8f95908e16
- Fix NaN detection: replace non-existent isnan() with ne(&self) (NaN≠NaN) - C51 gradient strip: copy().detach() replaces to_vec2→from_vec GPU→CPU→GPU roundtrip - GPU batch fast path: skip CPU fold + 5× from_vec when GpuBatch available - Action validation: GPU clamp() replaces CPU loop in GPU batch path - PER TD errors: keep as GPU Tensor when GpuPrioritized replay active - PER indices: use GpuBatch.indices directly instead of CPU Vec→Tensor - IS weights: cache GPU tensor from GpuBatch, reuse in both loss paths - Logging: Q-values 10→500 steps, diagnostics 100→1000 steps 2758 tests pass, 0 clippy warnings. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
ml
10-model ML ensemble for the Foxhunt HFT system, built on Candle v0.9.1.
Models
- DQN (Rainbow) — deep Q-network with prioritized replay, dueling heads, noisy nets
- PPO — proximal policy optimization with GAE, LSTM policies, clip-higher
- TFT — temporal fusion transformer for multi-horizon forecasting
- Mamba2 — state space model for sequence prediction
- Liquid Networks — biologically inspired networks for non-stationary data
- TLOB — transformer-based limit order book analysis
- KAN — Kolmogorov-Arnold networks
- xLSTM — extended LSTM architecture
- TGGN — temporal graph neural network
- Diffusion — diffusion-based generative model
Key Modules
ensemble— model ensemble coordination and confidence aggregationhyperopt— PSO-based hyperparameter optimization with per-model adapterstrainers— unified training loops (DQN, PPO, supervised)inference—InferenceAdaptertrait for predictioncheckpoint— model checkpointing and restorationevaluation— walk-forward evaluation pipeline
Usage
use ml::dqn::DQN;
use ml::ppo::PpoTrainer;