foxhunt

Files

jgrusewski c071489979 infra(smoke): --max-bars cap for train_baseline_rl + multi_fold smoke

Adds an optional --max-bars CLI cap to `examples/train_baseline_rl.rs`.
When >0, truncates the fxcache-loaded features/targets/OFI/timestamps in
lockstep per the data_loading.rs precedent (commit 2ac956298 OFI/bars
desync fix), then rebuilds FxCacheData with the capped bar_count. 0 =
unlimited, matching pre-change production behaviour.

Wires --max-bars=500000 into the multi_fold_convergence smoke test. Full
ES.FUT fxcache is ~697,732 bars (~290 MB on GPU with OFI per the
session_2026-04-20 memory note); the smoke's 3-fold × (6+2+2+2×step =
14-month total) walk-forward span needs ~350k bars at 1-min resolution,
so 500k gives ~43% headroom while freeing ~81 MB of VRAM — enough to
unblock RTX 3050 Ti 4 GB local runs that OOMed before.

Applied after fxcache load and before WalkForwardConfig so downstream
fold-range generation sees the capped range. Zero effect on production
training (default 0 = unlimited).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

2026-04-22 22:23:29 +02:00

baseline_common

fix(ml): wire spread_cost_bps into RL training — commission + spread = total tx cost

2026-03-01 22:10:43 +01:00

cuda_test.rs

fix: wire real GPU backprop + SPSA gradients, fix checkpoint loading, eliminate candle from examples

2026-03-20 15:41:42 +01:00

download_baseline.rs

fix: add retry with backoff to Databento download_quarter

2026-03-31 22:57:43 +02:00

evaluate_baseline.rs

test(policy-quality): Task 0.16 surrogate_noise_check smoke (mandatory gate)