update ai-hub to first optimize model for Workbench

Remove old examples

README.md

@@ -65,6 +65,18 @@ sagemaker:
     entry_point: null       # Optional: script inside source_dir
     source_dir: null        # Optional: local dir packaged and uploaded automatically
     hyperparameters: {}
+
+aihub:
+  device:
+    name: Samsung Galaxy S25 (Family)
+  target_runtime: tflite
+  input_specs: {}           # Required before running qc-cli ai-hub commands
+  job_name: null            # Optional prefix for AI Hub Workbench jobs
+  model_name: null          # Optional name for uploaded local ONNX models
+  compile_options: null
+  profile_options: null
+  quantize_options: null
+  output_dir: build/qai-hub
 ```
 
 `qc-cli init` generates the `infra.stack_name` and `s3.bucket` namespace once and writes it to `config.yaml`. Keep these values stable for a deployment; changing them points the CLI at different infrastructure.
@@ -78,9 +90,13 @@ To provision an MLflow tracking server, set:
 ```yaml
 mlflow:
   mode: create
-  tracking_server_name: your-tracking-server-name
+  experiment_name: qc-cli-training
+  registered_model_name: qc-cli-model
+  register_trained_models: true
 ```
 
+In `create` mode, the CLI manages the tracking server name from `infra.stack_name`; you do not need to set `tracking_server_name`.
+
 To use an existing MLflow tracking server, set:
 
 ```yaml
@@ -89,6 +105,16 @@ mlflow:
   tracking_server_name: your-tracking-server-name
 ```
 
+When MLflow is enabled, `train start` creates an MLflow run for the SageMaker job. `train status` finalizes that run once the job reaches a terminal state and registers completed model artifacts as experiment model versions using the `experiment-latest` MLflow alias. An experiment version is an immutable trained-source artifact; it records that training produced a model, not that the model is better than earlier versions or ready for release.
+
+To open the managed SageMaker MLflow UI, request a fresh presigned URL:
+
+```bash
+qc-cli mlflow open --config config.yaml
+```
+
+This opens a browser to a fresh presigned URL. It works for `mode: create` and for `mode: existing` when the existing server is managed by Amazon SageMaker. In `create` mode, the command uses the CLI-managed tracking server name. In `existing` mode, it uses `mlflow.tracking_server_name`. If the existing MLflow server is external to SageMaker, open it with that server's own URL instead.
+
 ## Commands
 
 ### `init`
@@ -99,6 +125,12 @@ qc-cli init --output <path>  Write config to a custom path
 qc-cli init --force          Overwrite an existing config file
 ```
 
+### `mlflow`
+
+```
+qc-cli mlflow open  Open a presigned MLflow UI URL in a browser
+```
+
 ### `infra`
 
 ```
@@ -140,6 +172,80 @@ qc-cli train list --limit 3     Show a custom number of recent jobs
 
 The expected output artifact is SageMaker’s `model.tar.gz`, normally containing the trained model file your container writes to `/opt/ml/model`.
 
+### `ai-hub`
+
+```
+qc-cli ai-hub upload <calibration.npz|calibration-dir> <inputs.npz|inputs.npy>
+qc-cli ai-hub upload <calibration> <inputs> --from-step validate
+qc-cli ai-hub optimize [--onnx-path PATH] [--model-s3-uri URI] [--from-job NAME]
+qc-cli ai-hub quantize <calibration.npz|calibration-dir> [--model-id ID] [--onnx-path PATH] [--model-s3-uri URI] [--from-job NAME]
+qc-cli ai-hub compile [--model-id ID] [--onnx-path PATH] [--model-s3-uri URI] [--from-job NAME]
+qc-cli ai-hub validate <inputs.npz|inputs.npy> [--model-id ID] [--input-name NAME]
+qc-cli ai-hub profile [--model-id ID]
+qc-cli ai-hub download [--model-id ID] [--output PATH]
+```
+
+`ai-hub upload` optimizes to ONNX, quantizes, validates, and profiles. When `aihub.target_runtime` is not `onnx`, it
+also compiles the quantized model to that deployment runtime. The initial ONNX optimization gives external models
+Workbench provenance and applies compiler optimization passes before quantization.
+
+Resume behavior:
+
+```text
+--from-step optimize  Run optimize, quantize, optional final compile, validate, and profile.
+--from-step quantize  Quantize the last optimized ONNX, then optionally compile, validate, and profile.
+--from-step compile   Skip optimize and quantize; finalize the last quantized model for the target runtime.
+--from-step validate  Skip optimize, quantize, and compile; validate the last compiled model.
+--from-step profile   Skip optimize, quantize, compile, and validate; profile the last compiled model.
+```
+
+When a step runs in the current command, `upload` passes its returned model ID directly to the next step. When a step is skipped, the next step resolves the needed model ID from `.qc-cli.json`. This avoids re-running earlier AI Hub jobs when you only need to continue from a later step.
+
+`ai-hub optimize` compiles an external model with `--target_runtime onnx`. `ai-hub quantize` uses an explicit
+`--model-id`, the last optimized ONNX model, or an explicit/local model source in that order. `ai-hub compile` resolves
+model sources in this order: `--model-id`, explicit source options, last quantized model, then the last training job.
+For `target_runtime: onnx`, upload treats the quantized ONNX as the final model and skips a redundant second compile.
+`ai-hub download` remains separate because downloading is outside the Workbench processing loop.
+
+AI Hub authentication currently uses the local `qai-hub` SDK configuration. A planned follow-up is to support AWS Systems Manager Parameter Store `SecureString` for team-managed tokens, where `config.yaml` stores only a parameter name such as `/qc-cli/aihub/token`, AWS KMS encrypts the token at rest, and the CLI retrieves it at runtime with `ssm:GetParameter` plus `kms:Decrypt` permissions.
+
+## Model lifecycle
+
+The CLI uses neutral experiment naming for trained artifacts and reserves release terminology for an explicit promotion step.
+
+Current behavior:
+
+1. `qc-cli train start` submits a SageMaker training job.
+2. `qc-cli train status` finalizes the MLflow run after the job reaches a terminal state.
+3. If the job completed and `mlflow.register_trained_models` is enabled, the SageMaker `model.tar.gz` is registered as a new MLflow model version with:
+   - `qc_cli.stage=experiment`
+   - `qc_cli.artifact_kind=trained_source`
+   - `qc_cli.source=sagemaker`
+4. The MLflow alias `experiment-latest` points at the most recently registered experiment version.
+5. AI Hub upload commands create deployable derived artifacts from a trained-source experiment or local ONNX model.
+
+Future release aliases such as `v1` or `production` can point at a selected deployable artifact.
+
+Example future metadata:
+
+```text
+qc-cli-model version 12
+qc_cli.stage=experiment
+qc_cli.artifact_kind=trained_source
+qc_cli.source=sagemaker
+
+qc-cli-model-aihub version 3
+qc_cli.stage=ai_hub_compiled
+qc_cli.artifact_kind=deployable
+qc_cli.parent_registered_model_name=qc-cli-model
+qc_cli.parent_model_version=12
+qc_cli.runtime=tflite
+qc_cli.quantization=int8
+qc_cli.target_device=Samsung Galaxy S25
+```
+
+In that flow, `experiment-latest` remains a training convenience alias. Release selection is a separate promotion decision based on the derived artifact, not on the experiment name.
+
 ## AWS permissions required
 
 The IAM user or role running the CLI needs:

examples/meter-detection/README.md

@@ -0,0 +1,266 @@
+# YOLO26 Electric Meter Detection Example
+
+This example trains a YOLO26 object detection model on the Roboflow Universe electric meter dataset using the existing `qc-cli` SageMaker training flow.
+
+The workflow is intentionally command driven. Run each step yourself so you can inspect the dataset, update `config.yaml`, and decide when to submit the SageMaker job.
+
+Dataset:
+
+```text
+https://universe.roboflow.com/kemals-workspace-kbc8l/electric-meter-detection-o4tfi/dataset/1
+```
+
+## Prerequisites
+
+- Install or sync the project dependencies: `uv sync`
+- The virtual environment is activated.
+- AWS credentials configured for the profile in `config.yaml`
+- Infrastructure already deployed with `qc-cli infra setup`
+
+## 1. Download The Dataset
+
+Register or sign in to Roboflow, then open the dataset page:
+
+```text
+https://universe.roboflow.com/kemals-workspace-kbc8l/electric-meter-detection-o4tfi/dataset/1
+```
+
+Download the dataset in YOLOv26 format from the Roboflow UI, then extract the downloaded archive into:
+
+```text
+examples/meter-detection/data/electric-meter-detection
+```
+
+The `data.yaml` file should be directly under that folder:
+
+```text
+examples/meter-detection/data/electric-meter-detection/data.yaml
+```
+
+Do not move `data.yaml` into the `train/` split folder.
+
+After extracting, confirm the dataset has a YOLO data file and image splits:
+
+```bash
+find examples/meter-detection/data/electric-meter-detection -maxdepth 2 -type d | sort
+find examples/meter-detection/data/electric-meter-detection -name data.yaml -print
+```
+
+Open `examples/meter-detection/data/electric-meter-detection/data.yaml` and make sure the split paths are relative to that folder:
+
+```yaml
+path: .
+train: train/images
+val: valid/images
+test: test/images
+```
+
+If your downloaded dataset does not include a `test/` folder, remove the `test:` line.
+
+The expected layout is similar to:
+
+```text
+examples/meter-detection/data/electric-meter-detection/
+  data.yaml
+  train/
+  valid/
+  test/
+```
+
+## 2. Configure SageMaker Training
+
+Update `config.yaml` so the training section points at this example's source directory:
+
+```yaml
+sagemaker:
+  training:
+    image_uri: 763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training:2.6-cpu-py312-ubuntu22.04-sagemaker-v1
+    instance_type: ml.g4dn.xlarge
+    instance_count: 1
+    source_dir: examples/meter-detection/source
+    entry_point: train.py
+    hyperparameters:
+      model: yolo26n.pt
+      epochs: 25
+      imgsz: 640
+      batch: 16
+      workers: 2
+```
+
+Use `yolo26n.pt` for a lightweight first YOLO26 run. If those weights are unavailable in the installed Ultralytics package, use `yolo11n.pt` as the established fallback:
+
+```yaml
+      model: yolo11n.pt
+```
+
+The `source/requirements.txt` file is installed by the SageMaker PyTorch container before running `train.py`.
+
+For a CPU smoke test, use a CPU instance and reduce the workload:
+
+```yaml
+sagemaker:
+  training:
+    image_uri: 763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training:2.6-cpu-py312-ubuntu22.04-sagemaker-v1
+    instance_type: ml.m4.xlarge
+    instance_count: 1
+    source_dir: examples/meter-detection/source
+    entry_point: train.py
+    hyperparameters:
+      model: yolo26n.pt
+      epochs: 1
+      imgsz: 320
+      batch: 4
+      workers: 2
+```
+
+## 3. Check Infrastructure
+
+Confirm the CLI can see the configured SageMaker role and S3 bucket:
+
+```bash
+qc-cli infra status
+```
+
+## 4. Upload The Dataset
+
+Upload the downloaded Roboflow dataset to the `s3.data_prefix` configured in `config.yaml`:
+
+```bash
+qc-cli upload examples/meter-detection/data/electric-meter-detection
+```
+
+Directory uploads preserve paths relative to the uploaded directory, so SageMaker receives the dataset root with `data.yaml` plus the split directories.
+
+In SageMaker, this uploaded dataset root is mounted at `/opt/ml/input/data/train`. That `train` path is the SageMaker channel name, not the YOLO `train/` split folder.
+
+## 5. Start Training
+
+Submit the SageMaker training job:
+
+```bash
+qc-cli train start
+```
+
+The command prints the submitted SageMaker job name. Check progress with:
+
+```bash
+qc-cli train status
+```
+
+Or pass the job name explicitly:
+
+```bash
+qc-cli train status qc-cli-YYYYMMDD-HHMMSS
+```
+
+## SageMaker Outputs
+
+When the job completes, SageMaker packages the files written under `/opt/ml/model` into `model.tar.gz`.
+
+This example writes:
+
+```text
+best.pt
+model.onnx
+metrics.json
+```
+
+The archive is stored under the configured `s3.model_prefix`.
+
+## 6. Configure Qualcomm AI Hub
+
+Authenticate with Qualcomm AI Hub:
+
+```bash
+qai-hub configure --api_token
+```
+
+Add AI Hub settings to `config.yaml`. The input name and image size must match the ONNX model exported by this example:
+
+```yaml
+aihub:
+  device:
+    name: Dragonwing IQ-9075 EVK
+  target_runtime: onnx
+  input_specs:
+    images: [[1, 3, 640, 640], float32]
+  job_name: meter-detection
+  model_name: meter-detection
+  output_dir: build/qai-hub/meter-detection
+```
+
+The ONNX graph is the source of truth. The export normally uses the same value as `sagemaker.training.hyperparameters.imgsz`, but changing `config.yaml` after training does not resize an existing model. For example, a model exported with `imgsz: 320` requires `images: [[1, 3, 320, 320], float32]`.
+
+## 7. Prepare AI Hub Inputs
+
+Generate calibration samples and a validation input from the downloaded dataset:
+
+```bash
+uv run python examples/meter-detection/prepare_aihub_inputs.py --image-size 640
+```
+
+This writes:
+
+```text
+examples/meter-detection/data/aihub_calibration/*.npy
+examples/meter-detection/data/inputs.npz
+```
+
+The script applies the preprocessing expected by the exported YOLO model: aspect-ratio-preserving letterboxing, RGB channel order, channel-first layout, and pixel values normalized to `[0, 1]`.
