StatefulSymbolicContext

class torch.fx.experimental.symbolic_shapes.StatefulSymbolicContext(dynamic_sizes, dynamic_strides=None, constraint_sizes=None, constraint_strides=None, view_base_context=None, tensor_source=None, shape_env_to_source_to_symbol_cache=None)

Create symbols in create_symbolic_sizes_strides_storage_offset via a symbolic_context determination as given by a cache of Source:Symbol. A cache hit will reuse a stored symbol, and a cache miss will write to this cache.

This behaves like StatelessSymbolicContext, except the cache supersedes the other values - dynamic_sizes and constraint_sizes will not be read if we cache hit.

It is the cache owners responsibility to maintain the lifecycle of the cache w/r/t different shape_envs, clearing, etc.