Skip to content

Latest commit

 

History

History
48 lines (31 loc) · 1.38 KB

File metadata and controls

48 lines (31 loc) · 1.38 KB

TrainFake

中文 README | Documentation index

TrainFake hero

Cinematic AI training logs for demos, tests, teaching, and suspiciously serious terminals.

TrainFake is a preset-first terminal simulator for believable AI training runs. It does not train a real model, download weights, or touch a GPU. It stages the terminal experience around environment scans, warmup, training loops, incidents, recovery logs, eval-like metrics, checkpoints, and summaries.

Quick Start

Python 3.8 or newer is required.

python3 -m pip install .
trainfake presets
trainfake run boss-is-watching --step-delay 0
trainfake tui boss-is-watching

Run a custom YAML script:

trainfake run --config tests/fixtures/sample_run.yaml

What It Includes

  • Built-in scenes for LLM pretraining, Diffusers LoRA, vLLM launch rehearsals, BERT fine-tuning, RLHF reward models, RAG evals, multimodal pretraining, and GPU cluster drills.
  • Classic streaming logs with trainfake run.
  • Split-screen Textual dashboard with trainfake tui.
  • YAML run scripts for custom stages, events, log styles, checkpoint cadence, and reproducible seeds.

Docs

Demo

TrainFake demo

License

MIT