+
+## 8. Upload To Qualcomm AI Hub
+
+Use the SageMaker job name printed by `qc-cli train start`:
+
+```bash
+qc-cli ai-hub upload \
+  examples/meter-detection/data/aihub_calibration \
+  examples/meter-detection/data/inputs.npz \
+  --from-job qc-cli-YYYYMMDD-HHMMSS
+```
+
+The command downloads the job's `model.tar.gz`, finds `model.onnx`, and runs the following AI Hub workflow:
+
+1. Compile the external ONNX to a Workbench-optimized ONNX model.
+2. Quantize the optimized ONNX model.
+3. Compile the quantized model when the configured deployment runtime is not `onnx`.
+4. Validate and profile the final model.
+
+The training example sanitizes the Ultralytics ONNX export before saving `model.onnx`. This removes graph input or output names, such as `output0`, that are duplicated in the ONNX `value_info` metadata and rejected by AI Hub.
+
+For a model already downloaded by a failed upload attempt, sanitize the extracted ONNX file and retry using the local model. Replace the job name in both paths:
+
+```bash
+uv run --with onnx python examples/meter-detection/source/sanitize_onnx.py \
+  build/qai-hub/meter-detection/qc-cli-YYYYMMDD-HHMMSS/source/extracted/model.onnx \
+  --output build/qai-hub/meter-detection/model.aihub.onnx
+
+qc-cli ai-hub upload \
+  examples/meter-detection/data/aihub_calibration \
+  examples/meter-detection/data/inputs.npz \
+  --onnx-path build/qai-hub/meter-detection/model.aihub.onnx
+```
+
+Download the compiled artifact after the workflow completes:
+
+```bash
+qc-cli ai-hub download --output build/qai-hub/meter-detection/model.tflite
+```
+
+## Training Hyperparameters
+
+Values under `sagemaker.training.hyperparameters` are passed to `source/train.py` as command-line arguments.
+
+| Name | Type | Default | Description |
+|---|---:|---:|---|
+| `model` | string | `yolo26n.pt` | Ultralytics model weights or model YAML. |
+| `epochs` | int | `25` | Number of training epochs. |
+| `imgsz` | int | `640` | Square training image size. |
+| `batch` | int | `16` | Images per training batch. |
+| `workers` | int | `2` | DataLoader worker count. |
+| `patience` | int | `20` | Early stopping patience. |
+| `device` | string | auto | Optional Ultralytics device value such as `0` or `cpu`. |
+| `data-yaml` | string | auto | Optional path to `data.yaml`; normally discovered from the uploaded dataset root. |
+| `dataset-dir` | string | `SM_CHANNEL_TRAIN` | Uploaded dataset root mounted by SageMaker. |
+
+Do not set `dataset-dir` or `model-dir` in normal SageMaker runs. SageMaker sets those automatically through `SM_CHANNEL_TRAIN` and `SM_MODEL_DIR`.

examples/meter-detection/prepare_aihub_inputs.py

@@ -0,0 +1,92 @@
+#!/usr/bin/env python3
+"""Prepare Qualcomm AI Hub calibration and validation inputs for the meter detector."""
+
+from __future__ import annotations
+
+import argparse
+from pathlib import Path
+
+import numpy as np
+from PIL import Image
+
+IMAGE_EXTENSIONS = {".jpg", ".jpeg", ".png"}
+
+
+def parse_args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument(
+        "--dataset-dir",
+        type=Path,
+        default=Path("examples/meter-detection/data/electric-meter-detection"),
+        help="Root of the extracted Roboflow dataset.",
+    )
+    parser.add_argument(
+        "--calibration-dir",
+        type=Path,
+        default=Path("examples/meter-detection/data/aihub_calibration"),
+        help="Directory where .npy calibration samples will be written.",
+    )
+    parser.add_argument(
+        "--input-file",
+        type=Path,
+        default=Path("examples/meter-detection/data/inputs.npz"),
+        help="Validation .npz input file for qc-cli ai-hub validate.",
+    )
+    parser.add_argument("--input-name", default="images", help="ONNX input name.")
+    parser.add_argument("--image-size", type=int, default=640, help="Square image size used for ONNX export.")
+    parser.add_argument("--samples", type=int, default=16, help="Number of calibration samples to write.")
+    return parser.parse_args()
+
+
+def preprocess_image(path: Path, image_size: int) -> np.ndarray:
+    """Apply Ultralytics-style letterboxing and produce an NCHW float32 tensor."""
+    with Image.open(path) as source:
+        image = source.convert("RGB")
+
+    scale = min(image_size / image.width, image_size / image.height)
+    resized_width = round(image.width * scale)
+    resized_height = round(image.height * scale)
+    image = image.resize((resized_width, resized_height), Image.Resampling.BILINEAR)
+
+    canvas = Image.new("RGB", (image_size, image_size), (114, 114, 114))
+    left = round((image_size - resized_width) / 2 - 0.1)
+    top = round((image_size - resized_height) / 2 - 0.1)
+    canvas.paste(image, (left, top))
+
+    array = np.asarray(canvas, dtype=np.float32) / 255.0
+    return np.transpose(array, (2, 0, 1))[None, ...].astype(np.float32)
+
+
+def main() -> None:
+    args = parse_args()
+    if args.image_size < 1:
+        raise SystemExit("--image-size must be at least 1")
+    if args.samples < 1:
+        raise SystemExit("--samples must be at least 1")
+
+    images = sorted(
+        path
+        for path in args.dataset_dir.rglob("*")
+        if path.is_file() and path.suffix.lower() in IMAGE_EXTENSIONS and path.parent.name == "images"
+    )
+    if not images:
+        raise SystemExit(f"No images found under {args.dataset_dir}")
+
+    args.calibration_dir.mkdir(parents=True, exist_ok=True)
+    args.input_file.parent.mkdir(parents=True, exist_ok=True)
+    for stale_sample in args.calibration_dir.glob("sample_*.npy"):
+        stale_sample.unlink()
+
+    prepared: list[np.ndarray] = []
+    for index, image_path in enumerate(images[: args.samples]):
+        sample = preprocess_image(image_path, args.image_size)
+        np.save(args.calibration_dir / f"sample_{index:03d}.npy", sample)
+        prepared.append(sample)
+
+    np.savez(args.input_file, **{args.input_name: prepared[0]})  # pyright: ignore[reportArgumentType]
+    print(f"Wrote {len(prepared)} calibration samples to {args.calibration_dir}")
+    print(f"Wrote validation input to {args.input_file}")
+
+
+if __name__ == "__main__":
+    main()

examples/meter-detection/source/requirements.txt

@@ -0,0 +1,3 @@
+ultralytics>=8.3.0
+pyyaml>=6.0.3
+onnx>=1.16.0

examples/meter-detection/source/sanitize_onnx.py

@@ -0,0 +1,38 @@
+#!/usr/bin/env python3
+"""Remove ONNX value_info entries that duplicate graph inputs or outputs."""
+
+from __future__ import annotations
+
+import argparse
+from pathlib import Path
+
+import onnx  # type: ignore[reportMissingImports]
+
+
+def sanitize_onnx(path: Path, output_path: Path | None = None) -> Path:
+    model = onnx.load(path)
+    io_names = {value.name for value in (*model.graph.input, *model.graph.output)}
+    retained_value_info = [value for value in model.graph.value_info if value.name not in io_names]
+
+    destination = output_path or path
+    if len(retained_value_info) != len(model.graph.value_info):
+        del model.graph.value_info[:]
+        model.graph.value_info.extend(retained_value_info)
+
+    destination.parent.mkdir(parents=True, exist_ok=True)
+    onnx.save(model, destination)
+    return destination
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("onnx_path", type=Path)
+    parser.add_argument("--output", type=Path)
+    args = parser.parse_args()
+
+    written = sanitize_onnx(args.onnx_path, args.output)
+    print(f"Saved sanitized ONNX model to {written}")
+
+
+if __name__ == "__main__":
+    main()

examples/meter-detection/source/train.py

@@ -0,0 +1,126 @@
+#!/usr/bin/env python3
+"""SageMaker entry point for YOLO electric meter detection training."""
+
+from __future__ import annotations
+
+import argparse
+import json
+import os
+import shutil
+from pathlib import Path
+from typing import Any
+
+import yaml
+from sanitize_onnx import sanitize_onnx
+from ultralytics import YOLO  # type: ignore[reportMissingImports]
+
+
+def parse_args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--model", default="yolo26n.pt")
+    parser.add_argument("--epochs", type=int, default=25)
+    parser.add_argument("--imgsz", type=int, default=640)
+    parser.add_argument("--batch", type=int, default=16)
+    parser.add_argument("--workers", type=int, default=2)
+    parser.add_argument("--patience", type=int, default=20)
+    parser.add_argument("--device", default=None)
+    parser.add_argument("--data-yaml", default=None)
+    parser.add_argument("--dataset-dir", default=os.environ.get("SM_CHANNEL_TRAIN", "/opt/ml/input/data/train"))
+    parser.add_argument("--train-dir", dest="dataset_dir", help=argparse.SUPPRESS)
+    parser.add_argument("--model-dir", default=os.environ.get("SM_MODEL_DIR", "/opt/ml/model"))
+    return parser.parse_args()
+
+
+def find_data_yaml(dataset_dir: Path, explicit_path: str | None) -> Path:
+    if explicit_path:
+        data_yaml = Path(explicit_path)
+        if data_yaml.is_file():
+            return data_yaml
+        raise FileNotFoundError(f"Configured data.yaml does not exist: {data_yaml}")
+
+    matches = sorted(dataset_dir.rglob("data.yaml"))
+    if not matches:
+        raise FileNotFoundError(f"Could not find data.yaml under {dataset_dir}")
+    if len(matches) > 1:
+        print(f"Found multiple data.yaml files; using {matches[0]}")
+    return matches[0]
+
+
+def prepare_data_yaml(data_yaml: Path) -> Path:
+    """Write a SageMaker-local data file rooted at the uploaded dataset."""
+    dataset_root = data_yaml.parent
+    data = yaml.safe_load(data_yaml.read_text(encoding="utf-8"))
+    if not isinstance(data, dict):
+        raise ValueError(f"Expected a mapping in {data_yaml}")
+
+    normalized = dict(data)
+    normalized["path"] = str(dataset_root)
+    if "val" not in normalized and "valid" in normalized:
+        normalized["val"] = normalized.pop("valid")
+
+    prepared_path = dataset_root / "data.sagemaker.yaml"
+    prepared_path.write_text(yaml.safe_dump(normalized, sort_keys=False), encoding="utf-8")
+    print(f"Prepared dataset config: {prepared_path}")
+    return prepared_path
+
+
+def copy_if_exists(source: Path, destination: Path) -> None:
+    if source.exists():
+        shutil.copy2(source, destination)
+        print(f"Saved {destination}")
+
+
+def main() -> None:
+    args = parse_args()
+    dataset_dir = Path(args.dataset_dir)
+    model_dir = Path(args.model_dir)
+    model_dir.mkdir(parents=True, exist_ok=True)
+
+    data_yaml = prepare_data_yaml(find_data_yaml(dataset_dir, args.data_yaml))
+    model = YOLO(args.model)
+
+    train_kwargs: dict[str, Any] = {
+        "data": str(data_yaml),
+        "epochs": args.epochs,
+        "imgsz": args.imgsz,
+        "batch": args.batch,
+        "workers": args.workers,
+        "patience": args.patience,
+        "project": str(model_dir / "runs"),
+        "name": "train",
+        "exist_ok": True,
+    }
+    if args.device:
+        train_kwargs["device"] = args.device
+
+    results = model.train(**train_kwargs)
+    save_dir = Path(results.save_dir)
+    best_pt = save_dir / "weights" / "best.pt"
+    last_pt = save_dir / "weights" / "last.pt"
+    trained_weights = best_pt if best_pt.exists() else last_pt
+    if not trained_weights.exists():
+        raise FileNotFoundError(f"Could not find trained weights in {save_dir / 'weights'}")
+
+    copy_if_exists(trained_weights, model_dir / "best.pt")
+    trained_model = YOLO(str(trained_weights))
+    onnx_path = Path(trained_model.export(format="onnx", imgsz=args.imgsz))
+    saved_onnx_path = sanitize_onnx(onnx_path, model_dir / "model.onnx")
+    print(f"Saved {saved_onnx_path}")
+
+    metrics = {
+        "model": args.model,
+        "epochs": args.epochs,
+        "imgsz": args.imgsz,
+        "batch": args.batch,
+        "workers": args.workers,
+        "patience": args.patience,
+        "data_yaml": str(data_yaml),
+        "weights": str(trained_weights),
+        "onnx": str(saved_onnx_path),
+    }
+    (model_dir / "metrics.json").write_text(json.dumps(metrics, indent=2), encoding="utf-8")
+    print(f"Saved model artifacts to {model_dir}")
+
+
+if __name__ == "__main__":
+    main()

examples/training/README.md

@@ -1,89 +0,0 @@
-# SageMaker Training Example
-
-This example downloads a small image-classification dataset, uploads it through `qc-cli`, and submits a live SageMaker training job.
