b10cdebfb1
The GPU path (build_batch_states / build_state_tensor) only concatenated market[40] + portfolio[3] = 43 dims, while state_dim was set to 51 when OFI was enabled. This caused shape mismatch [128,43] vs [51,1024]. - Add ofi_features: Option<Tensor> field to DqnGpuData - Add upload_ofi() method for MBP-10 order book features (8 dims/bar) - Concatenate OFI in build_batch_states: [count,40]+[count,3]+[count,8]=[count,51] - Concatenate OFI in build_state_tensor: [1,40]+[1,3]+[1,8]=[1,51] - Wire trainer to call upload_ofi() after GPU data upload - Fix CPU path to zero-pad regime_features when OFI enabled but data missing 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;