jgrusewski 76cfc0d185 perf: IQN quantiles 64→32 — saves 4.3GB VRAM, halves IQN GEMMs ...

Academic literature shows diminishing returns past 32 quantiles.
64 quantiles allocated 8.6GB for tiled intermediates (B×Q×hidden).
32 quantiles: 4.3GB (half), ~13 GEMMs halved in aux_child.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

2026-04-18 22:28:23 +02:00

..

backtesting

refactor: collapse 9-level to 7-level ExposureLevel — eliminate degenerate Flat variants

2026-04-11 11:54:09 +02:00

common

…

config

…

ctrader-openapi

…

data

…

database

…

market-data

…

ml

perf: IQN quantiles 64→32 — saves 4.3GB VRAM, halves IQN GEMMs

2026-04-18 22:28:23 +02:00

ml-asset-selection

…

ml-backtesting

…

ml-checkpoint

…

ml-core

cleanup: purge all bf16 naming remnants — pure f32/TF32 pipeline

2026-04-18 10:30:04 +02:00

ml-data

…

ml-data-validation

…

ml-dqn

cleanup: purge all bf16 naming remnants — pure f32/TF32 pipeline

2026-04-18 10:30:04 +02:00

ml-ensemble

…

ml-explainability

…

ml-features

fix: zen precommit — 2 criticals + 3 lows resolved

2026-04-17 02:05:20 +02:00

ml-hyperopt

…

ml-labeling

…

ml-observability

…

ml-paper-trading

…

ml-ppo

cleanup: rename all _bf16 kernel functions and variables — pure f32 pipeline

2026-04-18 10:19:12 +02:00

ml-regime

…

ml-regime-detection

…

ml-risk

…

ml-security

…

ml-stress-testing

…

ml-supervised

cleanup: remove half crate dependency from entire workspace

2026-04-10 18:54:14 +02:00

ml-universe

…

ml-validation

…

model_loader

…

risk

…

risk-data

…

storage

…

trading_engine

…

trading-data

…

training_uploader

…