-
-## Prerequisites
-
-- AWS credentials configured for the profile in `config.yaml`
-- Infrastructure already deployed with `qc-cli infra setup`
-- `config.yaml` updated with:
-
-```yaml
-s3:
-  bucket: your-bucket-name
-
-sagemaker:
-  training:
-    image_uri: 763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training:2.6-cpu-py312-ubuntu22.04-sagemaker-v1
-    instance_type: ml.m4.xlarge
-    instance_count: 1
-    source_dir: examples/training/source
-    entry_point: train.py
-    hyperparameters:
-      epochs: 1
-      batch-size: 32
-      learning-rate: 0.001
-      image-size: 160
-      validation-split: 0.2
-```
-
-## Training Hyperparameters
-
-Values under `sagemaker.training.hyperparameters` are passed to the training entry point as command-line arguments. For this example, they map to arguments defined in [source/train.py](source/train.py).
-
-Supported by this example:
-
-| Name | Type | Default | Description |
-|---|---:|---:|---|
-| `epochs` | int | `1` | Number of training epochs. |
-| `batch-size` | int | `32` | Images per training batch. |
-| `learning-rate` | float | `0.001` | Adam optimizer learning rate. |
-| `image-size` | int | `160` | Resize images to square `image-size x image-size`. |
-| `validation-split` | float | `0.2` | Fraction of data used for validation. |
-| `max-samples` | int | `0` | Optional cap for smoke tests; `0` means use all images. |
-| `seed` | int | `13` | Random seed for reproducible splitting. |
-| `num-workers` | int | `2` | DataLoader worker count. |
-
-Do not set `train-dir` or `model-dir` in normal SageMaker runs. SageMaker sets those automatically through `SM_CHANNEL_TRAIN` and `SM_MODEL_DIR`.
-
-## 1. Download The Dataset
-
-```bash
-bash examples/training/download_flower_photos.sh
-```
-
-This creates:
-
-```text
-examples/training/data/flower_photos_sagemaker/
-  daisy/
-  dandelion/
-  roses/
-  sunflowers/
-  tulips/
-```
-
-## 2. Run Training
-
-Run the training script and wait until it finishes:
-
-```bash
-bash examples/training/run_training.sh --config config.yaml --wait
-```
-
-Use a dataset that is already uploaded to `s3.data_prefix`:
-
-```bash
-bash examples/training/run_training.sh \
-  --config config.yaml \
-  --skip-upload \
-  --wait
-```
-
-## Notes
-
-- The default dataset path is `examples/training/data/flower_photos_sagemaker`.
-- Uploaded data uses the `s3.bucket` and `s3.data_prefix` values from `config.yaml`.
-- Training artifacts are written under `s3://<bucket>/<model_prefix>/`.
-- The SageMaker `model.tar.gz` contains `model.onnx`, `model.pt`, `class_to_idx.json`, and `metrics.json`.
-- SageMaker packages `examples/training/source`, installs `requirements.txt`, and runs `train.py`.

examples/training/download_flower_photos.sh

@@ -1,40 +0,0 @@
-#!/usr/bin/env bash
-set -euo pipefail
-
-DATASET_URL="https://storage.googleapis.com/download.tensorflow.org/example_images/flower_photos.tgz"
-DEST_DIR="${1:-examples/training/data}"
-ARCHIVE_PATH="${DEST_DIR}/flower_photos.tgz"
-RAW_DATASET_DIR="${DEST_DIR}/flower_photos"
-DATASET_DIR="${DEST_DIR}/flower_photos_sagemaker"
-CLASS_NAMES=("daisy" "dandelion" "roses" "sunflowers" "tulips")
-
-mkdir -p "${DEST_DIR}"
-
-if [[ -d "${DATASET_DIR}" ]]; then
-  echo "Dataset already exists: ${DATASET_DIR}"
-  echo "Use this path with run_training.py:"
-  echo "  ${DATASET_DIR}"
-  exit 0
-fi
-
-echo "Downloading TensorFlow flower_photos dataset..."
-if command -v curl >/dev/null 2>&1; then
-  curl -L "${DATASET_URL}" -o "${ARCHIVE_PATH}"
-elif command -v wget >/dev/null 2>&1; then
-  wget -O "${ARCHIVE_PATH}" "${DATASET_URL}"
-else
-  echo "Either curl or wget is required." >&2
-  exit 1
-fi
-
-echo "Extracting dataset..."
-tar -xzf "${ARCHIVE_PATH}" -C "${DEST_DIR}"
-
-echo "Preparing SageMaker directory layout..."
-mkdir -p "${DATASET_DIR}"
-for class_name in "${CLASS_NAMES[@]}"; do
-  cp -R "${RAW_DATASET_DIR}/${class_name}" "${DATASET_DIR}/${class_name}"
-done
-
-echo "Dataset ready: ${DATASET_DIR}"
-find "${DATASET_DIR}" -mindepth 1 -maxdepth 1 -type d -print | sort

examples/training/run_training.sh

@@ -1,111 +0,0 @@
-#!/usr/bin/env bash
-set -euo pipefail
-
-CONFIG_PATH="config.yaml"
-DATASET_DIR="examples/training/data/flower_photos_sagemaker"
-WAIT=false
-SKIP_UPLOAD=false
-POLL_SECONDS=60
-
-usage() {
-  cat <<EOF
-Usage: $0 [options]
-
-Options:
-  --config PATH          Path to qc-cli config file. Default: config.yaml
-  --dataset-dir PATH     Dataset directory to upload. Default: ${DATASET_DIR}
-  --skip-upload          Train against data already uploaded to s3.data_prefix.
-  --wait                 Poll until training completes.
-  -h, --help             Show this help.
-EOF
-}
-
-while [[ $# -gt 0 ]]; do
-  case "$1" in
-    --config)
-      CONFIG_PATH="$2"
-      shift 2
-      ;;
-    --dataset-dir)
-      DATASET_DIR="$2"
-      shift 2
-      ;;
-    --skip-upload)
-      SKIP_UPLOAD=true
-      shift
-      ;;
-    --wait)
-      WAIT=true
-      shift
-      ;;
-    -h|--help)
-      usage
-      exit 0
-      ;;
-    *)
-      echo "Unknown option: $1" >&2
-      usage >&2
-      exit 1
-      ;;
-  esac
-done
-
-if [[ ! -f "${CONFIG_PATH}" ]]; then
-  echo "Config not found: ${CONFIG_PATH}" >&2
-  exit 1
-fi
-
-if [[ "${SKIP_UPLOAD}" == false && ! -d "${DATASET_DIR}" ]]; then
-  echo "Dataset not found: ${DATASET_DIR}" >&2
-  echo "Run: bash examples/training/download_flower_photos.sh" >&2
-  exit 1
-fi
-
-run() {
-  echo "+ $*"
-  "$@"
-}
-
-run uv run qc-cli infra status --config "${CONFIG_PATH}"
-
-if [[ "${SKIP_UPLOAD}" == false ]]; then
-  run uv run qc-cli upload "${DATASET_DIR}" --config "${CONFIG_PATH}"
-fi
-
-TRAIN_OUTPUT="$(uv run qc-cli train start --config "${CONFIG_PATH}")"
-echo "${TRAIN_OUTPUT}"
-
-JOB_NAME="$(printf '%s\n' "${TRAIN_OUTPUT}" | grep -Eo 'qc-cli-[0-9]{8}-[0-9]{6}' | tail -n 1)"
-if [[ -z "${JOB_NAME}" ]]; then
-  echo "Could not find training job name in qc-cli output." >&2
-  exit 1
-fi
-
-echo "Submitted SageMaker training job: ${JOB_NAME}"
-
-if [[ "${WAIT}" == false ]]; then
-  run uv run qc-cli train status "${JOB_NAME}" --config "${CONFIG_PATH}"
-  exit 0
-fi
-
-while true; do
-  STATUS_OUTPUT="$(uv run qc-cli train status "${JOB_NAME}" --config "${CONFIG_PATH}")"
-  echo "${STATUS_OUTPUT}"
-
-  if printf '%s\n' "${STATUS_OUTPUT}" | grep -q 'Status:.*Completed'; then
-    echo "Training completed successfully."
-    exit 0
-  fi
-
-  if printf '%s\n' "${STATUS_OUTPUT}" | grep -q 'Status:.*Failed'; then
-    echo "Training failed." >&2
-    exit 1
-  fi
-
-  if printf '%s\n' "${STATUS_OUTPUT}" | grep -q 'Status:.*Stopped'; then
-    echo "Training stopped." >&2
-    exit 1
-  fi
-
-  sleep "${POLL_SECONDS}"
-done

examples/training/source/requirements.txt

@@ -1 +0,0 @@
-onnx==1.21.0

examples/training/source/train.py

@@ -1,192 +0,0 @@
-#!/usr/bin/env python3
-"""SageMaker entry point for CPU image-classification training."""
-
-from __future__ import annotations
-
-import argparse
-import json
-import os
-import random
-from pathlib import Path
-
-import torch
-from torch import nn
-from torch.utils.data import DataLoader, Subset, random_split
-from torchvision import datasets, transforms
-
-
-class SmallImageClassifier(nn.Module):
-    def __init__(self, class_count: int) -> None:
-        super().__init__()
-        self.features = nn.Sequential(
-            nn.Conv2d(3, 16, kernel_size=3, padding=1),
-            nn.ReLU(inplace=True),
-            nn.MaxPool2d(2),
-            nn.Conv2d(16, 32, kernel_size=3, padding=1),
-            nn.ReLU(inplace=True),
-            nn.MaxPool2d(2),
-            nn.Conv2d(32, 64, kernel_size=3, padding=1),
-            nn.ReLU(inplace=True),
-            nn.MaxPool2d(2),
-            nn.AdaptiveAvgPool2d((1, 1)),
-        )
-        self.classifier = nn.Linear(64, class_count)
-
-    def forward(self, x: torch.Tensor) -> torch.Tensor:
-        x = self.features(x)
-        x = torch.flatten(x, 1)
-        return self.classifier(x)
-
-
-def parse_args() -> argparse.Namespace:
-    parser = argparse.ArgumentParser()
-    parser.add_argument("--epochs", type=int, default=1)
-    parser.add_argument("--batch-size", type=int, default=32)
-    parser.add_argument("--learning-rate", type=float, default=0.001)
-    parser.add_argument("--image-size", type=int, default=160)
-    parser.add_argument("--validation-split", type=float, default=0.2)
-    parser.add_argument("--max-samples", type=int, default=0)
-    parser.add_argument("--seed", type=int, default=13)
-    parser.add_argument("--num-workers", type=int, default=2)
-    parser.add_argument("--train-dir", default=os.environ.get("SM_CHANNEL_TRAIN", "/opt/ml/input/data/train"))
-    parser.add_argument("--model-dir", default=os.environ.get("SM_MODEL_DIR", "/opt/ml/model"))
-    return parser.parse_args()
-
-
-def build_datasets(args: argparse.Namespace) -> tuple[Subset, Subset, dict[str, int]]:
-    transform = transforms.Compose(
-        [
-            transforms.Resize((args.image_size, args.image_size)),
-            transforms.ToTensor(),
-            transforms.Normalize(mean=(0.485, 0.456, 0.406), std=(0.229, 0.224, 0.225)),
-        ]
-    )
-    dataset = datasets.ImageFolder(args.train_dir, transform=transform)
-    if len(dataset.classes) < 2:
-        raise ValueError(f"Expected at least two classes in {args.train_dir}. Found: {dataset.classes}")
-
-    if args.max_samples > 0 and args.max_samples < len(dataset):
-        indices = list(range(len(dataset)))
-        random.Random(args.seed).shuffle(indices)
-        dataset = Subset(dataset, indices[: args.max_samples])
-
-    validation_size = max(1, int(len(dataset) * args.validation_split))
-    train_size = len(dataset) - validation_size
-    if train_size < 1:
-        raise ValueError("Not enough images to create a train/validation split.")
