Add backup.

Author: benedictchen

run_new_tests.py

@@ -92,7 +92,12 @@ def run_all_new_tests(verbose: bool = False) -> Dict[str, Any]:
         'tests/test_adaptive_episode_sampler.py', 
         'tests/test_curriculum_sampler.py',
         'tests/test_omniglot_dataset.py',
-        'tests/test_integration_workflow.py'
+        'tests/test_integration_workflow.py',
+        'tests/test_phase4_ml_enhancements.py',
+        'tests/test_error_handling_comprehensive.py',
+        'tests/test_evaluation_comprehensive.py',
+        'tests/test_research_patches_determinism.py',
+        'tests/test_performance_benchmarking.py'
     ]
 
     results = []
@@ -145,6 +150,11 @@ def run_all_new_tests(verbose: bool = False) -> Dict[str, Any]:
         print("✅ CurriculumSampler operational")
         print("✅ OmniglotDataset implemented")
         print("✅ Complete workflow integration successful")
+        print("✅ Phase 4 ML-powered enhancements functional")
+        print("✅ Error handling and recovery mechanisms working")
+        print("✅ Evaluation harness and metrics operational")
+        print("✅ Research patches and determinism validated")
+        print("✅ Performance benchmarking comprehensive")
     else:
         print(f"\n💥 {failed} test file(s) failed - see details above")
 
@@ -171,6 +181,11 @@ def main():
         print("  - CurriculumSampler for progressive learning")
         print("  - OmniglotDataset implementation")
         print("  - Complete workflow integration tests")
+        print("  - Phase 4 ML-powered enhancement components")
+        print("  - Error handling and recovery mechanisms")
+        print("  - Evaluation harness and metrics")
+        print("  - Research patches and determinism hooks")
+        print("  - Performance benchmarking and scalability")
         print()
         print("Options:")
         print("  --verbose, -v    Show detailed test output")

src/meta_learning/__init__.py

@@ -32,229 +32,61 @@
 
 from ._version import __version__
 
-# Core components - always available
+# Core essential components that should always work
 from .shared.types import Episode
 from .core.utils import partition_task, remap_labels
 from .core.seed import seed_all
-from .core.bn_policy import freeze_batchnorm_running_stats
 from .core.math_utils import pairwise_sqeuclidean, cosine_logits
 
-# Data handling - essential utilities  
+# Essential data utilities
 from .data_utils import InfiniteIterator, OnDeviceDataset
-from .data import SyntheticFewShotDataset, CIFARFSDataset, MiniImageNetDataset, make_episodes
 
-# Models - import conditionally to prevent crashes
-try:
-    from .models.conv4 import Conv4
-    CONV4_AVAILABLE = True
-except ImportError:
-    Conv4 = None
-    CONV4_AVAILABLE = False
+# Model architectures - recently implemented and tested
+from .models.conv4 import Conv4, ResNet12, WideResNet, MetaModule, MetaLinear, MetaConv2d
 
-# Algorithms - now with integrated advanced features
-from .algos.protonet import ProtoHead  # Now includes uncertainty estimation
-from .algos.maml import inner_adapt_and_eval, meta_outer_step, ContinualMAML
+# Basic algorithms
+from .algos.protonet import ProtoHead
+from .algos.maml import inner_adapt_and_eval, meta_outer_step
 
