Alpha Style Guide
Comprehensive UI reference for the Alpha GPT training system.
Buttons
Variants
Sizes (Comparative)
sm (h-7)
md (h-9)
lg (h-11)
icon
States
Forms & Inputs
Text Inputs
Select Inputs
Charts & Data
0.4521
Loss?
12.4k
Throughput?
tokens/sec
14.2 GB
VRAM
85% utilized
42.1%
MFU?
Sparklines
Active Training
Validation Loss
Detail Table
Optimizer?AdamW
Learning Rate6.0e-4
Batch Size512
Complex Interactive Charts
Interactive Loss Chart
?
?
?
?
Training Analytics
Charts available for every training run. These visualize throughput, timing, gradient health, perplexity, and convergence dynamics.
Throughput (tok/s)
Step Time (ms/iter)
Perplexity
Train/Val Gap
Gradient Clipping
GPU Operations
Smoothed Loss (EMA)
Loss Velocity
Timing Phase Breakdown
Backward / Forward Ratio
Step Time Breakdown (Stacked)
Forward
Backward
Grad Norm
Optimizer
GPU Sync
Data
GPU & VRAM
Learning Rate Schedule
Gradient Norm
Evolutionary Analytics
Evolutionary Analysis (Symbiogenesis)
4.93
Wt Entropy
bits
28.5
Eff. Rank
-1.8800
Free Energy
1.200
Pop Entropy
nats
0.4000
Complexity
1.7950
Fitness
0
CUSUM
alerts
-
Batch Size
adaptive
CUSUM Statistical Monitors
Information Bottleneck (MI)
MI Analysis Pending
Transformer Layer Analysis
Transformer Layer Analysis
Gradient Norm Heatmap
Per-Layer Gradient Evolution
Feedback & States
Loading Spinners
sm
md
lg
Empty State
No training runs found
Sync runs from disk or start a new training job to see data here.
Badges & Status
Status & Categorization
Default
Secondary
Outline
Completed
Stale
Failed
Active
Documentation & Architecture
API Documentation
GETPOST
/v1/modelsSome Fieldrequired
{
"model": "alpha-1"
}Architecture Elements
CPUGPUDispatch
ForwardBackward
gemm_coop
Cooperative Tensor Core Kernel
Typography & Colors
Theme Palette
accent
green
yellow
red
bg
surface
surface-2
border
Text Hierarchy
Heading 1
3xl / bold / text-primaryHeading 2
xl / semibold / text-primaryStandard paragraph text.
base / text-secondaryLabel Text
sm / mono / uppercase / text-muted