-
-    generator = torch.Generator().manual_seed(args.seed)
-    train_dataset, validation_dataset = random_split(dataset, [train_size, validation_size], generator=generator)
-    return train_dataset, validation_dataset, getattr(dataset, "dataset", dataset).class_to_idx
-
-
-def run_epoch(
-    model: nn.Module,
-    data_loader: DataLoader,
-    criterion: nn.Module,
-    optimizer: torch.optim.Optimizer | None,
-    device: torch.device,
-) -> tuple[float, float]:
-    training = optimizer is not None
-    model.train(training)
-
-    total_loss = 0.0
-    total_correct = 0
-    total_examples = 0
-
-    for images, labels in data_loader:
-        images = images.to(device)
-        labels = labels.to(device)
-
-        with torch.set_grad_enabled(training):
-            logits = model(images)
-            loss = criterion(logits, labels)
-
-            if training:
-                optimizer.zero_grad()
-                loss.backward()
-                optimizer.step()
-
-        total_loss += loss.item() * images.size(0)
-        total_correct += (logits.argmax(dim=1) == labels).sum().item()
-        total_examples += images.size(0)
-
-    return total_loss / total_examples, total_correct / total_examples
-
-
-def export_onnx(model: nn.Module, model_dir: Path, image_size: int) -> None:
-    model.eval()
-    dummy_input = torch.randn(1, 3, image_size, image_size)
-    torch.onnx.export(
-        model,
-        dummy_input,
-        model_dir / "model.onnx",
-        export_params=True,
-        opset_version=17,
-        do_constant_folding=True,
-        input_names=["input"],
-        output_names=["logits"],
-        dynamic_axes={
-            "input": {0: "batch_size"},
-            "logits": {0: "batch_size"},
-        },
-    )
-
-
-def main() -> None:
-    args = parse_args()
-    random.seed(args.seed)
-    torch.manual_seed(args.seed)
-
-    train_dataset, validation_dataset, class_to_idx = build_datasets(args)
-    train_loader = DataLoader(
-        train_dataset,
-        batch_size=args.batch_size,
-        shuffle=True,
-        num_workers=args.num_workers,
-    )
-    validation_loader = DataLoader(
-        validation_dataset,
-        batch_size=args.batch_size,
-        shuffle=False,
-        num_workers=args.num_workers,
-    )
-
-    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
-    model = SmallImageClassifier(class_count=len(class_to_idx)).to(device)
-    criterion = nn.CrossEntropyLoss()
-    optimizer = torch.optim.Adam(model.parameters(), lr=args.learning_rate)
-
-    print(f"Training on {device}. Classes: {sorted(class_to_idx)}")
-    metrics = []
-    for epoch in range(1, args.epochs + 1):
-        train_loss, train_accuracy = run_epoch(model, train_loader, criterion, optimizer, device)
-        validation_loss, validation_accuracy = run_epoch(model, validation_loader, criterion, None, device)
-        epoch_metrics = {
-            "epoch": epoch,
-            "train_loss": train_loss,
-            "train_accuracy": train_accuracy,
-            "validation_loss": validation_loss,
-            "validation_accuracy": validation_accuracy,
-        }
-        metrics.append(epoch_metrics)
-        print(json.dumps(epoch_metrics, sort_keys=True))
-
-    model_dir = Path(args.model_dir)
-    model_dir.mkdir(parents=True, exist_ok=True)
-    torch.save(
-        {
-            "model_state_dict": model.cpu().state_dict(),
-            "class_to_idx": class_to_idx,
-            "image_size": args.image_size,
-        },
-        model_dir / "model.pt",
-    )
-    export_onnx(model, model_dir, args.image_size)
-    (model_dir / "class_to_idx.json").write_text(json.dumps(class_to_idx, indent=2), encoding="utf-8")
-    (model_dir / "metrics.json").write_text(json.dumps(metrics, indent=2), encoding="utf-8")
-    print(f"Saved model artifacts to {model_dir}")
-
-
-if __name__ == "__main__":
-    main()

pyproject.toml

@@ -5,15 +5,19 @@ build-backend = "hatchling.build"
 [project]
 name = "qc-cli"
 version = "0.1.0"
-description = "CLI for SageMaker ONNX training and Qualcomm AI Hub optimization"
+description = "CLI for training and deploying models for Qualcomm AI Hub"
 requires-python = ">=3.13"
 dependencies = [
     "aws-cdk-lib>=2.180.0",
     "typer==0.25.0",
     "boto3>=1.34,<1.42",
     "constructs>=10.0.0",
+    "mlflow>=3.0",
+    "numpy>=1.26",
     "pydantic>=2.13.3",
     "pyyaml>=6.0.3",
+    "qai-hub>=0.49.0",
+    "sagemaker-mlflow>=0.4.0",
 ]
 
 [project.scripts]

src/aws/mlflow.py

@@ -17,3 +17,20 @@ def describe_tracking_server(region: str, profile: str, name: str) -> dict[str,
         ):
             return None
         raise
+
+
+def get_tracking_server_arn(region: str, profile: str, name: str) -> str:
+    server = describe_tracking_server(region, profile, name)
+    if not server:
+        raise ValueError(f"MLflow tracking server not found: {name}")
+
+    arn = server.get("TrackingServerArn")
+    if not arn:
+        raise ValueError(f"MLflow tracking server has no ARN: {name}")
+    return str(arn)
+
+
+def create_presigned_tracking_server_url(region: str, profile: str, name: str) -> str:
+    client = boto3.Session(profile_name=profile, region_name=region).client("sagemaker")
+    response = client.create_presigned_mlflow_tracking_server_url(TrackingServerName=name)
+    return str(response["AuthorizedUrl"])

src/aws/s3.py

@@ -21,6 +21,24 @@ def upload_file(
     return f"s3://{bucket}/{s3_key}"
 
 
+def download_file(
+    region: str,
+    profile: str,
+    s3_uri: str,
+    local_path: str,
+) -> str:
+    if not s3_uri.startswith("s3://"):
+        raise ValueError(f"Expected S3 URI, got: {s3_uri}")
+    bucket_key = s3_uri.removeprefix("s3://")
+    bucket, _, key = bucket_key.partition("/")
+    if not bucket or not key:
+        raise ValueError(f"Expected S3 URI with bucket and key, got: {s3_uri}")
+    dest = Path(local_path)
+    dest.parent.mkdir(parents=True, exist_ok=True)
+    _client(region, profile).download_file(bucket, key, str(dest))
+    return str(dest)
+
+
 def upload_dir(
     region: str,
     profile: str,

src/aws/sagemaker.py

@@ -36,6 +36,7 @@ class TrainingJobStatus:
     modified: datetime | None
     model_artifacts: str | None
     failure_reason: str | None
+    raw: dict[str, Any] = field(default_factory=dict)
 
 
 def _sm(session: Boto3SessionKwargs) -> SageMakerClient:
@@ -116,9 +117,20 @@ def get_training_job_status(session: Boto3SessionKwargs, job_name: str) -> Train
         modified=resp.get("LastModifiedTime"),
         model_artifacts=resp.get("ModelArtifacts", {}).get("S3ModelArtifacts"),
         failure_reason=resp.get("FailureReason"),
+        raw=dict(resp),
     )
 
 
+def get_model_artifacts(region: str, profile: str, job_name: str) -> str:
+    resp = boto3.Session(profile_name=profile, region_name=region).client("sagemaker").describe_training_job(
+        TrainingJobName=job_name
+    )
+    artifact = resp.get("ModelArtifacts", {}).get("S3ModelArtifacts")
+    if not artifact:
+        raise RuntimeError(f"Training job '{job_name}' does not have model artifacts yet.")
+    return str(artifact)
+
+
 def list_training_jobs(
     session: Boto3SessionKwargs,
     max_results: int = 10,

src/commands/ai_hub.py

@@ -0,0 +1,567 @@
+from collections.abc import Mapping, Sequence
+from dataclasses import dataclass
+from datetime import datetime
+from enum import StrEnum
+from pathlib import Path
+from typing import Any
+
+import qai_hub.hub as hub
+import typer
+from qai_hub.client import Device
+
+from src import state as state_ops
+from src.commands.utils import CONFIG_OPT, CONSOLE, load_cfg
+from src.config import Config
+from src.qualcomm import aihub_jobs
+from src.qualcomm.artifacts import ResolvedOnnx, resolve_onnx
+
+app = typer.Typer(help="Optimize, quantize, compile, validate, profile, and download models with Qualcomm Workbench")
+
+_RUNTIME_EXTENSIONS = {
+    "tflite": "tflite",
+    "qnn_context_binary": "bin",
+    "onnx": "onnx",
+}
+
+
+class UploadStep(StrEnum):
+    optimize = "optimize"
+    quantize = "quantize"
+    compile = "compile"
+    validate = "validate"
+    profile = "profile"
+
+
+@dataclass(frozen=True)
+class ResolvedModelSource:
+    model: str | Path
+    model_artifact: str | None = None
+
+
+def _input_specs(cfg: Config) -> dict[str, tuple[tuple[int, ...], str]]:
+    specs = {name: (tuple(shape), dtype) for name, (shape, dtype) in cfg.aihub.input_specs.items()}
+    if not specs:
+        CONSOLE.print("[red]aihub.input_specs must define at least one input.[/red]")
+        raise typer.Exit(1)
+    return specs
+
+
+def _load_inputs(
+    input_file: Path,
+    specs: Mapping[str, tuple[Sequence[int], str]],
+    input_name: str | None = None,
+) -> dict[str, Any]:
+    import numpy as np
+
+    if not input_file.exists():
+        raise FileNotFoundError(f"File not found: {input_file}")
+
+    if input_file.suffix == ".npz":
+        loaded = np.load(input_file)
+        missing = set(specs) - set(loaded.files)
+        if missing:
+            raise ValueError(f"Missing input(s) in NPZ: {', '.join(sorted(missing))}")
+        return {name: loaded[name] for name in specs}
+
+    if input_file.suffix == ".npy":
+        if input_name is None:
+            if len(specs) != 1:
+                raise ValueError("--input-name is required when config has multiple inputs")
+            input_name = next(iter(specs))
+        if input_name not in specs:
+            raise ValueError(f"Input name '{input_name}' is not defined in aihub.input_specs")
+        return {input_name: np.load(input_file)}
+
+    raise ValueError("Input file must be .npz or .npy")
+
+
+def _load_calibration(path: Path, specs: Mapping[str, tuple[Sequence[int], str]]) -> dict[str, Any]:
+    import numpy as np
+
+    if path.is_file():
+        return _load_inputs(path, specs)
+
+    if not path.is_dir():
+        raise FileNotFoundError(f"Calibration path not found: {path}")
+
+    if len(specs) != 1:
+        raise ValueError("Directory calibration data is supported only for single-input models.")
+    input_name = next(iter(specs))
+    samples = [np.load(p) for p in sorted(path.glob("*.npy"))]
+    if not samples:
+        raise ValueError(f"No .npy calibration samples found in {path}")
+    return {input_name: samples}
+
+
+def _job_name(cfg: Config, operation: str) -> str | None:
+    if not cfg.aihub.job_name:
+        return None
+    return f"{cfg.aihub.job_name}-{operation}"
+
+
+def _model_id_or_state(config_path: str, model_id: str | None, *, quantized: bool = False) -> str:
+    st = state_ops.store(config_path)
+    resolved = model_id or (st.get_last_quantized_model_id() if quantized else st.get_last_compiled_model_id())
+    if not resolved:
+        source = "quantized" if quantized else "compiled"
+        CONSOLE.print(f"[red]No {source} model found. Pass --model-id or run the previous AI Hub step first.[/red]")
+        raise typer.Exit(1)
+    return resolved
+
+
+def _resolve_model_source(
+    cfg: Config,
+    config_path: str,
+    *,
+    model_id: str | None = None,
+    previous_model_id: str | None = None,
+    from_job: str | None = None,
+    model_s3_uri: str | None = None,
+    onnx_path: str | None = None,
+) -> ResolvedModelSource:
+    if model_id:
+        return ResolvedModelSource(model_id)
+
+    has_explicit_source = bool(from_job or model_s3_uri or onnx_path)
+    if previous_model_id and not has_explicit_source:
+        return ResolvedModelSource(previous_model_id)
+
+    resolved = _resolve_onnx_source(
+        cfg,
+        config_path,
+        from_job=from_job,
+        model_s3_uri=model_s3_uri,
+        onnx_path=onnx_path,
+    )
+    return ResolvedModelSource(resolved.onnx_path, resolved.model_artifact)
+
+
+def _resolve_onnx_source(
+    cfg: Config,
+    config_path: str,
+    *,
+    from_job: str | None = None,
+    model_s3_uri: str | None = None,
+    onnx_path: str | None = None,
+) -> ResolvedOnnx:
+    st = state_ops.store(config_path)
+    last_training_job = st.get_last_training_job()
+    saved_model_artifact = None
+    if not from_job and not model_s3_uri and not onnx_path and not last_training_job:
+        saved_model_artifact = st.get_last_model_artifact()
+
+    return resolve_onnx(
+        cfg=cfg,
+        output_dir=cfg.aihub.output_dir,
+        from_job=from_job,
+        model_s3_uri=model_s3_uri or saved_model_artifact,
+        onnx_path=onnx_path,
+        last_training_job=last_training_job,
+    )
+
+
+def _device_selector(device: Device) -> str:
+    parts: list[str] = []
+    if device.name:
+        parts.append(f"name={device.name!r}")
+    if device.os:
+        parts.append(f"os={device.os!r}")
+    if device.attributes:
+        parts.append(f"attributes={device.attributes!r}")
+    return ", ".join(parts) if parts else "empty selector"
+
+
+def _validate_device(cfg: Config) -> None:
+    device = cfg.aihub.device
+    try:
+        matches = hub.get_devices(name=device.name, os=device.os, attributes=device.attributes)
+    except Exception as e:
+        CONSOLE.print(f"[red]Unable to validate AI Hub device {_device_selector(device)}: {e}[/red]")
+        raise typer.Exit(1)
+
+    if matches:
+        return
+
+    CONSOLE.print(f"[red]AI Hub device not found: {_device_selector(device)}[/red]")
+    CONSOLE.print("Run [bold]qai-hub list-devices[/bold] to see valid device names.")
