class

MicroGPT::LookaheadModel

Constructors

new(model : MiniGPT, n_ahead : Int32 = 2)

extra_heads

generate(start_ids : Array(Int32), max_tokens : Int32, temperature : Float64 = 1.0) : Array(Int32)

masks

model

n_ahead

param_count

train_step(input_ids : Array(Int32), targets : Array(Array(Int32))) : Tuple(Float64, Array(Float64))

Training step with multiple output heads and zebra masks Returns {avg_loss, per_head_losses}