add phase-2 ultra-conservative sweep (J,K,L,M) + waiter that auto-launches after phase 1 from the best ckpt

add 9-config hparam sweep + new_layer_lr_mul param-groups support

fix scheduler bug: don't prepare scheduler with accelerate (was over-stepping cosine by num_processes); add grow40_winning_v2 config

fix OOM: chunked KL with checkpointing + PYTORCH_CUDA_ALLOC_CONF expandable_segments; add kl_chunk_size config key

add grow_layers, sweep configs (replicate_zero4, grow40_winning, grow40_simple), sweep runner

initial scaffold: distill.py + base/zero_14_17 configs + accelerate yaml