U <c01 @sBUdZddlZddlZddlZddlZddlmZmZmZm Z m Z m Z ddl m Z edZddZdZdd eee e e e d ffefd d d Zdd eee e e e d ffefd d dZdd eee e e e d ffefd ddZdd eee e e e d ffefd ddZdd eee e e e d ffefd ddZdd eee e e e d ffefd ddZdd eee e e e d ffefd ddZejeiZee e e e d ffefed<ee,ddlZeeej <eeej!ej"ej#f<W5QRXeee$<eee%<eee <ddZ&dS)a5"Contains definitions of the methods used by the _BaseDataLoaderIter workers to collate samples fetched from dataset into Tensor(s). These **needs** to be in global scope since Py2 doesn't support serializing static methods. `default_collate` and `default_convert` are exposed to users via 'dataloader.py'. N)CallableDictOptionalTupleTypeUnion)string_classesz[SaUO]csPt}ttjrS|jdkr`|jdkr`|jdkr`|jdkrVtjj dk rVSt Stt j j rz|fddDWStk rfddDYSXnttrtd r|d d DSttrd d DStt j jrHttsHz|dd DWStk rDdd DYSXnSdS)an Function that converts each NumPy array element into a :class:`torch.Tensor`. If the input is a `Sequence`, `Collection`, or `Mapping`, it tries to convert each element inside to a :class:`torch.Tensor`. If the input is not an NumPy array, it is left unchanged. This is used as the default function for collation when both `batch_sampler` and `batch_size` are NOT defined in :class:`~torch.utils.data.DataLoader`. The general input type to output type mapping is similar to that of :func:`~torch.utils.data.default_collate`. See the description there for more details. Args: data: a single data point to be converted Examples: >>> # Example with `int` >>> default_convert(0) 0 >>> # Example with NumPy array >>> # xdoctest: +SKIP >>> default_convert(np.array([0, 1])) tensor([0, 1]) >>> # Example with NamedTuple >>> Point = namedtuple('Point', ['x', 'y']) >>> default_convert(Point(0, 0)) Point(x=0, y=0) >>> default_convert(Point(np.array(0), np.array(0))) Point(x=tensor(0), y=tensor(0)) >>> # Example with List >>> default_convert([np.array([0, 1]), np.array([2, 3])]) [tensor([0, 1]), tensor([2, 3])] numpyZstr_Zstring_ndarrayNcsi|]}|t|qSdefault_convert.0keydatar C/tmp/pip-unpacked-wheel-gikjz4vx/torch/utils/data/_utils/collate.py Asz#default_convert..csi|]}|t|qSr r rrr rrDs_fieldscss|]}t|VqdSNr rdr r r Fsz"default_convert..cSsg|] }t|qSr r rr r r Hsz#default_convert..cSsg|] }t|qSr r rr r rrKscSsg|] }t|qSr r rr r rrNs)type isinstancetorchTensor __module____name__np_str_obj_array_patternsearchdtypestr as_tensor collectionsabcMapping TypeErrortuplehasattrSequencer)r elem_typer rrr s4     r z\default_collate: batch must contain tensors, numpy arrays, numbers, dicts or lists; found {}collate_fn_map.csd}t|}dk rX|kr0|dSD]"}t||r4|dSq4t|tjjrz|fdd|DWStk rfdd|DYSXnt|trt|dr|fdd tDSt|tjj rt }t t |t fd d |Dstd tt}t|trHfd d |DSz|fdd |DWStk rfdd |DYSXtt|dS)a General collate function that handles collection type of element within each batch and opens function registry to deal with specific element types. `default_collate_fn_map` provides default collate functions for tensors, numpy arrays, numbers and strings. Args: batch: a single batch to be collated collate_fn_map: Optional dictionary mapping from element type to the corresponding collate function. If the element type isn't present in this dictionary, this