0 +1,51 @@
+{
+  "all_finite": true,
+  "base_model": "nvidia/Cosmos3-Nano",
+  "base_revision": "e59a53c25979a090fa8706c9acc0c254a6e89b92",
+  "checkpoint_kind": "trainable_parameters_only_safetensors",
+  "checkpoint_sha256": "32cf0f8ffffa0a9ba2414acff6e3cadc6fc5dd70f2816294c2d0fccc1536c32a",
+  "checkpoint_source": "/scratch/shared/engs2907/logs/vgg_forecast_4d/train/cosmos3_nano_five_stream_stage_one_lora_r256_replicate_only_500k_20261103/attempt01/trainable_overlays/step_003000.safetensors",
+  "config_identity": "cosmos3_nano_omni4d_stage_one_lora_r256_native_layer_masks_fp32_rope_full_checkpoint_retained_forward_parameters_global32",
+  "dtype": "float32",
+  "exclusions": [
+    "pretrained base weights",
+    "optimizer state",
+    "full model checkpoint"
+  ],
+  "file_bytes_expected": 2990807184,
+  "lineage": {
+    "exact_training_resume": false,
+    "optimizer_state_included": false,
+    "parent_checkpoint_sha256": "ce44be10b68b39c0c0cd2a6ee9c37af52195d66ebb7e9aff04a787ba6df49e3e",
+    "parent_update": 2000,
+    "pretrained_base_included": false
+  },
+  "model": "Cosmos3-Nano Omni4D stage one",
+  "parameter_count": 747673344,
+  "run_id": "cosmos3_nano_stage_one_lora_r256_replicate_only_500k_step_003000_20261104",
+  "schema": "cosmos_trainable_overlay_backup_context_v1",
+  "source_commit": "8279c4a51ca8c42d21c4efcc58804de8b73a79e2",
+  "source_run": "/scratch/shared/engs2907/logs/vgg_forecast_4d/train/cosmos3_nano_five_stream_stage_one_lora_r256_replicate_only_500k_20261103/attempt01",
+  "source_windows_config": "source_windows.yaml",
+  "source_windows_sha256": "ae49ec6d723283ac0429363143cc6e7cebba2e79bdaf8a38c7540f880531dffd",
+  "tensor_count": 530,
+  "training_log": "/scratch/shared/engs2907/logs/vgg_forecast_4d/slurm/train/cosmos3_nano_five_stream_stage_one_lora_r256_replicate_only_500k_20261103/attempt01/slurm-71575-202.out",
+  "training_parameters": {
+    "global_batch": 32,
+    "gradient_accumulation": 1,
+    "lora_alpha": 256,
+    "lora_rank": 256,
+    "per_rank_batch": 16,
+    "streams": [
+      "video",
+      "pointmap",
+      "camera",
+      "dino",
+      "track_video"
+    ],
+    "training_frames": 9,
+    "world_size": 2
+  },
+  "training_step": "71575.202",
+  "training_update": 3000
+}