Add DeviceConfig and GpuCapabilities re-exports to ml::prelude so
downstream crates can import GPU management types via a single
`use ml::prelude::*` statement.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace hardcoded max_batch_size: 230 in CampaignConfig::dqn_default()
and ppo_default() with dynamic GPU detection via GpuCapabilities::detect()
and resolve_batch_size(). Update test to no longer assert <= 230.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace Device::cuda_if_available(0).unwrap_or(Device::Cpu) with
DeviceConfig::Auto.resolve().unwrap_or(Device::Cpu) in DQN, PPO, TFT,
and Mamba2 ensemble inference adapters so device selection goes through
the centralized GPU detection module.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace local DeviceConfig enum definition in liquid/candle_cfc.rs with
a re-export from the central crate::gpu::DeviceConfig, eliminating
duplication while preserving all existing API and tests.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
8 bite-sized tasks with TDD approach, exact file paths, and complete code.
Covers: DeviceConfig relocation, GpuCapabilities detection,
ModelMemoryEstimate per-model profiles, and wiring into PPO trainer
and hyperopt campaign.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Centralizes GPU device selection, capability detection, and per-model
batch size optimization to replace hardcoded RTX 3050 Ti limits.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- NoiseScheduler: precomputed cosine/linear alpha_bar schedules
- Denoiser: FC network with sinusoidal time embedding + SiLU + residual
- DDIMSampler: deterministic fast sampling (10 steps from 1000 timesteps)
- DiffusionTrainableAdapter: UnifiedTrainable for unified training pipeline
- Hyperopt adapter with ParameterSpace (9 params, batch ≤64 for 4GB GPU)
- ModelType::Diffusion registered in common + coordinator
- 41 tests passing (config=3, noise=7, denoiser=4, sampler=5, trainable=12, hyperopt=7)
- OOM-safe: FC denoiser instead of U-Net, small hidden dims, conservative defaults
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Phase 2-3 of ensemble expansion:
- KAN (Kolmogorov-Arnold Network): B-spline basis, layer, network, trainable adapter
- TLOB UnifiedTrainable adapter with 3D input support (batch, seq, features)
- Hyperopt adapters for both KAN and TLOB (ParameterSpace + metrics)
- ModelType::KAN variant registered in common, coordinator, lib.rs
- 44 new tests, all passing, zero warnings
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Closes the integration gap between Liquid CfC and DQN/PPO by adding:
- LiquidTrainer with gRPC progress callbacks, early stopping, and
checkpoint management (ml/src/trainers/liquid.rs)
- LiquidModel wrapper in enhanced_ml.rs for hot-loading from safetensors
- Ensemble weight rebalance: DQN 0.25, PPO 0.25, TFT 0.20, Mamba2 0.15,
Liquid-CfC 0.15
Verified: 81 liquid unit tests + 3 integration tests + 211 trading_service
tests pass, 0 compile errors across workspace.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Implements TGGNTrainableAdapter wrapping a candle-based projection
network (input→hidden→ReLU→output) with full UnifiedTrainable interface
including checkpoint save/load, validation, and gradient tracking.
11 tests covering forward, backward, optimizer step, checkpoint
roundtrip, validation, learning rate, and metrics collection.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Add load_from_safetensors() to DQNAgent for weight loading via VarMap
- Update RealDQNModel::from_checkpoint to try safetensors first, fall back to JSON
- Replace std::mem::forget(prediction_shutdown_tx) with proper Vec-based storage
that sends shutdown signal and drops senders during graceful shutdown
- Wire order_manager.get_open_orders() into emergency_stop response so callers
see which orders were active when the kill switch engaged
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace hardcoded position_size: 1000.0 with actual position quantities
from fetch_positions(). Compute contribution_pct as marginal VaR ratio
(symbol_var / portfolio_var * 100) instead of naive equal-weight split.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace 7 hardcoded 0.0 values with real calculations:
- target_quantity from last close price
- portfolio_volatility from log return stddev * sqrt(252)
- portfolio_sharpe from weighted returns / portfolio vol
- var_95 parametric VaR
- max_drawdown_estimate from vol approximation
- rebalance_delta as target - current (0 until positions available)
- per-asset volatility from price bars (was hardcoded 0.15)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Tasks 8-14 production hardening batch:
- Populate Order/Position/Execution proto messages from JSON payload in event
stream converters instead of returning None (Task 9)
- Compute max_drawdown from cumulative PnL samples in A/B testing pipeline
instead of hardcoded 0.0 (Task 11)
- Document feature pipeline integration blockers with detailed roadmap
comments in state.rs and trading.rs (Task 8)
- Document realized PnL gap: TradingPosition lacks the field, repository
has async method incompatible with Iterator::map (Task 10)
- Document ML order quantity gap in api_gateway proxy: MlOrderResponse
proto lacks quantity field (Task 12)
- Document per-symbol weight tracking roadmap in ensemble_coordinator (Task 13)
- Document OHLCV bar pipeline upgrade roadmap in state.rs (Task 14)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Add CandleCfCTrainer to training.rs with Candle-based gradient training
- Update mod.rs with full CfC v2 re-exports (CandleCfCNetwork, CfCCell, etc.)
