- init_observability now reads LOG_DIR env var for log directory (falls back
to relative logs/ path). Gracefully disables file logging if directory
cannot be created instead of failing the entire observability init.
- Add ServiceAccount, Role, and RoleBinding for ml-training-service to
create/get/list/watch/delete batch/v1 Jobs (K8s training dispatch).
- Add LOG_DIR=/app/logs env to ml-training-service, trading-service, and
backtesting-service deployment YAMLs.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Single model_memory_mb field now serves as the sole source of truth for
base model memory. Removes the redundant base_model_memory_mb field,
legacy dual-path branches in calculate_optimal_batch_size() and
max_safe_batch_size(), and the associated legacy test.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
AutoBatchSizer was computing an "optimal" batch_size (128 on L4) and
overriding the PSO-chosen value. On an L4 24GB this meant 3.5% VRAM
utilization per trial. Now the trainer only clamps if the batch_size
would actually OOM, letting PSO explore the full [64, 4096] range.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The GPU experience collector only supported DQNAgentType::Standard, but
enable_regime_qnetwork defaults to true, creating RegimeConditional
agents. This silently fell back to CPU with a debug! message invisible
at INFO log level.
- Add primary_head() accessor to RegimeConditionalDQN (returns trending head)
- Extract weights from primary head for both collector init and weight sync
- Upgrade debug! to warn! for missing GPU collection prerequisites
- Add warn! to PPO for silent GPU skip when raw market data not uploaded
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
These files were part of the Phase 2 cuda_pipeline but were untracked
and not included in previous commits. Required for compilation with
--features ml/cuda.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Phase 3 of the CUDA pipeline: both trainers now take a GPU-first path
for experience collection (128×500 = 64K experiences per kernel launch,
zero CPU-GPU roundtrips per timestep) with automatic CPU fallback.
- DQN: upload features alongside targets, GPU collection branch before
CPU loop, gpu_batch_to_experiences() conversion into replay buffer
- PPO: set_raw_market_data() for CudaSlice upload, GPU collection
branch bypasses collect_rollouts + prepare_training_batch entirely,
gpu_batch_to_trajectory_batch() conversion with in-kernel GAE
- Configurable GPU batch sizes (gpu_n_episodes, gpu_timesteps_per_episode)
and trading params (initial_capital, avg_spread) via hyperparameters
- SAFETY training diagnostics downgraded from warn! to debug!
- 4 new batch index-math validation tests in cuda_pipeline
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Move 10 shared device functions (gpu_random, leaky_relu, matvec_leaky_relu,
action_to_exposure, action_to_tx_cost, barrier_init/check/reset,
diversity_entropy, curiosity_inference) and shared constants from
dqn_experience_kernel.cu into a new common_device_functions.cuh header.
The DQN kernel now expects the common header to be prepended via NVRTC
source concatenation at compile time. DQN-specific functions
(q_forward_dueling, argmax_q) remain in the kernel file. This enables
the upcoming PPO kernel to reuse the same shared functions.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Adds GpuExperienceCollector field to DQNTrainer, initializes it when
dueling networks and curiosity module are present on CUDA device, and
syncs GPU weight copies after each training epoch.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Three fixes from spec review:
1. diversity_entropy returns -0.1 penalty (threshold < 1.0) instead of raw entropy
2. barrier_done included in episode termination (barrier hit ends episode)
3. step_in_episode counter resets properly on mid-loop episode reset
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace raw f64 price/quantity in SmallBatchOptimizer with typed
financial values. Conversion to f64 happens at the SIMD boundary
(_mm256_loadu_pd) only. Delete 3 orphaned dead-code files:
financial_safe.rs, simd_optimizations.rs, tests/financial_tests.rs.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Aliased re-exports avoid collision with existing types::Price/Quantity/Money
during migration. Will rename to canonical names in Phase 6 cleanup.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Ratio arithmetic (Add, Sub, Mul, Div) now clamps overflow to
f64::MAX/f64::MIN instead of producing Inf from finite inputs
- Cross-type Price*Quantity→Money and Money/Quantity→Price use
saturating i128→i64 conversion via clamp instead of silent truncation
- Added clamp_finite() and saturating_i128_to_i64() helper functions
- Added 9 edge-case tests: 89/89 passing, clippy clean
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Add Serialize/Deserialize to all four financial types
- Ratio::ln returns 0.0 for non-positive inputs (prevents NaN leak)
- Ratio::exp clamps to f64::MAX on overflow (prevents Inf leak)
- Ratio::powi clamps to f64::MAX/MIN on overflow
- Add 4 invariant-preservation tests (80 total)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace all `.to_string().parse::<f32/f64>()` patterns with
`num_traits::ToPrimitive` methods (`.to_f32()`, `.to_f64()`).
Each string roundtrip heap-allocated per conversion — fatal in
DQN hot loop (300K+ bars × epochs). Decimal stays as canonical
financial type; conversions happen at GPU/float boundaries only.
