jgrusewski f1e53797ca docs: H100 epoch optimization v2 spec — 47s → 19s (2.5x speedup) ...

9 optimizations: single-copy weight sync, graph update vs re-capture,
async validation, aux GEMM fusion, conditional frequency, target sync,
mega-graph, batch_size 16K, validation subsampling.

50 trials × 40 epochs: 26hr → 10.6hr ($45/run savings at $3/hr)

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

2026-04-08 01:49:01 +02:00

..

infra

feat(infra): GPU training launcher with local/cloud routing

2026-02-23 11:19:58 +01:00

plans

feat(dqn): enable Branching DQN with 45 factored actions (5×3×3)

2026-03-11 22:00:13 +01:00

superpowers

docs: H100 epoch optimization v2 spec — 47s → 19s (2.5x speedup)

2026-04-08 01:49:01 +02:00