- Fix CUDA variance bug (undefined variable) and kernel compilation stub
- Add 3 integration tests: full training loop, checkpoint roundtrip, validation
- 3 new unit tests for CandleCfCTrainer (creation, single epoch, loss decrease)
73 liquid tests pass, 0 errors, 0 clippy warnings in liquid module.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
4 tests validating the 51-dim feature extraction pipeline:
- DBN data loading (graceful skip if file absent)
- Synthetic bars: dimension check (51-dim), no NaN/Inf
- Value range bounds (-100 to 100)
- Streaming vs batch consistency (element-wise 1e-10 tolerance)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Remove 520 lines of ndarray-based placeholder code that never performed real
gradient descent. Production training uses Candle-based trainers in ml::trainers/.
Deleted: SimpleNeuralNetwork, TrainingPipeline, NetworkConfig, ActivationType,
TrainingMetrics, NetworkInterface, MockNetwork, and 9 associated tests.
Kept: DeviceCapabilities, TrainingConfig, sub-module re-exports.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Sequential build plan for 5 model integrations (TGGN, TLOB, KAN,
xLSTM, Diffusion) with TDD, worktree isolation, and full trait
implementations (UnifiedTrainable + ParameterSpace + hyperopt).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace load_and_split_data() mock that generated 2000 synthetic samples
with an error-returning stub directing users to ml_training_service.
The binary retains its real infrastructure (TFTTrainer, CLI, checkpoint
storage, progress callbacks) — only the fake data generation is removed.
-120 lines of mock data, +16 lines error stub with documentation.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Approved design for expanding the ensemble with KAN, xLSTM, and
Diffusion architectures plus bringing TGGN/TLOB to full integration.
Sequential build order, no god classes, modular decomposition.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Fix leftover conflict markers from rebase onto main. Align as_str()
with main's semantics (general-purpose model names), add separate
s3_prefix() for S3 storage paths, and fix to_db_string() to preserve
per-variant database values.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Extend checkpoint_roundtrip.rs with 4 new tests for sequence-buffered models:
- TFT adapter deterministic inference: verifies same adapter produces
identical direction/confidence on repeated calls with stable buffer
- TFT quantile metadata: confirms quantiles are absent during buffering
phase and present (with correct count) after buffer fills
- Mamba2 adapter deterministic inference: same pattern as TFT, verifies
direction/confidence stability and correct model name ("MAMBA-2")
- Mamba2 CheckpointManager roundtrip: saves/loads via Checkpointable
trait on Mamba2SSM, verifies metadata tags and hyperparameters
All 10 tests (6 existing DQN/PPO + 4 new TFT/Mamba2) pass consistently.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Full adapter bridging CandleCfCNetwork to the unified training pipeline
with VarMap-based checkpointing, AdamW optimizer, and gradient norm
tracking. Includes 10 unit tests covering creation, training steps,
validation, metrics, learning rate, checkpoint roundtrip, and error cases.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Bridge the ML crate's parking_lot::RwLock-based CircuitBreaker to the
shared CircuitBreakerTrait from common. Maps synchronous methods
(allow_request, record_success, etc.) to the async trait interface and
converts CircuitState variants to common::CircuitBreakerState.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Define an async trait that abstracts the circuit breaker state machine
(Closed -> Open -> HalfOpen) so different implementations can be used
polymorphically. Implement the trait for the existing tokio::Mutex-based
CircuitBreaker struct by delegating to its existing async methods.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Remove 8 warning prints (tracing::warn! and eprintln!) from Default impls
in config.rs. These fired on every test run and default construction,
creating noise. The migration notice in the module-level comment remains
as documentation.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
