befa7e2f55
CUDA kernel: cooperative_load_tile() + matvec_leaky_relu_shmem() cache weight matrices in shared memory (SRAM), reducing HBM traffic 32x (1.76TB→55GB). Block size increased to 256 threads for better SHMEM utilization. Trainer hot-path: replace to_vec2 full-tensor CPU readbacks with GPU-native flatten_all/min/max/mean_all (4 scalar readbacks). Merge two Q-diagnostic forward passes into one. Deferred max_priority flush (1 GPU→CPU sync/epoch instead of ~8/batch). Fix StagedGpuBuffer::flush() dimension mismatch: used raw experience state.len() (45) instead of gpu.state_dim() (48 aligned), causing slice_scatter crash [500,48] vs [n,45] when CPU experiences flush into GPU-aligned replay buffer. Fix hardcoded state_dim=48 in optimal_n_episodes() — now uses dynamic state_dim from network config (correct for OFI-enabled 56-dim states). Dynamic episode auto-scaling guard: only auto-scale when gpu_n_episodes≥128 (production), respecting explicit test overrides (gpu_n_episodes=2). Tests: 874 ml + 392 ml-dqn = 1266 pass, 0 fail. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Foxhunt
Production HFT trading system in Rust.
Architecture
The workspace contains 32 crates organized as follows:
Core Libraries (16)
| Crate | Purpose |
|---|---|
trading_engine |
Order processing, FIX 4.4, IB TWS, SIMD, RDTSC timing |
risk |
VaR, Kelly, circuit breakers, kill switches, compliance |
risk-data |
Risk data types and shared structures |
trading-data |
Trading data types |
ml |
DQN Rainbow, PPO, TFT, Mamba2, ensemble inference |
ml-data |
ML data types and feature definitions |
data |
Market data ingestion and storage |
backtesting |
Replay engine, strategy tester |
adaptive-strategy |
Ensemble execution, microstructure analysis |
common |
Shared types, resilience, error handling |
storage |
S3 and local model storage |
model_loader |
Model serialization and loading |
market-data |
Market data feed handlers |
database |
PostgreSQL access layer (SQLx) |
config |
Configuration management |
tli |
CLI commands and tooling |
Services (8)
| Service | Purpose |
|---|---|
backtesting_service |
gRPC backtesting service |
broker_gateway_service |
FIX routing, broker connectivity |
trading_service |
Core trading operations |
ml_training_service |
Model training orchestration |
data_acquisition_service |
Market data acquisition |
trading_agent_service |
Autonomous trading agents |
api_gateway |
gRPC API gateway with auth |
web-gateway |
Axum REST + WebSocket gateway |
Frontend
web-dashboard/ -- React 19 + TypeScript + Vite + TradingView charts.
Building
# Check compilation (no PostgreSQL required)
SQLX_OFFLINE=true cargo check --workspace
# Run tests for a specific crate
SQLX_OFFLINE=true cargo test -p <crate> --lib
# Clippy
SQLX_OFFLINE=true cargo clippy --workspace
ML Models
Four production model architectures on Candle v0.9.1 with CUDA:
- DQN Rainbow -- Deep Q-Network with prioritized replay, dueling heads, noisy nets
- PPO -- Proximal Policy Optimization with GAE and LSTM policies
- TFT -- Temporal Fusion Transformer for multi-horizon forecasting
- Mamba2 -- State space model for sequence prediction
Each model has a standalone trainer and a UnifiedTrainable adapter for the hyperopt pipeline.
Infrastructure
- Git: Gitea at
git.fxhnt.ai(Tailscale-only), Scaleway DEV1-S - Observability: OpenTelemetry OTLP (env
OTEL_EXPORTER_OTLP_ENDPOINT) - Database: PostgreSQL with SQLx offline mode for CI
License
Proprietary. All rights reserved.