Tree — 02

objectives/2

distributed/1

memory/1

optimisers/1

Training

Everything between random initialisation and a checkpoint worth keeping. Memory arithmetic that decides what fits, sharding strategies that decide what scales, and the preference objectives that have largely replaced hand-written reward functions.

Five entries. One is a stub and is marked as such — read them as open questions rather than answers.

Entries

Direct Preference Optimisation

Closed-form preference objective that removes the reward model.

Gradient Checkpointing

Trade recompute for activation memory at segment boundaries.

Proximal Policy Optimisation

Clipped surrogate advantage objective with KL control to a reference.

Second-Order Preconditioningstub

Shampoo, Muon and the orthogonalised-update family.

Partition optimiser state, gradients and parameters across ranks.