+    raise typer.Exit(1)
+
+
+def _quantize_step(
+    cfg: Config,
+    config_path: str,
+    calibration_path: Path,
+    *,
+    model_id: str | None = None,
+    from_job: str | None = None,
+    model_s3_uri: str | None = None,
+    onnx_path: str | None = None,
+) -> str:
+    st = state_ops.store(config_path)
+    specs = _input_specs(cfg)
+    try:
+        source = _resolve_model_source(
+            cfg,
+            config_path,
+            model_id=model_id,
+            previous_model_id=st.get_last_optimized_model_id(),
+            from_job=from_job,
+            model_s3_uri=model_s3_uri,
+            onnx_path=onnx_path,
+        )
+        calibration_data = _load_calibration(calibration_path, specs)
+    except (FileNotFoundError, ValueError) as e:
+        CONSOLE.print(f"[red]{e}[/red]")
+        raise typer.Exit(1)
+
+    try:
+        hub_model = (
+            hub.upload_model(str(source.model), name=cfg.aihub.model_name)
+            if isinstance(source.model, Path)
+            else hub.get_model(source.model)
+        )
+        result = aihub_jobs.submit_quantize_job(
+            hub_model,
+            calibration_data,
+            cfg.aihub.quantize_options,
+            job_name=_job_name(cfg, "quantize"),
+        )
+    except Exception as e:
+        CONSOLE.print(f"[red]AI Hub quantize failed: {e}[/red]")
+        raise typer.Exit(1)
+
+    updates: dict[str, Any] = {
+        "last_quantize_job_id": result["job_id"],
+        "last_quantized_model_id": result["model_id"],
+    }
+    if source.model_artifact:
+        updates["last_model_artifact"] = source.model_artifact
+    st.update(**updates)
+    CONSOLE.print(f"[green]✓[/green] Quantize job: [bold]{result['job_id']}[/bold]")
+    CONSOLE.print(f"[green]✓[/green] Quantized model: [bold]{result['model_id']}[/bold]")
+    return str(result["model_id"])
+
+
+def _optimize_step(
+    cfg: Config,
+    config_path: str,
+    from_job: str | None,
+    model_s3_uri: str | None,
+    onnx_path: str | None,
+) -> str:
+    st = state_ops.store(config_path)
+    _validate_device(cfg)
+    specs = _input_specs(cfg)
+    try:
+        source = _resolve_onnx_source(
+            cfg,
+            config_path,
+            from_job=from_job,
+            model_s3_uri=model_s3_uri,
+            onnx_path=onnx_path,
+        )
+    except (FileNotFoundError, ValueError) as e:
+        CONSOLE.print(f"[red]{e}[/red]")
+        raise typer.Exit(1)
+
+    try:
+        hub_model = hub.upload_model(str(source.onnx_path), name=cfg.aihub.model_name)
+        result = aihub_jobs.submit_compile_job(
+            model=hub_model,
+            device=cfg.aihub.device,
+            input_specs=specs,
+            target_runtime="onnx",
+            job_name=_job_name(cfg, "optimize"),
+        )
+    except Exception as e:
+        CONSOLE.print(f"[red]AI Hub ONNX optimization failed: {e}[/red]")
+        raise typer.Exit(1)
+
+    st.update(
+        last_model_artifact=source.model_artifact,
+        last_optimize_job_id=result["job_id"],
+        last_optimized_model_id=result["model_id"],
+    )
+    CONSOLE.print(f"[green]✓[/green] ONNX optimization job: [bold]{result['job_id']}[/bold]")
+    CONSOLE.print(f"[green]✓[/green] Optimized ONNX model: [bold]{result['model_id']}[/bold]")
+    return str(result["model_id"])
+
+
+def _compile_step(
+    cfg: Config,
+    config_path: str,
+    *,
+    model_id: str | None = None,
+    from_job: str | None = None,
+    model_s3_uri: str | None = None,
+    onnx_path: str | None = None,
+) -> str:
+    st = state_ops.store(config_path)
+    _validate_device(cfg)
+    specs = _input_specs(cfg)
+    try:
+        source = _resolve_model_source(
+            cfg,
+            config_path,
+            model_id=model_id,
+            previous_model_id=st.get_last_quantized_model_id(),
+            from_job=from_job,
+            model_s3_uri=model_s3_uri,
+            onnx_path=onnx_path,
+        )
+    except (FileNotFoundError, ValueError) as e:
+        CONSOLE.print(f"[red]{e}[/red]")
+        raise typer.Exit(1)
+
+    try:
+        hub_model = (
+            hub.upload_model(str(source.model), name=cfg.aihub.model_name)
+            if isinstance(source.model, Path)
+            else hub.get_model(source.model)
+        )
+        result = aihub_jobs.submit_compile_job(
+            model=hub_model,
+            device=cfg.aihub.device,
+            input_specs=specs,
+            target_runtime=cfg.aihub.target_runtime,
+            options=cfg.aihub.compile_options,
+            job_name=_job_name(cfg, "compile"),
+        )
+    except Exception as e:
+        CONSOLE.print(f"[red]AI Hub compile failed: {e}[/red]")
+        raise typer.Exit(1)
+
+    updates: dict[str, Any] = {
+        "last_compile_job_id": result["job_id"],
+        "last_compiled_model_id": result["model_id"],
+    }
+    if source.model_artifact:
+        updates["last_model_artifact"] = source.model_artifact
+    st.update(**updates)
+    CONSOLE.print(f"[green]✓[/green] Compile job: [bold]{result['job_id']}[/bold]")
+    CONSOLE.print(f"[green]✓[/green] Compiled model: [bold]{result['model_id']}[/bold]")
+    return str(result["model_id"])
+
+
+def _validate_step(
+    cfg: Config,
+    config_path: str,
+    input_file: Path,
+    model_id: str | None,
+    input_name: str | None,
+) -> str:
+    _validate_device(cfg)
+    specs = _input_specs(cfg)
+    resolved_model_id = _model_id_or_state(config_path, model_id)
+    try:
+        inputs = _load_inputs(input_file, specs, input_name)
+    except (FileNotFoundError, ValueError) as e:
+        CONSOLE.print(f"[red]{e}[/red]")
+        raise typer.Exit(1)
+
+    run = datetime.now().strftime("%Y%m%d-%H%M%S")
+    out_dir = Path(cfg.aihub.output_dir) / run / "validation"
+    try:
+        hub_model = hub.get_model(resolved_model_id)
+        result = aihub_jobs.submit_inference_job(
+            hub_model,
+            cfg.aihub.device,
+            inputs,
+            out_dir,
+            job_name=_job_name(cfg, "validate"),
+        )
+    except Exception as e:
+        CONSOLE.print(f"[red]AI Hub inference failed: {e}[/red]")
+        raise typer.Exit(1)
+
+    state_ops.store(config_path).update(last_inference_job_id=result["job_id"])
+    CONSOLE.print(f"[green]✓[/green] Inference job: [bold]{result['job_id']}[/bold]")
+    outputs = result.get("outputs")
+    if isinstance(outputs, dict):
+        for name, value in outputs.items():
+            CONSOLE.print(f"  {name}: shape={getattr(value, 'shape', '?')}")
+    CONSOLE.print(f"Outputs: [cyan]{out_dir}[/cyan]")
+    return str(result["job_id"])
+
+
+def _profile_step(cfg: Config, config_path: str, model_id: str | None) -> str:
+    _validate_device(cfg)
+    resolved_model_id = _model_id_or_state(config_path, model_id)
+    try:
+        hub_model = hub.get_model(resolved_model_id)
+        result = aihub_jobs.submit_profile_job(
+            hub_model,
+            cfg.aihub.device,
+            cfg.aihub.profile_options,
+            job_name=_job_name(cfg, "profile"),
+        )
+    except Exception as e:
+        CONSOLE.print(f"[red]AI Hub profile failed: {e}[/red]")
+        raise typer.Exit(1)
+    state_ops.store(config_path).update(last_profile_job_id=result["job_id"])
+    CONSOLE.print(f"[green]✓[/green] Profile job: [bold]{result['job_id']}[/bold]")
+    return str(result["job_id"])
+
+
+@app.command()
+def optimize(
+    from_job: str | None = typer.Option(None, "--from-job", help="Training job name whose model artifact should optimize"),
+    model_s3_uri: str | None = typer.Option(None, "--model-s3-uri", help="S3 URI of model.tar.gz to optimize"),
+    onnx_path: str | None = typer.Option(
+        None, "--onnx-path", help="Local ONNX path or ONNX path inside extracted artifact"
+    ),
+    config: str = CONFIG_OPT,
+) -> None:
+    """Optimize an external model into a Workbench-produced ONNX model."""
+    cfg = load_cfg(config)
+    _optimize_step(cfg, config, from_job, model_s3_uri, onnx_path)
+
+
+@app.command()
+def quantize(
+    calibration_path: Path = typer.Argument(..., help="Calibration .npz file or directory of .npy samples"),
+    model_id: str | None = typer.Option(None, "--model-id", help="AI Hub optimized ONNX model ID"),
+    from_job: str | None = typer.Option(None, "--from-job", help="Training job name whose model artifact should quantize"),
+    model_s3_uri: str | None = typer.Option(None, "--model-s3-uri", help="S3 URI of model.tar.gz to quantize"),
+    onnx_path: str | None = typer.Option(
+        None, "--onnx-path", help="Local ONNX path or ONNX path inside extracted artifact"
+    ),
+    config: str = CONFIG_OPT,
+) -> None:
+    """Quantize an ONNX model to INT8."""
+    cfg = load_cfg(config)
+    _quantize_step(
+        cfg,
+        config,
+        calibration_path,
+        model_id=model_id,
+        from_job=from_job,
+        model_s3_uri=model_s3_uri,
+        onnx_path=onnx_path,
+    )
+
+
+@app.command()
+def compile(
+    model_id: str | None = typer.Option(None, "--model-id", help="AI Hub model ID to compile"),
+    from_job: str | None = typer.Option(None, "--from-job", help="Training job name whose model artifact should compile"),
+    model_s3_uri: str | None = typer.Option(None, "--model-s3-uri", help="S3 URI of model.tar.gz to compile"),
+    onnx_path: str | None = typer.Option(
+        None, "--onnx-path", help="Local ONNX path or ONNX path inside extracted artifact"
+    ),
+    config: str = CONFIG_OPT,
+) -> None:
+    """Compile a model for the configured Qualcomm AI Hub target."""
+    cfg = load_cfg(config)
+    _compile_step(
+        cfg,
+        config,
+        model_id=model_id,
+        from_job=from_job,
+        model_s3_uri=model_s3_uri,
+        onnx_path=onnx_path,
+    )
+
+
+@app.command()
+def validate(
+    input_file: Path = typer.Argument(..., help="NumPy .npz or .npy inputs to run on device"),
+    model_id: str | None = typer.Option(None, "--model-id", help="AI Hub compiled model ID"),
+    input_name: str | None = typer.Option(None, "--input-name", help="Input name for .npy files"),
+    config: str = CONFIG_OPT,
+) -> None:
+    """Run an AI Hub inference job using sample inputs."""
+    cfg = load_cfg(config)
+    _validate_step(cfg, config, input_file, model_id, input_name)
+
+
+@app.command()
+def profile(
+    model_id: str | None = typer.Option(None, "--model-id", help="AI Hub compiled model ID"),
+    config: str = CONFIG_OPT,
+) -> None:
+    """Profile a compiled model on the configured AI Hub device."""
+    cfg = load_cfg(config)
+    _profile_step(cfg, config, model_id)
+
+
+@app.command()
+def upload(
+    calibration_path: Path = typer.Argument(..., help="Calibration .npz file or directory of .npy samples"),
+    input_file: Path = typer.Argument(..., help="Validation .npz or .npy inputs to run on device"),
+    from_step: UploadStep = typer.Option(UploadStep.optimize, "--from-step", help="Resume from this Workbench step"),
+    from_job: str | None = typer.Option(None, "--from-job", help="Training job name whose model artifact should upload"),
+    model_s3_uri: str | None = typer.Option(None, "--model-s3-uri", help="S3 URI of model.tar.gz to upload"),
+    onnx_path: str | None = typer.Option(
+        None, "--onnx-path", help="Local ONNX path or ONNX path inside extracted artifact"
+    ),
+    input_name: str | None = typer.Option(None, "--input-name", help="Input name for .npy validation files"),
+    config: str = CONFIG_OPT,
+) -> None:
+    """Optimize, quantize, optionally compile, validate, and profile a model."""
+    cfg = load_cfg(config)
+    steps = [UploadStep.optimize, UploadStep.quantize, UploadStep.compile, UploadStep.validate, UploadStep.profile]
+    selected = steps[steps.index(from_step) :]
+
+    optimized_model_id: str | None = None
+    quantized_model_id: str | None = None
+    compiled_model_id: str | None = None
+    if UploadStep.optimize in selected:
+        optimized_model_id = _optimize_step(cfg, config, from_job, model_s3_uri, onnx_path)
+    if UploadStep.quantize in selected:
+        if UploadStep.optimize not in selected:
+            optimized_model_id = state_ops.store(config).get_last_optimized_model_id()
+            if not optimized_model_id:
+                CONSOLE.print(
+                    "[red]No optimized ONNX model found. Resume from --from-step optimize or run "
+                    "'qc-cli ai-hub optimize' first.[/red]"
+                )
+                raise typer.Exit(1)
+        quantized_model_id = _quantize_step(
+            cfg,
+            config,
+            calibration_path,
+            model_id=optimized_model_id,
+        )
+    if UploadStep.compile in selected:
+        if cfg.aihub.target_runtime == "onnx":
+            compiled_model_id = quantized_model_id or state_ops.store(config).get_last_quantized_model_id()
+            if not compiled_model_id:
+                CONSOLE.print(
+                    "[red]No quantized ONNX model found. Resume from --from-step quantize or run "
+                    "'qc-cli ai-hub quantize' first.[/red]"
+                )
+                raise typer.Exit(1)
+            state_ops.store(config).update(last_compiled_model_id=compiled_model_id)
+            CONSOLE.print("[green]✓[/green] Target runtime is ONNX; skipping final compile.")
