- Layer 1 now targets experience_action_select (GPU-fused training path)
AND branching_action_select (backtest/fallback), not just the fallback
- Action masking covers both greedy AND random exploration paths
- Exposure index uses parameterized b0_size, not hardcoded 5 or 9
- Fixed end-of-episode variable name (total_bars, not timesteps_per_episode)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Root cause: backtest windows of 300K bars produced ±billions% returns via
multiplicative compounding, and sqrt(252) annualization was wrong for
1-minute bars.
Fixes:
- Window size capped to 10K bars (~25 trading days), evenly distributed
across the full validation set (was clustered in first 6%)
- Annualization: configurable bars_per_day field in GpuBacktestConfig
(default 390.0 for 1-min), produces sqrt(98280) ≈ 313.5
- tanh normalization recalibrated: Sharpe/5, Sortino/8 (was /2, /3)
- CVaR threshold scaled to per-bar: 0.003 with slope 1400 (was 0.05/200)
- VaR/CVaR strided sampling covers full window (was first 4096 only)
- financials.rs + ab_testing.rs: sqrt(252) → sqrt(98280) for consistency
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
RTH/ETH cost differential, market impact (done), book depth fill quality,
macro events, weekend gap risk, margin utilization feature.
All configurable via TOML. We trade against real markets — simulation must match.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
C51 for action selection (expected Q), IQN for risk quantification (CVaR).
Disagreement between C51 and IQN expected Q = uncertainty signal.
Architecture already 90% built — IQN head exists but output unused.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Tests must match reality. All profiles use the full 8-component
composite reward. No legacy PnL-only fallback.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
DSR + normalized PnL + drawdown penalty + idle penalty + regime-adaptive
scaling + asymmetric loss + position-time decay + transaction costs.
All computed per-step in the CUDA kernel. Zero CPU involvement.
12 floats per-episode state, ~25 FLOPs per step, zero extra kernel
launches. 7 new hyperopt dimensions (Phase Fast fixed, Phase Full
searchable). Regime scaling from existing ADX/CUSUM features.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Spec: 3-phase plan to reduce DQN training epoch from 37s to <5s on H100.
- Phase 1: eliminate 300 CPU roundtrips in PER sampling (GPU Philox RNG)
- Phase 2: increase kernel occupancy (batch_size 512+, cuBLAS profiling)
- Phase 3: pipeline overlap (dual-stream experience/training)
Profiling instrumentation added to training loop (init/experience/training/validation breakdown).
RTX 3050 baseline: training=97.2%, experience=2.3% — PER sampling
with CPU RNG + DtoH sync is the primary bottleneck.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Compile+test in same H100 pod (eliminates cross-node PVC transfer)
- New cargo-target-cuda-test PVC (30Gi) — zero contention with training
- onExit notify, podGC, activeDeadlineSeconds, fsGroup, gpu-warmup
- Continue-on-failure with per-model exit code capture
- CUDA_COMPUTE_CAP=90, complete change detection paths
- TEST_DATA_DIR marked as required prerequisite code change
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Spec for Argo WorkflowTemplate that compiles with --features cuda
on CPU node and runs full GPU/CUDA test suite on H100 with real
market data from a dedicated test-data-pvc.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Restore 45-action factored space via Branching DQN (Tavakoli 2018),
outputting 11 Q-values (5+3+3) instead of 45. This was reduced to 5
exposure-only actions during debugging and was never intended as permanent.
- Enable use_branching: true by default in DQNConfig and DQNHyperparameters
- Add branching paths to select_action_with_confidence and select_action_inference
- Update agent.rs select_action_factored for branching-aware selection
- Expand CountBonus to per-branch tracking with bonuses_branched()
- Add order_type + urgency distribution tracking in monitoring
- Add DQN_ORDER_ACTIONS=3, DQN_URGENCY_ACTIONS=3, DQN_TOTAL_ACTIONS=45 to CUDA header
- Fix 7 pre-existing clippy doc_markdown errors in regime_conditional.rs
- Fix pre-existing cognitive_complexity in replay_buffer_type.rs (extract helpers)
- Fix flaky GPU test OOM under parallel execution (CPU fallback + test VRAM safety)
- Delete unused flash_attention submodules (block_sparse, causal_masking, etc.)
- Add GPU hot-path guard scripts and ensemble/hyperopt adapter improvements
Tests: ml-dqn 416/0, ml 905/0, clippy 0 errors on both crates
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Defines architecture for eliminating all 10 GPU→CPU sync barriers from
the training loop via 4 new GPU components: Training Guard (pinned
memory predicates), Q-Value Monitor (on-device accumulator), GPU-resident
action selection, and async experience collector readback.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
