ml_switcheroo.frameworks.torch¶

PyTorch Adapter with Dynamic Introspection.

Attributes¶

torch

Classes¶

TorchAdapter

Adapter for PyTorch.

Module Contents¶

ml_switcheroo.frameworks.torch.torch = None¶

class ml_switcheroo.frameworks.torch.TorchAdapter¶

Adapter for PyTorch.

display_name: str = 'PyTorch'¶

inherits_from: None = None¶

ui_priority: int = 0¶

property search_modules: List[str]¶

property import_alias: Tuple[str, str]¶

property supported_tiers: List[ml_switcheroo.enums.SemanticTier]¶

property import_namespaces: Dict[str, Dict[str, str]]¶

Defines how to import specific submodules.

Example

“torch.nn” -> import torch.nn as nn (root=’torch’, sub=’nn’, alias=’nn’) “torch.nn.functional” -> import torch.nn.functional as F (root=’torch.nn.functional’, sub=None, alias=’F’)

property discovery_heuristics: Dict[str, List[str]]¶

property test_config: Dict[str, str]¶: Test templates for Torch.

property structural_traits: ml_switcheroo.frameworks.base.StructuralTraits¶

property definitions: Dict[str, ml_switcheroo.frameworks.base.StandardMap]¶: Comprehensive distributed definitions for PyTorch mappings. Moved from standards_internal.py.

property rng_seed_methods: List[str]¶

collect_api(category: ml_switcheroo.frameworks.base.StandardCategory) → List[ml_switcheroo.frameworks.base.GhostRef]¶: Collects API signatures dynamically.

get_device_syntax(device_type: str, device_index: None | str = None) → str¶

get_serialization_imports() → List[str]¶

get_serialization_syntax(op: str, file_arg: str, object_arg: None | str = None) → str¶

classmethod get_example_code() → str¶

get_tiered_examples() → Dict[str, str]¶

convert(data)¶

apply_wiring(snapshot: Dict[str, Any]) → None¶: Applies manual fixes.