Tao Train Depth Anything V2

Depth Net Mono

Monocular depth estimation using Metric Depth Anything v2 or Relative Depth Anything architectures. Predicts per-pixel depth from single RGB images.

Pretrained checkpoint loading varies by model variant and use case — see the Pretrained checkpoint loading — use case matrix in references/parameters.md.

The mono and stereo skills both invoke the unified TAO depth_net CLI inside the container; the mono/stereo family is selected via model.model_type (full parameter glossary in references/parameters.md).

For TAO Deploy TensorRT actions (gen_trt_engine, TensorRT evaluate, and TensorRT inference), read references/tao-deploy-depth-anything-v2.md first. The deploy spec template lives in this skill's references/spec_template_deploy.yaml.

Train Action Policy

This model is AutoML-enabled at the model layer. Before handling any train-stage request, read references/skill_info.yaml and resolve the run override from either an explicit automl_policy value or the user's workflow request. Treat phrases like "turn off AutoML", "disable AutoML", "no HPO", or "plain training" as automl_policy: off for this run only; otherwise default to auto. When automl_policy: auto, automl_enabled: true, and both schemas/train.schema.json and references/spec_template_train.yaml are packaged, route the train action through tao-skill-bank:tao-run-automl by default with this model's skill_dir. Preserve workflow/application overrides for datasets, specs, output directories, GPU/platform settings, parent checkpoints, and automl_policy. Use direct model training only when automl_policy: off or the packaged train schema/template is missing; in the missing-schema case, report that AutoML is enabled but not runnable for this model until schemas are generated.

Non-train actions such as evaluate, inference, export, and deploy flows stay in this model skill. The per-run automl_policy override does not change model metadata.

Workflow

Prerequisites — data accessibility

Your dataset (RGB images + GT depth files) must be reachable from inside the container:

• SDK runner: place files at the S3 paths the runner resolves (the S3_TRAIN / S3_EVAL placeholders shown in the spec overrides). The runner handles S3 → container-path mounting transparently.
• Direct docker run (e.g. local testing): mount the host dataset root read-only at the same in-container path:

docker run ... -v <host_data_root>:<host_data_root>:ro <container> ...

The same accessibility requirement applies to the <output_dir> written by all actions.

Step 1 — Annotation file

Per-line annotation file referenced by data_sources[*].data_file:

If you already have one, point to it. Otherwise generate via depth_net convert:

depth_net convert -e <convert_spec.yaml>

convert_spec.yaml template:

data_root: <directory whose immediate children are scene/sample folders that contain your image+depth files; convert walks data_root recursively but expects per-scene subdirectories at one level below>
image_dir_pattern: [<substring matching left/RGB image paths>]
depth_dir_pattern: [<substring matching GT depth paths>]
image_extension: ''     # optional .endswith filter, e.g. '.jpg'
depth_extension: ''     # optional, swapped during depth derivation, e.g. '.png'
split_ratio: 0.0        # 0.0/1.0 = test-only; 0.8 = 80/20 train+val

convert walks data_root recursively, selects paths whose path-string contains all substrings in image_dir_pattern (AND-filter), then derives the depth path by replacing image_dir_pattern[0] with depth_dir_pattern[0] and image_extension with depth_extension. Inspect your dataset's directory layout and identify the substring distinguishing RGB images from depth files (e.g. rgb_ vs sync_depth_).

data_root must point at the parent that contains the per-scene subdirectories (e.g. for NYU eval, use /data/nyu_v2/eval/test, not /data/nyu_v2/eval/test/bathroom — the latter limits the walk to a single scene). Always include the leading dot in image_extension / depth_extension (e.g. '.jpg' not 'jpg'); the substring swap is form-sensitive and a mismatch silently corrupts derived paths.

Step 2 — Pair model_type and dataset_name based on your data

Default — generic class for each task:

Dataset-specific class — switch when the data needs preprocessing the generic class does not perform:

Using a generic class on data that requires unit conversion (e.g. raw NYU uint16 PNGs) results in an empty valid mask and silent train_loss = NaN. Match the class to your data's encoding.

Step 3 — Write spec yaml from the spec overrides

Copy the action block from references/spec-overrides.md. Replace:

• model.model_type from Step 2
• dataset.<...>.data_sources[*].dataset_name from Step 2
• data_sources[*].data_file with the path from Step 1 (S3 path under SDK runner, host path for direct docker)
• For metric finetune: additionally apply the Metric Variant Finetuning Recipe in references/finetuning.md.

For mono training set train.precision: fp32 (recommended) or bf16 (Ampere SM80+, alternative).