Also fixes blocking_read() in async context (risk_integration.rs).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
P2-C reversal/cash warns fire thousands of times per hyperopt trial
during backtest simulation. Also removes stale Kelly patch file (already
applied).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
These fire every training step when portfolio isn't populated yet —
extremely noisy during hyperopt with parallel trials.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The previous "auto-detect" logic forced CPU which was wrong — GPU is
faster for forward/backward even with parallel trials (DQN/PPO use
<350MB of 24GB VRAM across 7 threads).
Changes:
- Remove --device flag, always require CUDA GPU
- Add DQNTrainer::new_with_device() to share CUDA context across trials
- Propagate hyperopt device to internal DQN trainer (was ignoring it)
- PPO/DQN adapters error on missing GPU instead of silent CPU fallback
- Downgrade batch-size clamping from warn to debug
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
DQN/PPO networks are tiny (3 layers × 128 neurons). Running parallel
hyperopt on GPU wastes cores because CUDA context serializes across
threads — 5 trials on L4 only used 2000m of 6000m requested CPU.
Changes:
- Add --device flag to hyperopt_baseline_rl (auto/cpu/cuda)
- Auto mode forces CPU for parallel runs (no CUDA contention)
- CPU mode uses all available cores (no 2-core reserve)
- Add with_device() builder to DQN/PPO hyperopt trainers
- Downgrade "portfolio value <= 0" and GPU utilization warnings
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Portfolio value <= 0 during early exploration is expected behavior,
not a warning. GPU memory utilization advisory is informational.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Enable concurrent trial evaluation for DQN/PPO hyperparameter
optimization via clone-per-particle pattern — each PSO particle
clones the trainer and trains independently, replacing the previous
Arc<Mutex> serialization bottleneck. On L4 (8 vCPU) this yields
~4-5x throughput improvement.
Changes:
- DQNTrainer/PPOTrainer: Clone with Arc<AtomicUsize> trial counter
- DQNTrainer: replace unsafe mutable aliasing with Arc<Mutex> for
best_trial tracking
- ArgminOptimizer: add optimize_parallel() with ParallelObjectiveFunction
and scoped-thread LHS evaluation
- CLI: --parallel 0 (auto-detect CPUs-2), --initial-capital 35000,
--tx-cost-bps 0.1 (IBKR ES all-in)
- CI: both hyperopt jobs use --parallel 0 + IBKR ES cost defaults
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
train_from_parquet and load_training_data_from_parquet had zero
callers — TFT hyperopt uses train_from_bars via DBN data.
Also removes unused NormalizationParams struct from this module.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Feature vectors are [f64; 51] after WAVE 10 (Proxy OFI removal):
[0..5] static, [5..15] known, [15..51] unknown (36 features)
Old code used fv[15..54].min(fv.len()) which silently produced 36
features per step but expected 39 in Array2 shape → ShapeError.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
small_batch_ring::test_performance_characteristics asserts <500ns
latency which is unreliable on shared L4 nodes running parallel jobs.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Three critical stability fixes for PPO training:
1. Gradient clipping: Added clip_grads() that scales gradients by
max_norm/norm when L2 norm exceeds max_grad_norm (0.5). Previously
the code only logged a warning — gradient norms of 27-171x above
the threshold were applied raw to weights, causing divergence.
Fixed in all 8 locations across PpoTrainer (6) and ContinuousPPO (2).
2. Return normalization: Value loss now normalizes returns to N(0,1)
before computing MSE. Raw cumulative returns (±1000s) caused enormous
value gradients that destabilized the critic network.
3. Hyperopt search space: Tightened value LR upper bound from 1e-3 to
1e-4 (1e-3 is documented unstable), policy LR from 1e-3 to 3e-4.
Also fixed LSTM path where optimizer.step() ran BEFORE gradient norm
check — now clip → step (not step → warn).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The adaptive-strategy crate (~28K lines, 22 source files) was an orphaned
framework with zero external consumers. Its only valuable piece — ensemble
uncertainty quantification — has been ported to ml/src/ensemble/confidence.rs.
Ported: ConfidenceAggregator, UncertaintyQuantifier, ReliabilityScorer,
IntervalCombiner, DisagreementTracker + all config/output types. Removed
gratuitous async from pure-math methods. 6 tests (4 ported + 2 edge cases).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Extend hyperopt infrastructure to support all 10 ML models (DQN, PPO +
8 supervised). Previously only TFT and Mamba2 had hyperopt trainers.
- Add HyperparameterOptimizable impl for Liquid, TGGN, TLOB, KAN, xLSTM, Diffusion
- Create shared_data.rs with common data prep utilities (build_flat_pairs,
build_sequence_pairs, write_trial_result_json)
- Extend hyperopt_baseline_supervised binary to dispatch all 8 models
(individual, "both" for tft+mamba2, "all" for all 8)
- Add CI jobs: 7 train-validate + 10 hyperopt jobs for all models
- Fix DiffusionMetrics NaN default, XLSTMMetrics serde, safe indexing
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
evaluate_supervised looked for checkpoints at {models_dir}/{model}_fold{N}_best
but training saves to {models_dir}/{model}/{model}_fold{N}_best (model subdirectory).
Also saves NormStats JSON per fold during training so evaluation uses
training-time normalization instead of computing from test data (data leakage).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