+        else:
+            compiled_model_id = _compile_step(
+                cfg,
+                config,
+                model_id=quantized_model_id,
+            )
+    if UploadStep.validate in selected:
+        _validate_step(cfg, config, input_file, compiled_model_id, input_name)
+    if UploadStep.profile in selected:
+        _profile_step(cfg, config, compiled_model_id)
+
+
+@app.command()
+def download(
+    model_id: str | None = typer.Option(None, "--model-id", help="AI Hub compiled model ID"),
+    output: Path | None = typer.Option(None, "--output", "-o", help="Destination file path"),
+    config: str = CONFIG_OPT,
+) -> None:
+    """Download the last compiled deployable artifact from AI Hub."""
+    cfg = load_cfg(config)
+    resolved_model_id = _model_id_or_state(config, model_id)
+    ext = _RUNTIME_EXTENSIONS.get(cfg.aihub.target_runtime, cfg.aihub.target_runtime)
+    dest = output or (Path(cfg.aihub.output_dir) / f"model.{ext}")
+
+    try:
+        written = aihub_jobs.download_model(resolved_model_id, dest)
+    except Exception as e:
+        CONSOLE.print(f"[red]AI Hub download failed: {e}[/red]")
+        raise typer.Exit(1)
+    state_ops.store(config).update(last_downloaded_model=written)
+    CONSOLE.print(f"[green]✓[/green] Downloaded model: [cyan]{written}[/cyan]")

src/commands/infra.py

@@ -77,7 +77,8 @@ def setup(
     if outputs.get("SageMakerRoleArn"):
         CONSOLE.print(f"[green]✓[/green] IAM role:  {outputs['SageMakerRoleArn']}")
     if cfg.mlflow.mode is MlflowMode.create and outputs.get("MlflowTrackingServerArn"):
-        CONSOLE.print(f"[green]✓[/green] MLflow:    {outputs['MlflowTrackingServerArn']}")
+        mlflow_name = outputs.get("MlflowTrackingServerName", cfg.managed_mlflow_tracking_server_name)
+        CONSOLE.print(f"[green]✓[/green] MLflow:    {mlflow_name}")
     elif cfg.mlflow.mode is MlflowMode.existing:
         CONSOLE.print(f"[green]✓[/green] MLflow:    {cfg.mlflow.tracking_server_name}")
     CONSOLE.print("\n[bold green]Infrastructure ready.[/bold green]")
@@ -102,7 +103,7 @@ def status(config: str = CONFIG_OPT) -> None:
         if cfg.mlflow.mode is not MlflowMode.disabled:
             table.add_row(
                 "MLflow",
-                cfg.mlflow.tracking_server_name or "-",
+                cfg.effective_mlflow_tracking_server_name or "-",
                 "[red]unknown[/red]",
                 "-",
             )
@@ -126,7 +127,7 @@ def status(config: str = CONFIG_OPT) -> None:
     if cfg.mlflow.mode is MlflowMode.create:
         table.add_row(
             "MLflow",
-            cfg.mlflow.tracking_server_name or "-",
+            outputs.get("MlflowTrackingServerName", cfg.managed_mlflow_tracking_server_name),
             "[green]managed[/green]",
             outputs.get("MlflowTrackingServerArn", outputs.get("MlflowArtifactUri", "-")),
         )
@@ -209,6 +210,7 @@ def _role_name(configured_name: str, role_arn: str) -> str:
         return role_arn.rsplit("/", 1)[-1]
     return "-"
 
+
 def _destroy_account_id(config_path: str, cfg: Config) -> str:
     config_dir = str(Path(config_path).parent)
     state = read_infra_state(config_dir)

src/commands/init.py

@@ -0,0 +1,40 @@
+import secrets
+from pathlib import Path
+
+import typer
+import yaml
+
+from src.commands.utils import CONSOLE
+from src.config import GENERATED_STACK_PREFIX, Config, InfraConfig, S3Config
+
+app = typer.Typer()
+
+
+@app.command()
+def init(
+    output: str = typer.Option("config.yaml", help="Destination path for the config file"),
+    force: bool = typer.Option(False, "--force", "-f", help="Overwrite an existing config file"),
+) -> None:
+    """Write a starter config.yaml to the current directory."""
+    dest = Path(output)
+    if dest.exists() and not force:
+        CONSOLE.print(f"[yellow]{dest} already exists.[/yellow] Use --force to overwrite.")
+        raise typer.Exit(1)
+
+    config = _new_isolated_config()
+    dest.parent.mkdir(parents=True, exist_ok=True)
+    config_data = config.model_dump(mode="json")
+    config_data["sagemaker"].pop("role_name", None)
+    with open(dest, "w") as f:
+        yaml.safe_dump(config_data, f, sort_keys=False)
+
+    CONSOLE.print(f"[green]✓[/green] Config written to [bold]{dest}[/bold]")
+    CONSOLE.print("Edit [cyan]sagemaker.training.image_uri[/cyan] before running training commands.")
+
+
+def _new_isolated_config() -> Config:
+    suffix = secrets.token_hex(6)
+    namespace = f"{GENERATED_STACK_PREFIX}{suffix}"
+    config = Config(infra=InfraConfig(stack_name=namespace))
+    config.s3 = S3Config(bucket=f"{namespace}-data")
+    return config

src/commands/mlflow.py

@@ -0,0 +1,41 @@
+import webbrowser
+
+import typer
+
+from src.aws import mlflow as aws_mlflow
+from src.commands.utils import CONFIG_OPT, CONSOLE, load_cfg
+
+app = typer.Typer(help="Manage MLflow tracking server access")
+
+
+@app.command(name="open")
+def open_mlflow(config: str = CONFIG_OPT) -> None:
+    """Open a presigned URL for the configured MLflow tracking server."""
+    cfg = load_cfg(config)
+    tracking_server_name = cfg.effective_mlflow_tracking_server_name
+    if not tracking_server_name:
+        CONSOLE.print("[red]MLflow is disabled in config.yaml.[/red]")
+        raise typer.Exit(1)
+
+    try:
+        url = aws_mlflow.create_presigned_tracking_server_url(
+            cfg.aws.region,
+            cfg.aws.profile,
+            tracking_server_name,
+        )
+    except Exception as e:
+        CONSOLE.print("[yellow]Could not create a SageMaker MLflow UI URL.[/yellow]")
+        CONSOLE.print(f"Tracking server: [cyan]{tracking_server_name}[/cyan]")
+        CONSOLE.print(f"Reason: {e}")
+        CONSOLE.print(
+            "This command can create presigned URLs only for MLflow tracking servers managed by "
+            "Amazon SageMaker. If this is an external MLflow server, open it with that server's own URL."
+        )
+        raise typer.Exit(1)
+
+    CONSOLE.print(f"MLflow tracking server: [cyan]{tracking_server_name}[/cyan]")
+    CONSOLE.print(f"MLflow UI: {url}")
+    if webbrowser.open(url):
+        CONSOLE.print("[green]✓[/green] Opened MLflow UI in your browser.")
+    else:
+        CONSOLE.print("[yellow]Could not open a browser automatically. Open the URL above manually.[/yellow]")

src/commands/train.py

@@ -8,8 +8,9 @@ from src import state as state_ops
 from src.aws import iam
 from src.aws import sagemaker as sm_ops
 from src.commands.utils import CONFIG_OPT, CONSOLE, load_cfg
-from src.config import Config
+from src.config import Config, MlflowMode
 from src.infra.state import read_infra_state
+from src.tracking.mlflow import MlflowTracker
 
 app = typer.Typer(help="Manage SageMaker training jobs")
 
@@ -22,6 +23,14 @@ _STATUS_COLOR = {
 }
 
 
+def _tracker(cfg):
+    try:
+        return MlflowTracker.from_config(cfg)
+    except Exception as e:
+        CONSOLE.print(f"[red]MLflow setup failed: {e}[/red]")
+        raise typer.Exit(1)
+
+
 def _config_dir(config_path: str) -> str:
     return str(Path(config_path).parent)
 
@@ -58,6 +67,7 @@ def start(config: str = CONFIG_OPT) -> None:
         CONSOLE.print(f"[red]{e}[/red]")
         raise typer.Exit(1)
 
+    tracker = _tracker(cfg)
     job_name = f"qc-cli-{datetime.now().strftime('%Y%m%d-%H%M%S')}"
     s3_train_uri = f"s3://{cfg.s3.bucket}/{cfg.s3.data_prefix}"
     s3_output = f"s3://{cfg.s3.bucket}/{cfg.s3.model_prefix}"
@@ -77,9 +87,21 @@ def start(config: str = CONFIG_OPT) -> None:
     )
     sm_ops.start_training_job(cfg.aws.boto3_session, training_job)
 
-    state_ops.write_state(_config_dir(config), last_training_job=job_name)
+    st = state_ops.store(config)
+    st.set_last_training_job(job_name)
+    run_id = tracker.start_training_run(
+        training_job,
+        region=cfg.aws.region,
+        profile=cfg.aws.profile,
+        role_arn=role_arn,
+    )
+    if run_id:
+        st.update_training_job(job_name, mlflow_run_id=run_id)
 
     CONSOLE.print(f"[green]✓[/green] Job submitted: [bold]{job_name}[/bold]")
+    if run_id:
+        CONSOLE.print(f"MLflow run: [cyan]{run_id}[/cyan]")
+        CONSOLE.print("Open MLflow: [cyan]qc-cli mlflow open[/cyan]")
     CONSOLE.print("Track progress: [cyan]qc-cli train status[/cyan]")
 
 
@@ -90,9 +112,10 @@ def status(
 ) -> None:
     """Show training job status."""
     cfg = load_cfg(config)
+    st = state_ops.store(config)
 
     if not job_name:
-        job_name = state_ops.get_last_training_job(_config_dir(config))
+        job_name = st.get_last_training_job()
         if not job_name:
             CONSOLE.print(
                 "[red]No training job found in state. Pass a job name or run 'qc-cli train start' first.[/red]"
@@ -111,6 +134,25 @@ def status(
     if status.failure_reason:
         CONSOLE.print(f"[red]Failure: {status.failure_reason}[/red]")
 
+    job_state = st.get_training_job(job_name)
+    run_id = job_state.get("mlflow_run_id")
+    already_registered = job_state.get("registered_model_version")
+    if run_id and not already_registered and status.status in {"Completed", "Failed", "Stopped"}:
+        tracker = _tracker(cfg)
+        version = tracker.finalize_training_run(
+            run_id=str(run_id),
+            training_job_status=status,
+        )
+        updates = {"mlflow_finalized_status": status.status}
+        if version:
+            updates["registered_model_version"] = version
+        st.update_training_job(job_name, **updates)
+        if version:
+            st.set_latest_experiment_model_version(version)
+            CONSOLE.print(f"MLflow model version: [cyan]{version}[/cyan] ([cyan]experiment-latest[/cyan])")
+    if run_id and cfg.mlflow.mode is not MlflowMode.disabled:
+        CONSOLE.print("Open MLflow: [cyan]qc-cli mlflow open[/cyan]")
+
 
 @app.command(name="list")
 def list_jobs(

src/commands/upload.py

@@ -0,0 +1,70 @@
+from pathlib import Path
+
+import typer
+from rich.progress import BarColumn, Progress, SpinnerColumn, TaskProgressColumn, TextColumn
+
+from src.aws import s3 as s3_ops
+from src.commands.utils import CONFIG_OPT, CONSOLE, load_cfg
+
+app = typer.Typer()
+
+
+@app.command()
+def upload(
+    path: Path = typer.Argument(..., help="Local file or directory to upload"),
+    s3_key: str | None = typer.Option(None, "--s3-key", help="S3 key for file uploads"),
+    config: str = CONFIG_OPT,
+) -> None:
+    """Upload a local file or directory to S3."""
