VISUAL & INTERACTIVE

Visualize

Four interactive labs that turn abstract concepts into concrete visuals. Each is a teaching simulation; production decisions need real measurement.

Teaching simulation
V01

Tokenizer Playground

Splits Turkish and English text into BPE-like pieces. Results are approximate; production decisions need real-tokenizer measurement.

TURKISH10 tokens

Turkish is more expensive in English tokenizers.

Tür#0kçe,#1İngi#2lizce#3toke#4nize#5rlarda#6daha#7paha#8lıdır.#9
ENGLISH12 tokens

Turkish is more expensive in English tokenizers.

Tur#0kish#1is#2more#3expe#4nsive#5in#6Eng#7lish#8tok#9eni#10zers.#11
ComparisonTurkish produces about 0.83× more tokens than its English counterpart. This directly affects the context window and inference cost.
V02

VRAM Budget Visualizer

Separates model, adapter, optimizer, gradient, and activation shares visually. Numbers are approximate; real measurement should use nvidia-smi.

1.9GiB estimate/ 16 GiB
  • Model weights1.86 GiB97%
  • Adapter0.00 GiB0%
  • Optimizer0.00 GiB0%
  • Gradients0.00 GiB0%
  • Activations0.06 GiB3%
Fits the budgetKV cache (inference) estimate: 0.00 GiB. KV cache adds memory at inference time; it is not counted during training.
V03

Training Loss Simulator

Change learning rate, batch, epoch, and overfit risk; watch train/val loss curves live. Models exponential decay + overfit rise; does not replace a real run.

0.51.01.52.02.53.0best step (420)overfit onset (421)train-- val
0.860Final train loss
2.183Final val loss
420Best val step
421Overfit onset
V04

Attention Heatmap

Shows how much each token in a sentence 'attends' to every other as a heatmap. Color intensity = attention weight. This is a simulation; it does not reflect a real model's attention.

The
model
tokenizes
Turkish
into
more
tokens
than
English
The
model
tokenizes
Turkish
into
more
tokens
than
English
·
10
10
5
5
5
5
4
1
12
·
10
8
6
8
6
6
4
10
10
·
10
8
10
7
7
3
8
8
10
·
10
10
10
8
4
6
6
10
10
·
10
10
12
6
6
8
10
10
10
·
10
10
11
5
6
8
10
10
10
·
14
12
4
4
5
5
5
10
10
·
17
2
2
2
2
2
2
3
3
·
Highest attention“English” → “English” (84%). The diagonal (self-attention) usually dominates.