7458f1be01
✅ Validation Results: - PPO training: 24.2s (1 epoch, 950 samples, dim=225) - Feature extraction: 105μs/bar (9.5x faster than target) - Model checkpoint: 293KB (147KB actor + 146KB critic) - GPU memory: 145MB used (96.4% headroom) - Zero dimension mismatches 📊 Success Criteria (5/5): ✅ Feature dimension = 225 (Wave C 201 + Wave D 24) ✅ Model state_dim = 225 ✅ Training completed without errors ✅ Checkpoint saved successfully ✅ No dimension mismatch errors 📁 Training Data Ready: - ES.FUT: 2.9MB, 180 days - NQ.FUT: 4.4MB, 180 days - 6E.FUT: 2.8MB, 180 days - ZN.FUT: 65KB, 90 days (clean) 🚀 Next: Full production model retraining (4 models, ~10min GPU time) 🤖 Generated with Claude Code (https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
120 lines
3.6 KiB
TOML
120 lines
3.6 KiB
TOML
[package]
|
|
name = "ml_training_service"
|
|
version.workspace = true
|
|
edition.workspace = true
|
|
rust-version.workspace = true
|
|
authors.workspace = true
|
|
license.workspace = true
|
|
description = "ML Training Service - Model training orchestration and lifecycle management for HFT trading"
|
|
|
|
[dependencies]
|
|
# Core async and utilities - USE WORKSPACE
|
|
tokio.workspace = true
|
|
uuid.workspace = true
|
|
serde.workspace = true
|
|
serde_json.workspace = true
|
|
chrono.workspace = true
|
|
thiserror.workspace = true
|
|
anyhow.workspace = true
|
|
num_cpus.workspace = true
|
|
clap.workspace = true
|
|
rust_decimal.workspace = true
|
|
|
|
# gRPC and protocol buffers - USE WORKSPACE
|
|
tonic.workspace = true
|
|
tonic-prost.workspace = true
|
|
tonic-reflection.workspace = true
|
|
prost.workspace = true
|
|
prost-types.workspace = true
|
|
|
|
# Database and compression - USE WORKSPACE
|
|
sqlx.workspace = true
|
|
flate2.workspace = true
|
|
|
|
# Async streams and utilities - USE WORKSPACE
|
|
tokio-stream.workspace = true
|
|
tokio-util.workspace = true
|
|
async-stream.workspace = true
|
|
futures.workspace = true
|
|
async-trait.workspace = true
|
|
tokio-retry.workspace = true
|
|
|
|
# Logging, tracing and metrics - USE WORKSPACE
|
|
tracing.workspace = true
|
|
tracing-subscriber.workspace = true
|
|
metrics.workspace = true
|
|
metrics-exporter-prometheus.workspace = true
|
|
prometheus.workspace = true
|
|
once_cell.workspace = true
|
|
|
|
# Utilities - USE WORKSPACE
|
|
base64.workspace = true
|
|
rand.workspace = true
|
|
regex.workspace = true
|
|
axum.workspace = true # Health endpoint HTTP server
|
|
|
|
# Redis for job queue persistence
|
|
redis = { version = "0.27", features = ["tokio-comp", "connection-manager"] }
|
|
|
|
# Unix signal handling (for stopping Optuna subprocesses gracefully)
|
|
[target.'cfg(unix)'.dependencies]
|
|
nix = { version = "0.29", features = ["signal"] }
|
|
|
|
# Cryptography - Production-grade encryption
|
|
aes-gcm = "0.10"
|
|
chacha20poly1305 = "0.10"
|
|
pbkdf2 = { version = "0.12", features = ["simple"] }
|
|
sha2 = "0.10"
|
|
zeroize = { version = "1.6", features = ["alloc"] }
|
|
|
|
# X.509 certificate parsing for mTLS
|
|
x509-parser = "0.16"
|
|
reqwest = { version = "0.12", features = ["rustls-tls"], default-features = false }
|
|
rustls = { version = "0.23", features = ["ring"] }
|
|
|
|
# PyTorch dependencies REMOVED - unacceptable for HFT latency requirements
|
|
# All ML training now uses candle-core ecosystem only
|
|
|
|
# Internal workspace crates
|
|
trading_engine.workspace = true
|
|
risk.workspace = true
|
|
ml = { workspace = true, default-features = false, features = ["financial"] } # Minimal ML for compilation
|
|
data.workspace = true
|
|
config = { workspace = true, features = ["postgres"] }
|
|
common = { workspace = true, features = ["database"] }
|
|
storage.workspace = true # Add missing storage dependency
|
|
# Model functionality from ml-data
|
|
ml-data = { path = "../../ml-data" }
|
|
|
|
# Object store dependencies for S3 integration
|
|
object_store = { workspace = true, features = ["aws"] }
|
|
bytes.workspace = true
|
|
|
|
# DBN (Databento Binary) for real market data loading
|
|
dbn = "0.42.0"
|
|
|
|
[build-dependencies]
|
|
# NOTE: Tonic 0.14+ uses tonic-prost-build instead of tonic-build
|
|
tonic-prost-build.workspace = true
|
|
prost-build.workspace = true
|
|
|
|
[[bin]]
|
|
name = "ml_training_service"
|
|
path = "src/main.rs"
|
|
|
|
[dev-dependencies]
|
|
tempfile.workspace = true
|
|
tower.workspace = true # Use workspace version (0.4)
|
|
tower-test = "0.4.0" # Match workspace tower version
|
|
jsonwebtoken = "9.3"
|
|
sysinfo = "0.30" # For stress test memory monitoring
|
|
arrow.workspace = true # For test data generation (Parquet)
|
|
parquet.workspace = true # For Parquet file writing
|
|
|
|
[features]
|
|
default = ["minimal"] # Production default: real data loading
|
|
minimal = ["ml/financial"]
|
|
gpu = ["ml/simd"] # GPU features now use candle-core only
|
|
debug = []
|
|
mock-data = [] # Enable mock training data for testing (DO NOT USE IN PRODUCTION)
|