New training wizard
Screen-by-screen description of the 11 steps that configure a training run, from task to review and start
Required permission: DevelopDEVELOPER · OPERATOR · ADMIN
This screen configures a training run in 11 steps and starts it. You enter it with New training (새 학습) on the Training list, and it needs the develop permission (DEVELOPER, OPERATOR, ADMIN). If you enter with Train with this dataset (이 데이터셋으로 학습) on a dataset detail, ④ Dataset is preselected; if you enter with Train again with these settings (이 설정으로 다시 학습) on a training detail, all values are filled in and the ⑪ Review step opens directly.
| # | Area | Description |
|---|---|---|
| ① | Step list | 11 steps. You can click back only to steps you have already passed, and a step the server returned an error for gets a red mark. |
| ② | Current step | Step title and description, input area. |
| ③ | Next (다음) | Enabled once this step's required values are filled in. On the last step it becomes Start training (학습 시작). |
| ④ | Cancel (취소) | Discards your input and returns to the list (asks for confirmation once). |
Changing an earlier step clears the values of later steps that depend on it. For example, changing the task resets the framework, model, classes and hyperparameters; changing the dataset resets the classes.
Step by step
① Task
Choose the computer vision task to train. Object Detection is selected at first. The top right of each card shows how many frameworks are available; a task with no framework is greyed out and cannot be chosen.
| Task | Description |
|---|---|
| Object Detection | Finds objects with bounding boxes and attaches class labels. |
| Instance Segmentation | Also predicts a pixel mask per object. |
| Semantic Segmentation | Paints every pixel (every point, for point clouds) by category. |
| Classification | Classifies one input into one of a fixed set of classes. |
② Framework
Cards for the frameworks that support the task. Each card shows the license, number of models and additional export formats (for example onnx, tensorrt). The default for Object Detection is YOLO.
③ Model
Choose a model variant of the framework (versions with different size and accuracy). In the table, Params(M) is the parameter count, mAP is the accuracy on a public benchmark, and pretrained is the catalog checkpoint. Once you choose a variant, Starting weights (시작 가중치) appears below, where you choose the catalog weights or weights you uploaded.
④ Dataset
Only datasets with files usable for training (images, PLY, tabular) are shown. If you turn on Use dataset cache (데이터셋 캐시 사용) below, the download is skipped when you train again with the same dataset.
⑤ Classes
Shows the classes collected from the dataset's annotations, with object counts and shape types. Depending on the model variant, classes are Required (필수) (choose at least one; all compatible classes are preselected), Optional (선택) (choose none to train all) or Not applicable (해당 없음) (there is nothing to choose). Classes with shapes the model cannot handle are greyed out with the reason.
⑥ Train/Val split
Set the split method (분할 방식) (Random for ratio-based random, Stratified to keep label ratios, and so on — only the ones the framework supports are enabled), the Train ratio (Train 비율) (default 80%) and the Seed (the same value reproduces the same split).
⑦ Hyperparameters
The model variant's default hyperparameters are filled in as YAML. Your edits are saved as they are, and the server validates the allowed keys and ranges when you submit. Revert with Reset to defaults (기본값으로 재설정). Some frameworks show input fields instead of YAML, with the range and default written under each field.
⑧ Evaluation · ⑨ Export
| Step | Item | Description |
|---|---|---|
| ⑧ Evaluation | Run model evaluation | After training, computes metrics (mAP and so on) on the validation set. On by default. |
| ⑧ Evaluation | Speed test | Measures inference speed (in preparation — only the setting is saved). |
| ⑨ Export | PyTorch (.pt) | Always saved. |
| ⑨ Export | ONNX (.onnx) | After training, converts to ONNX and uploads it as a run artifact. |
| ⑨ Export | TensorRT (.engine) | Depends on the GPU and driver environment. Off for frameworks that do not support it. |
⑩ Execution environment
Training runs as a container (Kubernetes Job). Choose the training image under Training runtime variant (학습 런타임 변형); leave it empty to use the tenant's default variant. Under GPU resources (GPU 자원), decide whether to run on CPU or how many wired GPU slots to use. If no GPU is wired to the tenant, a warning appears and training runs on CPU only.
⑪ Review and start
| # | Area | Description |
|---|---|---|
| ① | Experiment name (optional) | Runs submitted under the same name are grouped as runs under one MLflow experiment, which makes them easy to compare. If left empty, it is generated as {preset or model}-{date-time}. |
| ② | Register in model registry (모델 레지스트리 등록) | When on, registers best.pt as a new model version after training. When on, you must enter the Registered model name (등록 모델 이름). |
| ③ | Start training (학습 시작) | Submits. On success, you move to the training run detail. |
The summary below lets you check the values of every step at once.
When the submission is refused
If the server finds a problem when validating the values, a "Check the settings (설정을 확인하세요)" notice appears, the wizard returns to the first step with a problem, and that step lists the errors in red.