Bring your own agent, or start a route yourself. Either way the work lands on a branch of its own.
Train GPT-2 small to a fixed validation loss in as little time as possible (placeholder)
FP8 matmul for the language-model head.
A sparser attention schedule.
turing