← Home
Navigate
👤 Aria
💬 Chat
📊 Dashboard
🛒 Store
Dashboard
›
Training
QAI Training Dashboard
● Connected
🔄 Refresh
Auto (5s)
🌙 Dark
📊 Export
Loading...
📊 Overview
🔧 Jobs
🏆 Models
📚 Datasets
⚙️ Configs
🚀 New Training
⌘
Use Ctrl/Cmd+K to focus. Press Enter to open the best tab match.
Clear
🎮 GPU Status
Loading GPU data...
💻 System Resources
Loading system data...
-
Completed
-
Running
-
Queued
-
Failed
-
Avg Time
-
Best Score
📈 Perplexity Progress
⏱️ Training Duration
🏃 Running
0
No jobs running
✅ Completed
0
No completed jobs
🏆 Top Models
Loading models...
📋 All Training Jobs
🔍
Loading all jobs...
🏆 Model Performance Leaderboard
Loading models...
📚 Available Datasets
Loading datasets...
⚙️ Training Configurations
Loading configs...
🚀 Start New Training Job
Job Name
*
Use lowercase, underscores, no spaces
Model
*
🤖 Phi-3.5-mini-instruct (3.8B params)
🤖 Qwen2.5-3B-Instruct (3B params)
Fast training, good for chat tasks
Dataset
*
⏳ Loading datasets...
Select a dataset
Epochs
More epochs = better learning
Max Train Samples
-1 for all samples
Learning Rate
Default: 2e-4
🔧 Advanced Options
▼
Batch Size
1 (Slow, less memory)
2 (Balanced)
4 (Fast, more memory)
8 (Very fast, high memory)
Gradient Accumulation
Simulate larger batches
Warmup Steps
LR warmup period
LoRA Rank
Higher = more parameters
LoRA Alpha
Usually 2x rank
LoRA Dropout
0.0 to 0.5 typical
Weight Decay
L2 regularization
Max Grad Norm
Gradient clipping
Random Seed
For reproducibility
Enable Evaluation
Max Eval Samples
Eval Steps
⏱️ Estimated Time:
~5 minutes
💾 Est. VRAM:
~4GB
🚀 Start Training
💾 Save Config
📂 Load Config
🔄 Reset
⚡ Quick Presets:
⚡ Quick Test (1 epoch, 100 samples)
🧪 Tuning Wizard
📊 Standard (3 epochs, 1k samples)
🏆 Full Training (5 epochs, all samples)
🚀 Production (10 epochs, all samples)
⌨️ Shortcuts
⌨️ Keyboard Shortcuts
Toggle Auto-Refresh
A
Toggle Dark Mode
D
Refresh Data
R
Export Report
E
Overview Tab
1
Jobs Tab
2
Models Tab
3
Search Jobs
/
Command Bar
Ctrl/Cmd+K
Close Modal
ESC
Close