Dashbord¶

1. Page Overview¶

The navigation page serves as the central data overview interface of the system, providing users with comprehensive monitoring of training and evaluation activities. This page utilizes multi-dimensional data visualization components to display key performance indicators (KPIs) and system usage patterns in real-time.

2. Core Metrics Dashboard¶

2.1 Basic Statistical Indicators¶

Total Fine-tuning Sessions: Cumulative number of model fine-tuning tasks executed
Model Evaluation Count: Total number of completed model performance evaluation and testing tasks
Internal Dataset Count: Total number of proprietary dataset resources managed by the system
Knowledge Base Count: Number of constructed knowledge graph and document library instances
Base Model Count: Total number of available base pre-trained models
Customized Model Count: Number of personalized models that have undergone fine-tuning
Complete Model Count: Number of mature models that have completed the full training pipeline
Live Model Count: Number of models already deployed to the production environment

3. Data Visualization Components¶

3.1 Training Activity Trend Chart¶

Component Type: Time Series Bar Chart

Data Dimension: Past 7 days

Metric Description:

X-axis: Date timestamps (accurate to the day)
Y-axis: Number of fine-tuning training sessions executed per day
Data Points: Daily training task counts

3.2 Base Model Usage Distribution¶

Component Type: Donut Chart

Statistical Dimension: Selection frequency of each base model in fine-tuning tasks

Data Source: Fine-tuning task configuration logs

3.3 Evaluation Mode Distribution Chart¶

Component Type: Proportional Tag Treemap

Visual Form: Multiple colored rectangular blocks tiled in arrangement

Data Encoding:

Rectangle Area: Represents the usage frequency or proportion of the corresponding evaluation mode
Rectangle Color: Distinguishes between different types of evaluation modes (e.g., Judge Mode, Evaluation Mode, Single Model Mode, etc.)
Inline Text: Labels the evaluation type name along with its specific percentage value

Business Value: Intuitively displays the proportion of each evaluation mode, enabling quick identification of the currently predominant evaluation strategies.

3.4 Third-party Model Vendor Analysis¶

Component Type: Sunburst Chart (Multi-layer Pie Chart)

Hierarchical Structure:

Inner Layer: Distribution of model vendors (e.g., Ollama, Silicon Flow, Black Forest Labs, etc.)
Outer Layer: Usage details of specific models under each vendor

Color Coding: Different vendors use contrasting color schemes, while models under the same vendor use gradient shades.