These are historical training campaigns. Some examples use the retired 2B model
or older checkpoint selectors. Before running one on current Flash,
choose a current model, use explicit
/final or /step-N
checkpoint IDs, and validate with flash train config.toml --dry-run. Historical
scores do not transfer to a replacement model without a new evaluation.RESULTS.md.
Explore the flash-example repository
Browse the environments, training configs, frozen evaluations, and the
task-specific numerical record in RESULTS.md.