Resource Management
Feature Overview
The Resource Management module provides compute support for the full LLM lifecycle. In the latest navigation, these capabilities are distributed across the Model Inference, Model Training & Evaluation, Development Environment, and Resource Management menus, covering four major capabilities:
| Feature | Description |
|---|---|
| Development Environment (Notebook Instances) | One-click creation of interactive development environments; supports JupyterLab, VS Code, Eclipse Theia |
| Model Inference | Public inference (local deployments and cloud APIs) plus dedicated instances |
| Model Fine-tuning | Customize base models using LLaMA-Factory or MS-Swift frameworks |
| Model Evaluation | Benchmark testing with optional custom framework args, Prompt templates, and scoring plugins |
Development Environment
Development Environment Overview
The Development Environment provides one-click interactive development instances, allowing users to use platform compute resources directly for data analysis, model training, and experimentation.
Create a Development Instance
How to create a development instance on the Industry AI Model Platform, selecting the development environment and compute resources.
Using Notebook Instances
How to use development instances on the Industry AI Model Platform, including launching the Notebook, viewing logs, billing details, and stopping or deleting instances.
Model Inference
Model Inference Overview
The platform provides one-click inference to help users quickly allocate compute and start inference services without complex configuration.
Create Dedicated Inference Instance
How to create a dedicated inference instance for a model on the Industry AI Model Platform.
Using Dedicated Inference Instances
How to use dedicated inference instances on the Industry AI Model Platform, including API calls, Playground testing, real-time monitoring, and billing management.
Model Inference FAQ
Frequently asked questions about dedicated model inference instances.
Public Inference Overview
Overview of public inference on the Industry AI Model Platform: local deployments and cloud API models, and how to discover and call them.
Inference Framework Overview
Overview of the inference frameworks supported by the Industry AI Model Platform, including text generation, image generation, text-to-speech, and video generation frameworks.
Use Public Inference
How to test and call local deployments and cloud API models from public inference.
Manage Public Inference
How platform administrators configure local public inference deployments and cloud API channels (vendor, key, model list, and mapping).
Text Generation
How to use dedicated inference instances for text generation tasks, including conversation completion and API usage.
Text to Image
How to use dedicated inference instances for text-to-image generation tasks, including API usage.
Image Text to Text
How to use dedicated inference instances for image-text-to-text tasks (multimodal understanding), including API usage.
Feature Extraction
How to use dedicated inference instances for text feature extraction (Embeddings), including API usage.
Model Fine-tuning
Model Fine-tuning Overview
The platform provides GPU-accelerated fine-tuning instance hosting, supporting LLaMA-Factory and MS-Swift frameworks.
Create a Fine-tuning Instance
How to create a fine-tuning instance for a model on the Industry AI Model Platform.
Using Fine-tuning and Monitoring
How to use fine-tuning instances on the Industry AI Model Platform for model training, including Web UI operation, Notebook development, training monitoring, and billing management.
Model Fine-tuning FAQ
Frequently asked questions about model fine-tuning.
Export Fine-tuned Models
How to export fine-tuned models on the Industry AI Model Platform using LLaMA-Factory Web UI or MS-Swift CLI.
Fine-tuning Framework Overview
Overview of fine-tuning frameworks supported by the Industry AI Model Platform, including LLaMA-Factory and MS-Swift features, use cases, and comparison.
Model Evaluation
Create Model Evaluation Task
How to create model evaluation tasks on the Industry AI Model Platform using standard benchmark tests to assess model performance.
Using Model Evaluation
How to use the model evaluation feature on the Industry AI Model Platform, including evaluation status descriptions, viewing details, downloading results, and deleting tasks.
Model Evaluation FAQ
Frequently asked questions about model evaluation.
Evaluation Framework Overview
Overview of the three mainstream evaluation frameworks supported by the platform: lm-evaluation-harness, OpenCompass, and EvalScope.
Custom Evaluation Datasets
How to use custom datasets to evaluate model performance on the platform, supporting OpenCompass, EvalScope, and lm-evaluation-harness frameworks.
Evaluation Custom Parameters
Custom parameters on the create-evaluation page: framework args, vLLM args, and advanced Prompt templates and scoring plugins.
Scoring Plugins and Prompt Templates
How advanced options load default prompt templates and scoring plugins from the selected framework and dataset, and when you should override them.
Metrics Configuration
How scoring plugins map to native metrics in OpenCompass, lm-evaluation-harness, and EvalScope, and why Accuracy is not comparable across frameworks.