jgrusewski
8c7e907fcf
perf(cuda): eliminate GPU→CPU roundtrips from backtest evaluate loop
- gather_states: replace memcpy_dtoh + Tensor::from_vec with DtoD copy
(cuMemcpyDtoDAsync) — state tensor stays on device, zero CPU touch
- actions: replace to_vec1 + memcpy_htod with DtoD copy from argmax
tensor directly into actions_buf — eliminates per-step PCIe upload
- batch_q_values (RegimeConditionalDQN): replace CPU-side regime
classification (to_vec2 + serial loop + sub-batch re-upload) with
on-device classify_regime_masks_gpu + all-heads forward + masked blend
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-03-11 13:05:02 +01:00
..
2026-03-01 22:47:39 +01:00
2026-03-11 08:46:58 +01:00
2026-03-05 00:22:04 +01:00
2026-03-01 22:47:39 +01:00
2026-03-07 23:28:18 +01:00
2026-03-01 22:47:39 +01:00
2026-03-01 22:47:39 +01:00
2026-03-11 13:05:02 +01:00
2026-03-10 11:17:51 +01:00
2026-03-10 13:18:57 +01:00
2026-03-10 13:18:57 +01:00
2026-03-11 00:46:53 +01:00
2026-03-01 22:47:39 +01:00
2026-03-10 13:18:57 +01:00
2026-03-11 13:05:02 +01:00
2026-03-10 13:18:57 +01:00
2026-03-10 13:18:57 +01:00
2026-03-10 13:18:57 +01:00
2026-03-11 00:46:53 +01:00
2026-03-10 11:51:31 +01:00
2026-03-10 11:51:31 +01:00
2026-03-08 15:17:22 +01:00
2026-03-10 13:18:57 +01:00
2026-03-10 12:12:14 +01:00
2026-03-10 13:18:57 +01:00
2026-03-08 15:17:22 +01:00
2026-03-10 11:17:51 +01:00
2026-03-10 13:18:57 +01:00
2026-03-10 13:18:57 +01:00
2026-03-10 13:18:57 +01:00
2026-03-10 13:18:57 +01:00
2026-03-01 22:47:39 +01:00
2026-03-05 23:19:38 +01:00
2026-03-01 22:47:39 +01:00
2026-03-01 22:47:39 +01:00
2026-03-05 00:18:58 +01:00
2026-03-01 22:47:39 +01:00
2026-03-03 22:16:35 +01:00