huawei-cloud-modelarts-training-management

Huawei Cloud ModelArts Training Management

Full lifecycle management for ModelArts training jobs and related resources via hcloud CLI — 52 API interfaces across 8 functional domains.

---

Overview

This skill enables users to manage Huawei Cloud ModelArts training jobs and related resources through the hcloud CLI. It covers the complete training lifecycle including training job CRUD, algorithm management, training experiments, auto search (hyperparameter tuning), model import, and training image save.

Architecture

User Request → Agent → hcloud ModelArts  --cli-region={region} [--params] → Huawei Cloud ModelArts API

Applicable Scenarios

• Training Job Management: Create, list, show, stop, delete training jobs; query logs, metrics, engines, flavors, quotas
• Algorithm Management: Create, list, show, update, delete training algorithms; publish to gallery
• Training Job Tags: Create, show, delete tags for training jobs
• Training Experiments: Create, list, show, update, delete, check training experiments
• Training Job Events: Query job events, stages, tasks; list system events and scheduled events
• Model Import: Import, list, show, delete AI models; create ModelArts agency
• Auto Search: Query hyperparameter search trials, parameters analysis, yaml templates
• Training Image Save: Create and query training job image save tasks

---

Prerequisites

1. hcloud CLI installed and authenticated — Reference: https://support.huaweicloud.com/qs-hcli/hcli_02_003.html
2. Huawei Cloud AK/SK configured via hcloud or environment variables
3. ModelArts service enabled in the target region
4. IAM permissions — See references/iam-policies.md

---

Workflow

Step 1: Identify the Operation

Determine which training operation the user needs based on their request:

Step 2: Execute CLI Command

hcloud ModelArts  --cli-region={region} [--key=value ...]

{region} must be replaced with the actual region, e.g., cn-north-4. Do NOT hardcode the region.

Step 3: Handle Write Operations

For all write operations (Create/Update/Delete/Stop/Change/Patch/Notify/Accept/Batch), prompt the user for confirmation before execution.

---

KooCLI Command Format Standard

hcloud ModelArts  --cli-region={region} [--key=value ...]

Note: --project_id is auto-resolved from authentication credentials if omitted. Include it explicitly only when targeting a specific project.

Complex parameters: For complex nested parameters (e.g., config, spec), use --cli-jsonInput=/path/to/file.json. The JSON file must wrap body in {"body": {...}} envelope.

---

Core Commands

All 52 CLI command examples across 8 functional domains are documented in a separate reference file.

📖 For detailed command syntax, parameters, and examples, read references/cli-command-examples.md

Quick Index

When executing any command, always refer to the reference file for exact parameter names, required/optional flags, and usage patterns.

---

Parameter Confirmation

---

SDK Fallback

If a CLI operation fails due to a CLI bug, fall back to SDK:

from huaweicloudsdkcore.auth.credentials import BasicCredentials
from huaweicloudsdkmodelarts.v2.modelarts_client import ModelArtsClient
from huaweicloudsdkmodelarts.v2.region.modelarts_region import ModelArtsRegion

credentials = BasicCredentials(ak="{AK}", sk="{SK}", project_id="{project_id}")
client = ModelArtsClient.new_builder() \
    .with_credentials(credentials) \
    .with_region(ModelArtsRegion.value_of("{region}")) \
    .build()

---

Reference Documents

---

Known Issues Summary

• Complex nested params: Use --cli-jsonInput with JSON file for CreateTrainingJob, CreateAlgorithm, CreateModel, etc.
• --cli-jsonInput syntax: Use file path directly (no @ prefix), JSON must be wrapped in {"body": {...}}
• Training job logs: ShowObsUrlOfTrainingJobLogs returns a temporary OBS URL (valid for 5 minutes)
• StopTrainingJob: Can only stop jobs in creating, waiting, or running state
• Auto search: Trial early stop only works on running trials

See references/known-issues.md for full details.

---

Notes

• All write operations (Create/Update/Delete/Stop/Change/Patch/Notify/Accept) require user confirmation before execution
• Region is not hardcoded — uses {region} placeholder
• project_id is auto-resolved when omitted
• No hardcoded AK/SK in any file — credentials read from environment variables
• SDK fallback available when CLI encounters bugs
• Complex nested parameters use --cli-jsonInput with JSON file