 # Evaluation
 from .eval import evaluate
 
 # Benchmarking
 from .bench import run_benchmark
 
-
-# Hardware acceleration and research integrity (imported separately for CLI)
-try:
-    from .hardware_utils import (
-        HardwareConfig, HardwareDetector, MemoryManager, ModelOptimizer,
-        HardwareProfiler, create_hardware_config, setup_optimal_hardware
-    )
-    from .leakage_guard import (
-        LeakageGuard, LeakageType, LeakageViolation, create_leakage_guard
-    )
-    INTEGRATED_ADVANCED_AVAILABLE = True
-except ImportError:
-    INTEGRATED_ADVANCED_AVAILABLE = False
-
-# Legacy standalone modules (for backward compatibility)
-try:
-    from .continual_meta_learning import (
-        OnlineMetaLearner, ContinualMetaConfig, FisherInformationMatrix,
-        EpisodicMemoryBank, create_continual_meta_learner
-    )
-    from .few_shot_modules.uncertainty_components import (
-        UncertaintyAwareDistance, MonteCarloDropout, DeepEnsemble,
-        EvidentialLearning, UncertaintyConfig, create_uncertainty_aware_distance
-    )
-    STANDALONE_MODULES_AVAILABLE = True
-except ImportError:
-    STANDALONE_MODULES_AVAILABLE = False
-
-try:
-    # Basic  
-    from .algorithms.ttc_scaler import TestTimeComputeScaler
-    from .algorithms.ttc_config import TestTimeComputeConfig
-    from .algorithms.maml_research_accurate import ResearchMAML, MAMLConfig, MAMLVariant
-    
-    # Import research patches  
-    from .research_patches.batch_norm_policy import apply_episodic_bn_policy, EpisodicBatchNormPolicy
-    from .research_patches.determinism_hooks import setup_deterministic_environment, DeterminismManager
-    
-    # Import evaluation harness
-    from .evaluation.few_shot_evaluation_harness import FewShotEvaluationHarness
-    
-    # External research features available
-    EXTERNAL_RESEARCH_AVAILABLE = True
-    
-except ImportError as e:
-    # External research modules not available - core restored functionality still works!
-    import warnings
-    warnings.warn(f"External research modules not available: {e}. Core restored functionality still available!")
-    
-    TestTimeComputeScaler = None
-    TestTimeComputeConfig = None
-    ResearchMAML = None
-    MAMLConfig = None
-    MAMLVariant = None
-    apply_episodic_bn_policy = None
-    EpisodicBatchNormPolicy = None
-    setup_deterministic_environment = None
-    DeterminismManager = None
-    FewShotEvaluationHarness = None
-    EXTERNAL_RESEARCH_AVAILABLE = False
-
-# Core restored functionality is ALWAYS available now
-RESEARCH_AVAILABLE = True
-
-# Input validation and error handling (newly implemented)
+# Core validation functionality
 from .validation import (
     ValidationError, ConfigurationWarning, validate_episode_tensors,
     validate_few_shot_configuration, validate_distance_metric, 
-    validate_temperature_parameter, validate_learning_rate,
-    validate_model_parameters, validate_maml_config, validate_uncertainty_config,
-    validate_optimizer_config, validate_episodic_config, validate_regularization_config,
-    validate_complete_config, ValidationContext, check_episode_quality
-)
-
-# Error recovery and fault tolerance (newly implemented)
-from .error_recovery import (
-    RecoveryError, ErrorRecoveryManager, with_retry, safe_tensor_operation,
-    handle_numerical_instability, recover_from_dimension_mismatch,
-    RobustPrototypeNetwork, create_robust_episode, FaultTolerantTrainer, safe_evaluate
+    validate_temperature_parameter, validate_learning_rate
 )
-from .core.seed import ReproducibilityManager, distributed_seed_sync, benchmark_reproducibility_overhead, validate_seed_effectiveness
-from .core.bn_policy import apply_episodic_bn_policy, validate_bn_compatibility, EpisodicBatchNormPolicy
-from .algos.ttcs import TTCSWarningSystem, ttcs_with_fallback, ttcs_for_learn2learn_models, TTCSProfiler
-
-# High-level toolkit API - core component, should always work
-from .toolkit import MetaLearningToolkit, create_meta_learning_toolkit, quick_evaluation
-
-# Dataset ecosystem - newly implemented
-try:
-    from .data_utils.datasets import (
-        BenchmarkDatasetManager, OnDeviceDataset, InfiniteEpisodeIterator,
-        MiniImageNetDataset, SyntheticFewShotDataset, DatasetRegistry,
-        BaseMetaLearningDataset
-    )
-    DATASET_ECOSYSTEM_AVAILABLE = True
-except ImportError:
-    BenchmarkDatasetManager = None
-    OnDeviceDataset = None
-    InfiniteEpisodeIterator = None
-    MiniImageNetDataset = None
-    SyntheticFewShotDataset = None
-    DatasetRegistry = None
-    BaseMetaLearningDataset = None
-    DATASET_ECOSYSTEM_AVAILABLE = False
-
-# Error handling and monitoring - newly implemented
-try:
-    from .error_handling import (
-        IntelligentErrorRecovery, PerformanceMonitor, WarningManager,
-        ErrorType, WarningCategory, with_error_recovery, monitor_performance
-    )
-    ERROR_HANDLING_AVAILABLE = True
-except ImportError:
-    IntelligentErrorRecovery = None
-    PerformanceMonitor = None
-    WarningManager = None
-    ErrorType = None
-    WarningCategory = None
-    with_error_recovery = None
-    monitor_performance = None
-    ERROR_HANDLING_AVAILABLE = False
 
