Spaces:
Sleeping
Sleeping
| """ | |
| Q-TensorFormer: Closed-Loop Information-to-Resource Adaptive Transformer. | |
| Dynamically allocates model capacity, computation depth, attention complexity, | |
| tensor rank, KV-cache precision, and memory residency according to the | |
| information value of each token under real deployment constraints. | |
| """ | |
| from .config import ( | |
| ModelConfig, TrainingConfig, BudgetConfig, ExperimentConfig, | |
| PRESETS, tiny_config, small_config, medium_config, production_config, | |
| validate_model_config, | |
| ) | |
| from .information_state import TokenInformationState | |
| from .resource_allocator import ( | |
| InformationValueAllocator, AllocationBudget, AllocatorAction, PIDDualSubgradientController, | |
| ConstrainedDecisionEngine, MarginalValueModel, | |
| ) | |
| from .kv_cache import ( | |
| AdaptiveKVCache, KVPrecision, QuantizedKVTensor, | |
| HierarchicalAdaptiveKVCache, KVResidencyTier, KVTransferStats, | |
| ) | |
| from .counterfactual_learning import ( | |
| CounterfactualDatasetGenerator, PolicyBenchmarkEvaluator, CounterfactualRecord, | |
| ) | |
| from .phase_aware_profiler import ( | |
| PhaseAwareProfiler, PhaseProfileResult, | |
| ) | |
| from .matched_budget_analyzer import ( | |
| MatchedBudgetAnalyzer, MatchedBudgetComparison, ConstraintSatisfactionResult, ParetoHypervolumeResult, | |
| ) | |
| from .quantum_backend import ( | |
| QuantumBackend, BackendType, compute_meyer_wallach_entanglement, compute_quantum_expressibility, | |
| ) | |
| from .tensor_layers import TTLinear, TTFeedForward, factorize_dim | |
| from .attention import MultiHeadAttention | |
| from .blocks import HybridBlock | |
| from .hardware_cost_model import ( | |
| HardwareCostModel, HardwareMeasurement, HardwareRooflineAnalyzer, RooflinePoint, KernelBreakdownProfiler, | |
| ) | |
| from .models import QTensorFormer, DenseBaseline, create_model | |
| from .validator import ScientificClaimValidator | |
| from .tokenizer import QTensorFormerTokenizer | |
| from .checkpoint import ( | |
| save_canonical_checkpoint, | |
| load_canonical_checkpoint, | |
| inspect_checkpoint_compatibility, | |
| ) | |
| from .metrics import evaluate_model | |
| from .training import LanguageModelTrainer, DistillationTrainer | |
| # Frontier Research Extensions | |
| from .speculative_decoding import ZeroSVDSpeculativeDecoder | |
| from .semantic_kv_cache import SemanticAnchorKVCache | |
| from .bandit_allocator import ContextualBanditAllocator | |
| from .adaptive_vision import AdaptiveVisionProjector | |
| from .dynamic_grid import DynamicGridController | |
| from .theory import TheoreticalAnalyzer | |
| # Systems & Enterprise Serving Extensions (NVIDIA Hopper/Blackwell/vLLM/TRT-LLM) | |
| from .vllm_adapter import VLLMContinuousBatchingAdapter, QTensorFormerPagedAttentionMetadata | |
| from .trt_llm_plugin import QTensorFormerTRTPlugin | |
| from .roofline_analyzer import EnterpriseRooflineAnalyzer, DecodingRooflineProfile, HardwareRooflineSpec | |
| # Real-World Production & Physical Hardware Governance Extensions | |
| from .realworld_llm_converter import RealWorldLLMConverter | |
| from .hardware_telemetry import LiveHardwareTelemetrySensor, HardwareTelemetrySnapshot | |
| from .reasoning_benchmark import ReasoningComplexityAnalyzer, ReasoningBenchmarkResult, TokenComplexityRecord | |
| # Hugging Face Integration | |
| from .hf_model import QTensorFormerConfig, QTensorFormerForCausalLM | |
| __version__ = "4.0.0" | |
| __all__ = [ | |
| "QTensorFormer", | |
| "DenseBaseline", | |
| "create_model", | |
| "ZeroSVDSpeculativeDecoder", | |
| "SemanticAnchorKVCache", | |
| "ContextualBanditAllocator", | |
| "AdaptiveVisionProjector", | |
| "DynamicGridController", | |
| "TheoreticalAnalyzer", | |
| "VLLMContinuousBatchingAdapter", | |
| "QTensorFormerPagedAttentionMetadata", | |
| "QTensorFormerTRTPlugin", | |
| "EnterpriseRooflineAnalyzer", | |
| "DecodingRooflineProfile", | |
| "HardwareRooflineSpec", | |
| "RealWorldLLMConverter", | |
| "LiveHardwareTelemetrySensor", | |
| "HardwareTelemetrySnapshot", | |
| "ReasoningComplexityAnalyzer", | |
| "ReasoningBenchmarkResult", | |
| "TokenComplexityRecord", | |
| "ModelConfig", | |
| "TrainingConfig", | |
| "BudgetConfig", | |
| "ExperimentConfig", | |
| "validate_model_config", | |
| "TokenInformationState", | |
| "InformationValueAllocator", | |
| "AllocationBudget", | |
| "AllocatorAction", | |
| "PIDDualSubgradientController", | |
| "ConstrainedDecisionEngine", | |
| "MarginalValueModel", | |
| "AdaptiveKVCache", | |
| "HierarchicalAdaptiveKVCache", | |
| "KVPrecision", | |
| "KVResidencyTier", | |
| "KVTransferStats", | |
| "QuantizedKVTensor", | |
| "CounterfactualDatasetGenerator", | |
| "PolicyBenchmarkEvaluator", | |
| "PhaseAwareProfiler", | |
| "PhaseProfileResult", | |
| "MatchedBudgetAnalyzer", | |
| "MatchedBudgetComparison", | |
| "ConstraintSatisfactionResult", | |
| "ParetoHypervolumeResult", | |
| "QuantumBackend", | |
| "BackendType", | |
| "compute_meyer_wallach_entanglement", | |
| "compute_quantum_expressibility", | |
| "TTLinear", | |
| "TTFeedForward", | |
| "factorize_dim", | |
| "MultiHeadAttention", | |
| "HybridBlock", | |
| "HardwareCostModel", | |
| "HardwareMeasurement", | |
| "HardwareRooflineAnalyzer", | |
| "RooflinePoint", | |
| "KernelBreakdownProfiler", | |
| "QTensorFormerConfig", | |
| "QTensorFormerForCausalLM", | |
| "ScientificClaimValidator", | |
| "QTensorFormerTokenizer", | |
| "save_canonical_checkpoint", | |
| "load_canonical_checkpoint", | |
| "inspect_checkpoint_compatibility", | |
| "evaluate_model", | |
| "LanguageModelTrainer", | |
| "DistillationTrainer", | |
| "__version__", | |
| ] | |