29b6ec4dd9
All 6 loss/gradient CUDA kernels converted from float* to __nv_bfloat16*: - c51_loss_batched: BF16 logits (12 inputs), rewards, dones, IS weights, outputs - c51_grad_kernel: BF16 d_logits output, atomicAddBF16 for gradient accumulation - mse_loss_batched: BF16 logits, rewards, dones, IS weights, outputs - mse_grad_kernel: BF16 d_logits output, atomicAddBF16 - expected_q_kernel: BF16 logits in, BF16 Q-values out - q_stats_kernel: BF16 Q-values in (monitoring scalars stay float) Pattern: BF16 storage, F32 arithmetic (cast on load/store). Shared memory stays float for softmax/log/exp precision. total_loss stays float* (atomicAdd doesn't support BF16). 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;