Build an LLMFrom equations to execution
Demo mode

This public demo has limited functionality.
Run locally for training and inference.

Select Model

Load a checkpoint for inference.

Generation Settings

Sampling controls for text generation.

Demo mode: inference disabled.

Understand

Sampling equations and references.

Equations

Autoregressive generation

ti∼sample(logitsi)t_i \sim \text{sample}(\text{logits}_i)

Temperature

logitsscaled=logitsT\text{logits}_{scaled} = \frac{\text{logits}}{T}

Top-k

Keep the top k logits and set the rest to −∞-\infty.

Top-p

Keep the smallest set of tokens whose cumulative probability exceeds p.

Code Snippets

No snippets available yet.

Output

Model output based on your prompt.

Internals

Attention, logit lens, and layer norms.

Load a checkpoint to view diagnostics.