Step 4 — Run

docker run --gpus 'device=0' --shm-size 16G --ipc=host \
  --user $(id -u):$(id -g) \
  -v <data_root>:<data_root>:ro \
  -v <output_dir>:<output_dir> \
  <container> \
  depth_net <action> -e <spec.yaml>

Without --user $(id -u):$(id -g) the container writes outputs as nobody:nogroup, blocking host-side cleanup and retry.

Step 5 — Verify

• Container exit code 0
• status.json kpi block populated
• For train: inspect per-step train_loss directly — the entrypoint reports Execution status: PASS even when train_loss = NaN (see the Sanity-run PASS criteria in references/finetuning.md)
• For evaluate / inference: artifacts under results_dir

For TAO Deploy TensorRT actions (gen_trt_engine, TensorRT evaluate, and TensorRT inference), read references/tao-deploy-depth-anything-v2.md first. Deploy spec templates live in this skill's references/ folder with the spec_template_deploy_*.yaml prefix.

Training Requirements

• Valid dataset_name values for mono data_sources (case-insensitive): ThreeDVLM, FSD, NvCLIP, IssacStereo, Crestereo, Middlebury, NYUDV2, NYUDV2Relative, RelativeMonoDataset, MetricMonoDataset. NYUDV2 carries metric depth GT (meters) — pair with MetricDepthAnything; NYUDV2Relative is the same data with relative-depth conventions — pair with RelativeDepthAnything.
• Monitoring metric: val/loss

Per-Action Dataset Requirements

Spec Overrides

Data source overrides are mandatory for every action — construct the data source paths from the Per-Action Dataset Requirements table above and include them in spec_overrides; each data_sources entry is a dict with the two mandatory fields data_file and dataset_name. See references/spec-overrides.md for the full per-action train / evaluate / export / inference / quantize override blocks and the precision recommendations.

Eval Dataset

Optional. Val dataset configured via dataset.val_dataset.data_sources (each entry needs data_file and dataset_name).

Important Parameters

Full parameter glossary (model.*, train.*, dataset.*, export.*, inference.* fields with options, defaults, and sources) plus the Pretrained checkpoint loading — use case matrix live in references/parameters.md. Key starting points: model.model_type (default MetricDepthAnything), model.encoder (default vitl), train.optim.lr (default 1e-4, AdamW), train.precision (fp32 recommended), dataset.{train,val,test,infer}_dataset.augmentation.crop_size (default [518, 518]).

Finetuning Recipes

Relative and Metric variant finetuning recipes — including required spec keys, the metric dataset.{normalize_depth, min_depth, max_depth} block required in both train AND export specs, trainer-enforced defaults (clip_grad_norm: 0.1, warmup_steps: 20, weight_decay: 1e-4), sanity-run overrides, and the Sanity-run PASS criteria for catching silent train_loss = NaN — are in references/finetuning.md. Both recipes use train.optim.lr: 5e-6 with LambdaLR (the AdamW default 1e-4 is too aggressive when finetuning from a converged/pretrained backbone).

Multi-GPU / Multi-Node

Launch method: Lightning-managed (single python process, Lightning spawns workers).

• ddp with activation checkpointing: find_unused_parameters=False
• ddp without: find_unused_parameters=True
• fsdp forces precision to FP16

Multi-node env vars (set by orchestrator): WORLD_SIZE, NODE_RANK, MASTER_ADDR, MASTER_PORT, NUM_GPU_PER_NODE.

Export / TRT Defaults

• TRT data types: FP32, BF16 (Ampere SM80+). FP16 is not supported for the ViT-L mono backbone.
• Recommended TRT precision: bf16. Use fp32 if BF16 hardware is unavailable.

Full TAO Deploy reference: tao-deploy-depth-anything-v2.

Hardware

Minimum 1 GPU(s), recommended 2 GPU(s). 24GB+ VRAM per GPU. ViT-Large encoder is memory intensive. Use fp32 (recommended) or bf16 (Ampere SM80+, alternative) for training. Activation checkpointing is available for larger inputs.

Error Patterns

Common failure signatures and fixes — depth range mismatch, missing pretrained weights, Key 'encoder' not in 'MonoBackBone', missing dataset_name, depth_net_mono: not found, metric variant hyperparameter sourcing, and the export refuse-to-overwrite ONNX error — are documented in references/troubleshooting.md.

Spec Param / Parent Model Inference

Model-specific inference mappings (the full depth_net_mono.config.json per-action spec-field → inference-function table, plus parent_model / parent_job_id resolution guidance) are in references/spec-param-inference.md. These mappings belong in MD, not in config.json; generated runners should read that reference and apply the mappings with SDK helpers before create_job().