0e2f82ab54
Integration of 7 hive agents: - gpu_replay_buffer: 103 Candle refs → 0 (14 new CUDA kernels) - gpu_action_selector: 27 refs → CudaSlice API - signal_adapter: 26 refs → 3 new CUDA kernels - gpu_experience_collector: 5 refs → CudaSlice output - gpu_weights+iql+guard: 13 refs eliminated - DQN forward: new forward_only_kernel for inference - VarMap: F32 contiguous enforcement, fast-path extraction New modules: - ml-core/cuda_autograd: GpuTensor, GpuVarStore, GpuLinear, GpuAdamW - ml-ppo/cuda_nn: CudaLinear, CudaLSTM, CudaAdam, networks - ml-supervised/gpu_tensor: GpuTensor + cuBLAS for KAN, Diffusion cudarc 0.17.3 → 0.19.3 (via candle 0.9.1 → 0.9.2) safetensors 0.4 → 0.7 Zero errors, zero warnings workspace-wide. Co-Authored-By: Claude Opus 4.6 (1M context) <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;