51dd200e39
State padding: - pad128() helper for CUTLASS 128-element K-tile alignment - pad_states_kernel: scatter-copy contiguous states to padded [B, pad128(SD)] layout - states_buf, next_states_buf: allocated with pad128(state_dim) stride - gemmex_bf16_ldb: layer 1 GemmEx uses ldb=pad128(state_dim) for B-matrix - forward_online_raw, forward_target_raw: use padded ldb for first layer - compute_q_stats: padded states buffer uses pad128(state_dim) - state_dim_padded field on CublasForward Compile fix: - compile_training_kernels return type: 13 → 14 CudaFunction (pad_states_kernel) Compute-sanitizer: 1684 → 1137 errors (33% reduction). Remaining reads: bias vectors adjacent to weight matrices — harmless. 895/895 unit + 9/9 smoke tests pass. 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;