+    cfg = load_cfg(config)
+
+    if path.is_file():
+        key = s3_key or f"{cfg.s3.data_prefix.rstrip('/')}/{path.name}"
+        try:
+            with CONSOLE.status(f"Uploading {path.name}..."):
+                uri = s3_ops.upload_file(cfg.aws.region, cfg.aws.profile, cfg.s3.bucket, str(path), key)
+        except Exception as e:
+            CONSOLE.print(f"[red]Upload failed: {e}[/red]")
+            raise typer.Exit(1)
+
+        CONSOLE.print(f"[green]✓[/green] {path.name} -> {uri}")
+        return
+
+    if path.is_dir():
+        if s3_key is not None:
+            CONSOLE.print("[red]--s3-key can only be used when uploading a single file.[/red]")
+            raise typer.Exit(1)
+
+        files = [file for file in path.rglob("*") if file.is_file()]
+        if not files:
+            CONSOLE.print("[yellow]No files found in directory.[/yellow]")
+            raise typer.Exit(0)
+
+        prefix = cfg.s3.data_prefix
+        CONSOLE.print(f"Uploading {len(files)} files to s3://{cfg.s3.bucket}/{prefix.rstrip('/')}/")
+        try:
+            with Progress(
+                SpinnerColumn(),
+                TextColumn("[progress.description]{task.description}"),
+                BarColumn(),
+                TaskProgressColumn(),
+                console=CONSOLE,
+            ) as progress:
+                task = progress.add_task("Uploading...", total=len(files))
+                count = s3_ops.upload_dir(
+                    cfg.aws.region,
+                    cfg.aws.profile,
+                    cfg.s3.bucket,
+                    str(path),
+                    prefix,
+                    on_progress=lambda: progress.advance(task),
+                )
+        except Exception as e:
+            CONSOLE.print(f"[red]Upload failed: {e}[/red]")
+            raise typer.Exit(1)
+
+        CONSOLE.print(f"[green]✓[/green] Uploaded {count} files to s3://{cfg.s3.bucket}/{prefix.rstrip('/')}/")
+        return
+
+    CONSOLE.print(f"[red]Path not found: {path}[/red]")
+    raise typer.Exit(1)

src/config.py

@@ -1,19 +1,20 @@
 import re
-from enum import Enum
+from enum import StrEnum
 from typing import Any, Literal, TypedDict
 
 from mypy_boto3_s3.literals import BucketLocationConstraintType
 from mypy_boto3_sagemaker.literals import TrainingInstanceTypeType
-from pydantic import BaseModel, Field, model_validator
+from pydantic import BaseModel, Field, field_validator, model_validator
+from qai_hub.client import Device
 
 
-class MlflowMode(str, Enum):
+class MlflowMode(StrEnum):
     disabled = "disabled"
     create = "create"
     existing = "existing"
 
 
-class MlflowServerSize(str, Enum):
+class MlflowServerSize(StrEnum):
     small = "Small"
     medium = "Medium"
     large = "Large"
@@ -80,9 +81,31 @@ class SageMakerConfig(BaseModel):
     training: TrainingConfig = Field(default_factory=TrainingConfig)
 
 
+class AIHubConfig(BaseModel):
+    device: Device = Field(default_factory=lambda: Device("Samsung Galaxy S25 (Family)"))
+    target_runtime: str = "tflite"
+    input_specs: dict[str, tuple[list[int], str]] = Field(default_factory=dict)
+    job_name: str | None = None
+    model_name: str | None = None
+    compile_options: str | None = None
+    profile_options: str | None = None
+    quantize_options: str | None = None
+    output_dir: str = "build/qai-hub"
+
+    @field_validator("device", mode="before")
+    @classmethod
+    def parse_device(cls, value: Any) -> Any:
+        if isinstance(value, str):
+            return Device(value)
+        return value
+
+
 class MlflowConfig(BaseModel):
     mode: MlflowMode = MlflowMode.disabled
     tracking_server_name: str | None = None
+    experiment_name: str = "qc-cli-training"
+    registered_model_name: str = "qc-cli-model"
+    register_trained_models: bool = True
     artifact_prefix: str = "mlflow/"
     tracking_server_size: MlflowServerSize = MlflowServerSize.small
     mlflow_version: str | None = None
@@ -91,8 +114,8 @@ class MlflowConfig(BaseModel):
 
     @model_validator(mode="after")
     def require_tracking_server_name(self) -> "MlflowConfig":
-        if self.mode in {MlflowMode.create, MlflowMode.existing} and not self.tracking_server_name:
-            raise ValueError("mlflow.tracking_server_name is required when mlflow.mode is create or existing")
+        if self.mode is MlflowMode.existing and not self.tracking_server_name:
+            raise ValueError("mlflow.tracking_server_name is required when mlflow.mode is existing")
         return self
 
 
@@ -101,4 +124,17 @@ class Config(BaseModel):
     aws: AwsConfig = Field(default_factory=AwsConfig)
     s3: S3Config = Field(default_factory=S3Config)
     sagemaker: SageMakerConfig = Field(default_factory=SageMakerConfig)
+    aihub: AIHubConfig = Field(default_factory=AIHubConfig)
     mlflow: MlflowConfig = Field(default_factory=MlflowConfig)
+
+    @property
+    def managed_mlflow_tracking_server_name(self) -> str:
+        return f"{self.infra.stack_name}-mlflow"
+
+    @property
+    def effective_mlflow_tracking_server_name(self) -> str | None:
+        if self.mlflow.mode is MlflowMode.disabled:
+            return None
+        if self.mlflow.mode is MlflowMode.existing:
+            return self.mlflow.tracking_server_name
+        return self.managed_mlflow_tracking_server_name

src/infra/stack.py

@@ -74,6 +74,7 @@ class QCStack(Stack):
         CfnOutput(self, "SageMakerRoleArn", value=role.attr_arn)
 
         if config.mlflow.mode is MlflowMode.create:
+            tracking_server_name = config.managed_mlflow_tracking_server_name
             artifact_prefix = config.mlflow.artifact_prefix.strip("/")
             artifact_uri = (
                 f"s3://{data_bucket.bucket_name}/{artifact_prefix}/"
@@ -145,14 +146,14 @@ class QCStack(Stack):
                 "MlflowTrackingServer",
                 artifact_store_uri=artifact_uri,
                 role_arn=mlflow_role.attr_arn,
-                tracking_server_name=config.mlflow.tracking_server_name or "",
+                tracking_server_name=tracking_server_name,
                 automatic_model_registration=config.mlflow.automatic_model_registration,
                 mlflow_version=config.mlflow.mlflow_version,
                 tracking_server_size=config.mlflow.tracking_server_size.value,
                 weekly_maintenance_window_start=config.mlflow.weekly_maintenance_window_start,
             )
 
-            CfnOutput(self, "MlflowTrackingServerName", value=config.mlflow.tracking_server_name or "")
+            CfnOutput(self, "MlflowTrackingServerName", value=tracking_server_name)
             CfnOutput(self, "MlflowTrackingServerArn", value=tracking_server.attr_tracking_server_arn)
             CfnOutput(self, "MlflowArtifactUri", value=artifact_uri)
             CfnOutput(self, "MlflowRoleArn", value=mlflow_role.attr_arn)

src/main.py

@@ -1,114 +1,14 @@
-import secrets
-from pathlib import Path
-
 import typer
-import yaml
-from rich.console import Console
-from rich.progress import BarColumn, Progress, SpinnerColumn, TaskProgressColumn, TextColumn
 
-from src.aws import s3 as s3_ops
-from src.commands import infra, train
-from src.commands.utils import CONFIG_OPT, load_cfg
-from src.config import GENERATED_STACK_PREFIX, Config, InfraConfig, S3Config
+from src.commands import ai_hub, infra, init, mlflow, train, upload
 
 app = typer.Typer(
     help="qc-cli: End-to-end model managment for Qualcomm AI Hub.",
     no_args_is_help=True,
 )
+app.add_typer(init.app)
+app.add_typer(upload.app)
+app.add_typer(mlflow.app, name="mlflow")
 app.add_typer(infra.app, name="infra")
 app.add_typer(train.app, name="train")
-
-console = Console()
-
-
-@app.command()
-def init(
-    output: str = typer.Option("config.yaml", help="Destination path for the config file"),
-    force: bool = typer.Option(False, "--force", "-f", help="Overwrite an existing config file"),
-) -> None:
-    """Write a starter config.yaml to the current directory."""
-    dest = Path(output)
-    if dest.exists() and not force:
-        console.print(f"[yellow]{dest} already exists.[/yellow] Use --force to overwrite.")
-        raise typer.Exit(1)
-
-    config = _new_isolated_config()
-    dest.parent.mkdir(parents=True, exist_ok=True)
-    config_data = config.model_dump(mode="json")
-    config_data["sagemaker"].pop("role_name", None)
-    with open(dest, "w") as f:
-        yaml.safe_dump(config_data, f, sort_keys=False)
-
-    console.print(f"[green]✓[/green] Config written to [bold]{dest}[/bold]")
-    console.print(
-        "Edit [cyan]sagemaker.training.image_uri[/cyan] before running training commands."
-    )
-
-
-def _new_isolated_config() -> Config:
-    suffix = secrets.token_hex(6)
-    namespace = f"{GENERATED_STACK_PREFIX}{suffix}"
-    config = Config(infra=InfraConfig(stack_name=namespace))
-    config.s3 = S3Config(bucket=f"{namespace}-data")
-    return config
-
-
-@app.command()
-def upload(
-    path: Path = typer.Argument(..., help="Local file or directory to upload"),
-    s3_key: str | None = typer.Option(None, "--s3-key", help="S3 key for file uploads"),
-    config: str = CONFIG_OPT,
-) -> None:
-    """Upload a local file or directory to S3."""
-    cfg = load_cfg(config)
-
-    if path.is_file():
-        key = s3_key or f"{cfg.s3.data_prefix.rstrip('/')}/{path.name}"
-        try:
-            with console.status(f"Uploading {path.name}..."):
-                uri = s3_ops.upload_file(cfg.aws.region, cfg.aws.profile, cfg.s3.bucket, str(path), key)
-        except Exception as e:
-            console.print(f"[red]Upload failed: {e}[/red]")
-            raise typer.Exit(1)
-
-        console.print(f"[green]✓[/green] {path.name} -> {uri}")
-        return
-
-    if path.is_dir():
-        if s3_key is not None:
-            console.print("[red]--s3-key can only be used when uploading a single file.[/red]")
-            raise typer.Exit(1)
-
-        files = [file for file in path.rglob("*") if file.is_file()]
-        if not files:
-            console.print("[yellow]No files found in directory.[/yellow]")
-            raise typer.Exit(0)
-
-        prefix = cfg.s3.data_prefix
-        console.print(f"Uploading {len(files)} files to s3://{cfg.s3.bucket}/{prefix.rstrip('/')}/")
-        try:
-            with Progress(
-                SpinnerColumn(),
-                TextColumn("[progress.description]{task.description}"),
-                BarColumn(),
-                TaskProgressColumn(),
-                console=console,
-            ) as progress:
-                task = progress.add_task("Uploading...", total=len(files))
-                count = s3_ops.upload_dir(
-                    cfg.aws.region,
-                    cfg.aws.profile,
-                    cfg.s3.bucket,
-                    str(path),
-                    prefix,
-                    on_progress=lambda: progress.advance(task),
-                )
-        except Exception as e:
-            console.print(f"[red]Upload failed: {e}[/red]")
-            raise typer.Exit(1)
-
-        console.print(f"[green]✓[/green] Uploaded {count} files to s3://{cfg.s3.bucket}/{prefix.rstrip('/')}/")
-        return
-
-    console.print(f"[red]Path not found: {path}[/red]")
-    raise typer.Exit(1)
+app.add_typer(ai_hub.app, name="ai-hub")

src/qualcomm/aihub_jobs.py

@@ -0,0 +1,114 @@
+from pathlib import Path
+from typing import Any, TypedDict
+
+import qai_hub.hub as hub
+from qai_hub.client import CompileJob, Device, InferenceJob, Model, ProfileJob, QuantizeDtype, QuantizeJob
+
+
+class ModelJobResult(TypedDict):
+    job: CompileJob | QuantizeJob
+    job_id: str
+    model: Model
+    model_id: str
+
+
+class InferenceJobResult(TypedDict):
+    job: InferenceJob
+    job_id: str
+    outputs: Any
+
+
+class ProfileJobResult(TypedDict):
+    job: ProfileJob
+    job_id: str
+
+
+def _dataset_entries(inputs: dict[str, Any]) -> dict[str, list[Any]]:
+    return {name: value if isinstance(value, list) else [value] for name, value in inputs.items()}
+
+
+def submit_compile_job(
+    model: Model,
+    device: Device,
+    input_specs: dict[str, tuple[tuple[int, ...], str]],
+    target_runtime: str,
+    options: str | None = None,
+    job_name: str | None = None,
+) -> ModelJobResult:
+    compile_options = f"--target_runtime {target_runtime}"
+    if options:
+        compile_options = f"{compile_options} {options}"
+
+    job = hub.submit_compile_job(
+        model=model,
+        device=device,
+        name=job_name,
+        input_specs=input_specs,
+        options=compile_options,
+    )
+    target_model = job.get_target_model()
+    if target_model is None:
+        raise RuntimeError(f"Compile job {job.job_id} did not produce a target model.")
+    return {"job": job, "job_id": str(job.job_id), "model": target_model, "model_id": str(target_model.model_id)}
+
+
+def submit_inference_job(
+    model: Model,
+    device: Device,
+    inputs: dict[str, Any],
+    output_dir: str | Path,
+    job_name: str | None = None,
+) -> InferenceJobResult:
+    job = hub.submit_inference_job(
+        model=model,
+        device=device,
+        inputs=_dataset_entries(inputs),
+        name=job_name,
+    )
+    out = Path(output_dir)
+    out.mkdir(parents=True, exist_ok=True)
+    data = job.download_output_data(str(out))
+    return {"job": job, "job_id": str(job.job_id), "outputs": data}
+
+
+def submit_profile_job(
+    model: Model,
+    device: Device,
+    options: str | None = None,
+    job_name: str | None = None,
+) -> ProfileJobResult:
+    job = hub.submit_profile_job(
+        model=model,
+        device=device,
+        name=job_name,
+        options=options or "",
+    )
+    return {"job": job, "job_id": str(job.job_id)}
+
+
+def submit_quantize_job(
+    model: Model,
+    calibration_data: dict[str, Any],
+    options: str | None = None,
+    job_name: str | None = None,
+) -> ModelJobResult:
+    job = hub.submit_quantize_job(
+        model=model,
+        calibration_data=_dataset_entries(calibration_data),
+        weights_dtype=QuantizeDtype.INT8,
+        activations_dtype=QuantizeDtype.INT8,
+        name=job_name,
+        options=options or "",
+    )
+    target_model = job.get_target_model()
+    if target_model is None:
+        raise RuntimeError(f"Quantize job {job.job_id} did not produce a target model.")
