jgrusewski
9843de5e3d
feat(sp14): B.7 — trainer struct fields for EGF kernels + concat scratch
Adds 4 CudaFunction handles + 1 scratch buffer to GpuDqnTrainer:
- sp14_q_disagreement_update_kernel
- sp14_alpha_grad_compute_kernel
- sp14_gradient_hack_detect_kernel
- sp14_dir_concat_qaux_kernel
- sp14_dir_qaux_concat_scratch: CudaSlice<f32> [B * (SH2 + 1)]
All loaded from precompiled cubins in trainer construction, mirroring
the SP13 aux_pred_to_isv_tanh / aux_sign_label kernel-loading pattern.
Per feedback_no_partial_refactor: handles + scratch are held but not
yet wired in. Subsequent tasks (B.9+) launch them.
Build: cargo check --workspace clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-05 19:41:12 +02:00
..
2026-02-23 11:19:58 +01:00
2026-03-11 22:00:13 +01:00
2026-05-05 17:36:56 +02:00
2026-04-30 17:26:25 +02:00
2026-05-02 21:13:53 +02:00
2026-04-27 22:19:41 +02:00
2026-05-05 19:41:12 +02:00
2026-04-25 12:29:47 +02:00
2026-04-28 23:54:45 +02:00
2026-05-05 18:51:00 +02:00
2026-04-28 09:23:32 +02:00
2026-04-26 10:36:03 +02:00