foxhunt

gitadmin/foxhunt

Fork 0

Commit Graph

Author	SHA1	Message	Date
jgrusewski	b616d024ad	fix(dqn): IQN GPU PER weights, staged GPU buffer, CUDA default in all ML crates Three fixes for GPU PER hot path: 1. IQN quantile loss used empty CPU weights Vec instead of GPU-resident weights_tensor_cached — caused CUDA_ERROR_ILLEGAL_ADDRESS from uninitialized GPU memory. Now uses cached GPU tensor matching C51 and standard DQN paths. 2. GpuPrioritized add()/add_batch() replaced with StagedGpuBuffer: add() stages on CPU (Vec::push, zero GPU ops), sample() batch-flushes staging→GPU in one DMA before sampling. Production path (insert_batch_tensors) bypasses staging entirely — GPU→GPU with zero CPU. 3. All 9 ML sub-crates default to cuda feature so `cargo test -p ml-dqn` exercises GPU code paths on CUDA workstations. CI service crates use default-features=false, unaffected. Test results: 350 passed (was 343+7 failed), 0 failed, 1 ignored. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>	2026-03-08 22:31:49 +01:00
jgrusewski	7e5af20373	fix(ci): make CUDA non-default in 5 remaining ml sub-crates ml-supervised, ml-ensemble, ml-labeling, ml-explainability, ml-hyperopt all had default = ["cuda"] which pulled cudarc into the CPU services build, causing compile-services to fail with "nvcc not found". Changed all to default = [] and wired ml/Cargo.toml cuda feature to propagate to all 8 sub-crates (was only 3: ml-core, ml-dqn, ml-ppo). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>	2026-03-08 19:35:55 +01:00
jgrusewski	d313486dc2	refactor(ml): split monolith into 9 sub-crates + delete dead code Extract 9 new sub-crates from the ml monolith to enable parallel compilation across the workspace: New crates (this commit): - ml-features (282 tests): feature engineering, 21 modules - ml-labeling (45 tests): triple barrier, meta-labeling, fractional diff - ml-ensemble (116 tests): ensemble coordination, voting, confidence - ml-hyperopt (47 tests): core PSO/TPE optimizer, parameter space - ml-checkpoint (41 tests): checkpoint persistence, compression, signing - ml-regime (68 tests): CUSUM, Bayesian changepoint, regime classification - ml-data-validation (67 tests): FDR correction, CPCV, data quality - ml-risk (33 tests): neural VaR, Kelly criterion, circuit breakers - ml-validation (43 tests): statistical validation, walk-forward, DSR Extended existing crates: - ml-dqn: added evaluation/ (backtesting engine, metrics, reports) and checkpoint implementation - ml-supervised: added checkpoint implementations - ml-core: added shared types needed by new sub-crates Pattern: each module in ml/ becomes a thin facade (pub use subcrate::*) with bridge modules staying in ml for cross-model adapter code. Dead code deleted (~7K lines): - 13 undeclared files in microstructure/ (never compiled) - 7 undeclared files + tests/ in risk/ (never compiled) - parquet_io, cache_service, cache_storage, minio_integration (unused) - extraction_wave_d_impl.rs (bare fn outside impl block) All 2,746 sub-crate tests + 951 ml tests pass. Full workspace builds clean. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>	2026-03-08 15:17:22 +01:00

Author

SHA1

Message

Date

jgrusewski

b616d024ad

fix(dqn): IQN GPU PER weights, staged GPU buffer, CUDA default in all ML crates

Three fixes for GPU PER hot path:

1. IQN quantile loss used empty CPU weights Vec instead of GPU-resident
   weights_tensor_cached — caused CUDA_ERROR_ILLEGAL_ADDRESS from
   uninitialized GPU memory. Now uses cached GPU tensor matching C51
   and standard DQN paths.

2. GpuPrioritized add()/add_batch() replaced with StagedGpuBuffer:
   add() stages on CPU (Vec::push, zero GPU ops), sample() batch-flushes
   staging→GPU in one DMA before sampling. Production path (insert_batch_tensors)
   bypasses staging entirely — GPU→GPU with zero CPU.

3. All 9 ML sub-crates default to cuda feature so `cargo test -p ml-dqn`
   exercises GPU code paths on CUDA workstations. CI service crates use
   default-features=false, unaffected.

Test results: 350 passed (was 343+7 failed), 0 failed, 1 ignored.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

2026-03-08 22:31:49 +01:00

jgrusewski

7e5af20373

fix(ci): make CUDA non-default in 5 remaining ml sub-crates

ml-supervised, ml-ensemble, ml-labeling, ml-explainability, ml-hyperopt
all had default = ["cuda"] which pulled cudarc into the CPU services
build, causing compile-services to fail with "nvcc not found".

Changed all to default = [] and wired ml/Cargo.toml cuda feature to
propagate to all 8 sub-crates (was only 3: ml-core, ml-dqn, ml-ppo).

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

2026-03-08 19:35:55 +01:00

jgrusewski

d313486dc2

refactor(ml): split monolith into 9 sub-crates + delete dead code

Extract 9 new sub-crates from the ml monolith to enable parallel
compilation across the workspace:

New crates (this commit):
- ml-features (282 tests): feature engineering, 21 modules
- ml-labeling (45 tests): triple barrier, meta-labeling, fractional diff
- ml-ensemble (116 tests): ensemble coordination, voting, confidence
- ml-hyperopt (47 tests): core PSO/TPE optimizer, parameter space
- ml-checkpoint (41 tests): checkpoint persistence, compression, signing
- ml-regime (68 tests): CUSUM, Bayesian changepoint, regime classification
- ml-data-validation (67 tests): FDR correction, CPCV, data quality
- ml-risk (33 tests): neural VaR, Kelly criterion, circuit breakers
- ml-validation (43 tests): statistical validation, walk-forward, DSR

Extended existing crates:
- ml-dqn: added evaluation/ (backtesting engine, metrics, reports)
  and checkpoint implementation
- ml-supervised: added checkpoint implementations
- ml-core: added shared types needed by new sub-crates

Pattern: each module in ml/ becomes a thin facade (pub use subcrate::*)
with bridge modules staying in ml for cross-model adapter code.

Dead code deleted (~7K lines):
- 13 undeclared files in microstructure/ (never compiled)
- 7 undeclared files + tests/ in risk/ (never compiled)
- parquet_io, cache_service, cache_storage, minio_integration (unused)
- extraction_wave_d_impl.rs (bare fn outside impl block)

All 2,746 sub-crate tests + 951 ml tests pass.
Full workspace builds clean.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

2026-03-08 15:17:22 +01:00

3 Commits