+    return {"job": job, "job_id": str(job.job_id), "model": target_model, "model_id": str(target_model.model_id)}
+
+
+def download_model(model_id: str, output_path: str | Path) -> str:
+    dest = Path(output_path)
+    dest.parent.mkdir(parents=True, exist_ok=True)
+    model = hub.get_model(model_id)
+    result = model.download(str(dest))
+    return str(result or dest)

src/qualcomm/artifacts.py

@@ -0,0 +1,83 @@
+import tarfile
+from dataclasses import dataclass
+from pathlib import Path
+
+from src.aws import s3 as s3_ops
+from src.aws import sagemaker as sm_ops
+from src.config import Config
+
+
+@dataclass(frozen=True)
+class ResolvedOnnx:
+    onnx_path: Path
+    model_artifact: str | None
+    run_name: str
+
+
+def _safe_extract(tar: tarfile.TarFile, dest: Path) -> None:
+    dest_root = dest.resolve()
+    for member in tar.getmembers():
+        target = (dest / member.name).resolve()
+        if dest_root != target and dest_root not in target.parents:
+            raise ValueError(f"Unsafe tar member path: {member.name}")
+    tar.extractall(dest, filter="data")
+
+
+def _find_onnx(root: Path, explicit: str | None = None) -> Path:
+    if explicit:
+        p = Path(explicit)
+        if not p.is_absolute():
+            p = root / p
+        if not p.exists():
+            raise FileNotFoundError(f"ONNX file not found: {p}")
+        return p
+
+    matches = sorted(root.rglob("model.onnx"))
+    if not matches:
+        matches = sorted(root.rglob("*.onnx"))
+    if not matches:
+        raise FileNotFoundError(f"No ONNX file found under {root}")
+    if len(matches) > 1:
+        joined = ", ".join(str(p.relative_to(root)) for p in matches)
+        raise ValueError(f"Multiple ONNX files found ({joined}). Pass --onnx-path.")
+    return matches[0]
+
+
+def resolve_onnx(
+    cfg: Config,
+    output_dir: str,
+    from_job: str | None = None,
+    model_s3_uri: str | None = None,
+    onnx_path: str | None = None,
+    last_training_job: str | None = None,
+) -> ResolvedOnnx:
+    if onnx_path:
+        path = Path(onnx_path)
+        if path.exists():
+            return ResolvedOnnx(onnx_path=path, model_artifact=None, run_name=path.stem)
+
+    job = from_job or last_training_job
+    artifact = model_s3_uri
+    if not artifact:
+        if not job:
+            raise ValueError("No model source found. Pass --onnx-path, --model-s3-uri, --from-job, or run training first.")
+        artifact = sm_ops.get_model_artifacts(cfg.aws.region, cfg.aws.profile, job)
+
+    run_name = job or Path(artifact).name.removesuffix(".tar.gz").replace("/", "-")
+    root = Path(output_dir) / run_name / "source"
+    tar_path = root / "model.tar.gz"
+    s3_ops.download_file(cfg.aws.region, cfg.aws.profile, artifact, str(tar_path))
+
+    extract_dir = root / "extracted"
+    extract_dir.mkdir(parents=True, exist_ok=True)
+    try:
+        with tarfile.open(tar_path, "r:gz") as tar:
+            _safe_extract(tar, extract_dir)
+    except tarfile.TarError as e:
+        raise ValueError(f"Invalid model tarball: {tar_path}") from e
+
+    return ResolvedOnnx(
+        onnx_path=_find_onnx(extract_dir, onnx_path),
+        model_artifact=artifact,
+        run_name=run_name,
+    )

src/state.py

@@ -1,30 +1,85 @@
 import json
+from dataclasses import dataclass
 from pathlib import Path
 from typing import Any
 
 STATE_FILE = ".qc-cli.json"
 
 
-def _path(config_dir: str) -> Path:
-    return Path(config_dir) / STATE_FILE
+@dataclass(frozen=True)
+class CliStateStore:
+    config_dir: str = "."
+
+    @property
+    def path(self) -> Path:
+        return Path(self.config_dir) / STATE_FILE
+
+    def read(self) -> dict[str, Any]:
+        if not self.path.exists():
+            return {}
+        with open(self.path) as f:
+            value = json.load(f)
+        return dict(value) if isinstance(value, dict) else {}
+
+    def update(self, **updates: Any) -> None:
+        state = self.read()
+        state.update(updates)
+        self._write(state)
+
+    def get(self, key: str, default: Any = None) -> Any:
+        return self.read().get(key, default)
+
+    def get_last_training_job(self) -> str | None:
+        value = self.get("last_training_job")
+        return str(value) if value else None
+
+    def get_last_model_artifact(self) -> str | None:
+        value = self.get("last_model_artifact")
+        return str(value) if value else None
+
+    def get_last_optimized_model_id(self) -> str | None:
+        value = self.get("last_optimized_model_id")
+        return str(value) if value else None
+
+    def get_last_quantized_model_id(self) -> str | None:
+        value = self.get("last_quantized_model_id")
+        return str(value) if value else None
+
+    def get_last_compiled_model_id(self) -> str | None:
+        value = self.get("last_compiled_model_id")
+        return str(value) if value else None
+
+    def get_last_downloaded_model(self) -> str | None:
+        value = self.get("last_downloaded_model")
+        return str(value) if value else None
+
+    def set_last_training_job(self, job_name: str) -> None:
+        self.update(last_training_job=job_name)
+
+    def get_training_job(self, job_name: str) -> dict[str, Any]:
+        jobs = self._training_jobs(self.read())
+        value = jobs.get(job_name, {})
+        return dict(value) if isinstance(value, dict) else {}
+
+    def update_training_job(self, job_name: str, **updates: Any) -> None:
+        state = self.read()
+        jobs = self._training_jobs(state)
+        jobs[job_name] = {**jobs.get(job_name, {}), **updates}
+        state["training_jobs"] = jobs
+        self._write(state)
+
+    def set_latest_experiment_model_version(self, version: str) -> None:
+        self.update(latest_experiment_model_version=version)
+
+    def _write(self, state: dict[str, Any]) -> None:
+        with open(self.path, "w") as f:
+            json.dump(state, f, indent=2)
+
+    def _training_jobs(self, state: dict[str, Any]) -> dict[str, Any]:
+        value = state.get("training_jobs", {})
+        return dict(value) if isinstance(value, dict) else {}
 
 
-def read_state(config_dir: str = ".") -> dict[str, Any]:
-    path = _path(config_dir)
-    if not path.exists():
-        return {}
-    with open(path) as f:
-        return json.load(f)
-
-
-def write_state(config_dir: str = ".", **updates: str | None) -> None:
-    path = _path(config_dir)
-    state = read_state(config_dir)
-    state.update(updates)
-    with open(path, "w") as f:
-        json.dump(state, f, indent=2)
-
-
-def get_last_training_job(config_dir: str = ".") -> str | None:
-    value = read_state(config_dir).get("last_training_job")
-    return str(value) if value else None
+def store(config_path: str) -> CliStateStore:
+    config_dir = str(Path(config_path).parent)
+    return CliStateStore(config_dir)

src/tracking/__init__.py

@@ -0,0 +1,3 @@
+from src.tracking.mlflow import MlflowTracker, NoopTracker, Tracker
+
+__all__ = ["MlflowTracker", "NoopTracker", "Tracker"]

src/tracking/mlflow.py

@@ -0,0 +1,153 @@
+import os
+from dataclasses import dataclass
+from typing import Any, Protocol
+
+import mlflow
+from mlflow.tracking import MlflowClient
+
+from src.aws import mlflow as aws_mlflow
+from src.config import Config, MlflowMode
+
+
+class Tracker(Protocol):
+    def start_training_run(self, training_job: Any, *, region: str, profile: str, role_arn: str) -> str | None: ...
+
+    def finalize_training_run(self, *, run_id: str | None, training_job_status: Any) -> str | None: ...
+
+
+@dataclass(frozen=True)
+class NoopTracker:
+    def start_training_run(self, training_job: Any, *, region: str, profile: str, role_arn: str) -> str | None:
+        return None
+
+    def finalize_training_run(self, *, run_id: str | None, training_job_status: Any) -> str | None:
+        return None
+
+
+@dataclass(frozen=True)
+class MlflowTracker:
+    tracking_uri: str
+    experiment_name: str
+    registered_model_name: str
+    register_trained_models: bool
+
+    @classmethod
+    def from_config(cls, cfg: Config) -> Tracker:
+        if cfg.mlflow.mode is MlflowMode.disabled:
+            return NoopTracker()
+
+        os.environ.setdefault("MLFLOW_SUPPRESS_PRINTING_URL_TO_STDOUT", "true")
+
+        tracking_server_name = cfg.effective_mlflow_tracking_server_name
+        if not tracking_server_name:
+            raise RuntimeError("MLflow tracking server name could not be resolved.")
+
+        tracking_uri = aws_mlflow.get_tracking_server_arn(
+            cfg.aws.region,
+            cfg.aws.profile,
+            tracking_server_name,
+        )
+        mlflow.set_tracking_uri(tracking_uri)
+        mlflow.set_experiment(cfg.mlflow.experiment_name)
+
+        return cls(
+            tracking_uri=tracking_uri,
+            experiment_name=cfg.mlflow.experiment_name,
+            registered_model_name=cfg.mlflow.registered_model_name,
+            register_trained_models=cfg.mlflow.register_trained_models,
+        )
+
+    def start_training_run(self, training_job: Any, *, region: str, profile: str, role_arn: str) -> str | None:
+        run = mlflow.start_run(run_name=training_job.job_name)
+        run_id = str(run.info.run_id)
+
+        params = {
+            "aws.region": region,
+            "aws.profile": profile,
+            "sagemaker.role_arn": role_arn,
+            "sagemaker.job_name": training_job.job_name,
+            "sagemaker.training_image": training_job.image_uri,
+            "sagemaker.instance_type": training_job.instance_type,
+            "sagemaker.instance_count": training_job.instance_count,
+            "sagemaker.s3_train_uri": training_job.s3_train_uri,
+            "sagemaker.s3_output_path": training_job.s3_output_path,
+            "sagemaker.entry_point": training_job.entry_point,
+            "sagemaker.source_dir": training_job.source_dir,
+        }
+        self._log_params(params)
+        self._log_params({f"hyperparameters.{key}": value for key, value in training_job.hyperparameters.items()})
+        mlflow.set_tags(
+            {
+                "qc_cli.stage": "experiment",
+                "qc_cli.artifact_kind": "trained_source",
+                "qc_cli.source": "sagemaker",
+                "qc_cli.command": "train start",
+                "sagemaker.job_name": training_job.job_name,
+            }
+        )
+        mlflow.end_run()
+        return run_id
+
+    def finalize_training_run(self, *, run_id: str | None, training_job_status: Any) -> str | None:
+        if not run_id:
+            return None
+
+        with mlflow.start_run(run_id=run_id):
+            self._log_params(
+                {
+                    "sagemaker.training_status": training_job_status.status,
+                    "sagemaker.created_at": training_job_status.created,
+                    "sagemaker.modified_at": training_job_status.modified,
+                    "sagemaker.model_artifacts": training_job_status.model_artifacts,
+                    "sagemaker.failure_reason": training_job_status.failure_reason,
+                }
+            )
+            self._log_final_metrics(training_job_status.raw)
+            mlflow.set_tag("qc_cli.command", "train status")
+
+            if training_job_status.status != "Completed" or not training_job_status.model_artifacts:
+                mlflow.set_tag("qc_cli.training_terminal_status", training_job_status.status)
+                return None
+
+            if not self.register_trained_models:
+                return None
+
+            client = MlflowClient()
+            self._ensure_registered_model(client, self.registered_model_name)
+            version = client.create_model_version(
+                name=self.registered_model_name,
+                source=training_job_status.model_artifacts,
+                run_id=run_id,
+                tags={
+                    "qc_cli.stage": "experiment",
+                    "qc_cli.artifact_kind": "trained_source",
+                    "qc_cli.source": "sagemaker",
+                    "sagemaker.job_name": training_job_status.name,
+                },
+            )
+            version_number = str(version.version)
+            client.set_registered_model_alias(self.registered_model_name, "experiment-latest", version_number)
+            mlflow.set_tag("qc_cli.registered_model_name", self.registered_model_name)
+            mlflow.set_tag("qc_cli.registered_model_version", version_number)
+            return version_number
+
+    def _log_params(self, params: dict[str, Any]) -> None:
+        cleaned = {key: str(value) for key, value in params.items() if value is not None}
+        if cleaned:
+            mlflow.log_params(cleaned)
+
+    def _log_final_metrics(self, training_job: dict[str, Any]) -> None:
+        metrics = {}
+        for metric in training_job.get("FinalMetricDataList", []):
+            name = metric.get("MetricName")
+            value = metric.get("Value")
+            if name and value is not None:
+                metrics[str(name)] = float(value)
+        if metrics:
+            mlflow.log_metrics(metrics)
+
+    def _ensure_registered_model(self, client: MlflowClient, name: str) -> None:
+        try:
+            client.get_registered_model(name)
+        except Exception:
+            client.create_registered_model(name)