 __all__ = [
     "Episode",
     "remap_labels",
     "__version__",
     "partition_task",
+    "seed_all",
+    "pairwise_sqeuclidean",
+    "cosine_logits",
     "InfiniteIterator",
     "OnDeviceDataset",
+    "Conv4", 
+    "ResNet12",
+    "WideResNet", 
+    "MetaModule",
+    "MetaLinear",
+    "MetaConv2d",
+    "ProtoHead",
+    "inner_adapt_and_eval",
+    "meta_outer_step",
     "evaluate",
+    "run_benchmark",
+    "ValidationError",
+    "ConfigurationWarning",
     "validate_episode_tensors",
     "validate_few_shot_configuration",
     "validate_distance_metric",
     "validate_temperature_parameter",
-    "validate_learning_rate",
-    "validate_model_parameters",
-    "validate_maml_config",
-    "validate_uncertainty_config",
-    "validate_optimizer_config",
-    "validate_episodic_config",
-    "validate_regularization_config",
-    "validate_complete_config",
-    "validate_seed_effectiveness",
-    "validate_bn_compatibility",
-    "ValidationError",
-    "RecoveryError",
-    "ErrorRecoveryManager",
-    "IntelligentErrorRecovery",
-    "ErrorType",
-    "with_error_recovery",
-    "TestTimeComputeScaler",
-    "TestTimeComputeConfig",
-    "ResearchMAML",
-    "MAMLConfig",
-    "MAMLVariant",
-    "setup_deterministic_environment",
-    "DeterminismManager",
-    "FewShotEvaluationHarness",
-    "MetaLearningToolkit",
-    "create_meta_learning_toolkit",
-    "quick_evaluation",
-    "RESEARCH_AVAILABLE",
-    "EXTERNAL_RESEARCH_AVAILABLE",
-    "DATASET_ECOSYSTEM_AVAILABLE",
-    "ERROR_HANDLING_AVAILABLE",
-    "ConfigurationWarning",
-    "ValidationContext",
-    "check_episode_quality",
-    "with_retry",
-    "safe_tensor_operation",
-    "handle_numerical_instability",
-    "recover_from_dimension_mismatch",
-    "RobustPrototypeNetwork",
-    "create_robust_episode",
-    "FaultTolerantTrainer",
-    "safe_evaluate",
-    "ReproducibilityManager",
-    "distributed_seed_sync",
-    "benchmark_reproducibility_overhead",
-    "apply_episodic_bn_policy",
-    "EpisodicBatchNormPolicy",
-    "TTCSWarningSystem",
-    "ttcs_with_fallback",
-    "ttcs_for_learn2learn_models",
-    "TTCSProfiler",
-    "HardwareConfig",
-    "BenchmarkDatasetManager",
-    "InfiniteEpisodeIterator",
-    "MiniImageNetDataset",
-    "SyntheticFewShotDataset",
-    "DatasetRegistry",
-    "BaseMetaLearningDataset",
-    "PerformanceMonitor",
-    "WarningManager",
-    "WarningCategory",
-    "monitor_performance"
-]
+    "validate_learning_rate"
+]

src/meta_learning/algorithms/__init__.py

@@ -31,12 +31,20 @@
     MAMLVariant, 
     FunctionalModule
 )
+from .matching_networks import (
+    MatchingNetworks,
+    create_matching_networks,
+    matching_networks_loss
+)
 
 __all__ = [
     "TestTimeComputeScaler",
     "TestTimeComputeConfig", 
     "ResearchMAML",
     "MAMLConfig",
     "MAMLVariant",
-    "FunctionalModule"
+    "FunctionalModule",
+    "MatchingNetworks",
+    "create_matching_networks", 
+    "matching_networks_loss"
 ]