function will go through each key of the dictionary in the insertion order to invoke the corresponding collate function if the element type is a subclass of the key. Examples: >>> # Extend this function to handle batch of tensors >>> def collate_tensor_fn(batch, *, collate_fn_map): ... return torch.stack(batch, 0) >>> def custom_collate(batch): ... collate_map = {torch.Tensor: collate_tensor_fn} ... return collate(batch, collate_fn_map=collate_map) >>> # Extend `default_collate` by in-place modifying `default_collate_fn_map` >>> default_collate_fn_map.update({torch.Tensor: collate_tensor_fn}) Note: Each collate function requires a positional argument for batch and a keyword argument for the dictionary of collate functions as `collate_fn_map`. rNr.cs(i|] tfddDdqS)csg|] }|qSr r rrr rrs&collate...r.collaterbatchr/r0rrszcollate..cs(i|] tfddDdqS)csg|] }|qSr r rr0r rrsr1r.r2r4r5r0rrsrc3s|]}t|dVqdS)r.Nr2rZsamplesr.r rrszcollate..c3s|]}t|kVqdSr)len)relem) elem_sizer rrsz5each element in list of batch should be of equal sizecsg|]}t|dqSr.r2r7r.r rrszcollate..csg|]}t|dqSr;r2r7r.r rrscsg|]}t|dqSr;r2r7r.r rrs)rrr&r'r(r)r*r+zipr,iterr8nextall RuntimeErrorlistdefault_collate_err_msg_formatformat)r6r/r9r-Z collate_typeitZ transposedr )r6r/r:rr3Xs8    r3cCst|d}d}tjjdk rdtdd|D}|j||jd}||j t |ft | }tj |d|dS)Nrcss|]}|VqdSr)numel)rxr r rrsz$collate_tensor_fn..)device)out)rutilsrZget_worker_infosumstorageZ _new_sharedrGnewZresize_r8rAsizestack)r6r/r9rHrErKr r rcollate_tensor_fns"rOcCs@|d}t|jjdk r*tt|jtdd|D|dS)NrcSsg|]}t|qSr rr%)rbr r rrsz*collate_numpy_array_fn..r.)r!r"r#r$r)rBrCr3)r6r/r9r r rcollate_numpy_array_fnsrRcCs t|SrrPr5r r rcollate_numpy_scalar_fnsrScCstj|tjdS)N)r#)rtensorZfloat64r5r r rcollate_float_fnsrUcCs t|Sr)rrTr5r r rcollate_int_fnsrVcCs|Srr r5r r rcollate_str_fnsrWdefault_collate_fn_mapcCs t|tdS)a Function that takes in a batch of data and puts the elements within the batch into a tensor with an additional outer dimension - batch size. The exact output type can be a :class:`torch.Tensor`, a `Sequence` of :class:`torch.Tensor`, a Collection of :class:`torch.Tensor`, or left unchanged, depending on the input type. This is used as the default function for collation when `batch_size` or `batch_sampler` is defined in :class:`~torch.utils.data.DataLoader`. Here is the general input type (based on the type of the element within the batch) to output type mapping: * :class:`torch.Tensor` -> :class:`torch.Tensor` (with an added outer dimension batch size) * NumPy Arrays -> :class:`torch.Tensor` * `float` -> :class:`torch.Tensor` * `int` -> :class:`torch.Tensor` * `str` -> `str` (unchanged) * `bytes` -> `bytes` (unchanged) * `Mapping[K, V_i]` -> `Mapping[K, default_collate([V_1, V_2, ...])]` * `NamedTuple[V1_i, V2_i, ...]` -> `NamedTuple[default_collate([V1_1, V1_2, ...]), default_collate([V2_1, V2_2, ...]), ...]` * `Sequence[V1_i, V2_i, ...]` -> `Sequence[default_collate([V1_1, V1_2, ...]), default_collate([V2_1, V2_2, ...]), ...]` Args: batch: a single batch to be collated Examples: >>> # Example with a batch of `int`s: >>> default_collate([0, 1, 2, 3]) tensor([0, 1, 2, 3]) >>> # Example with a batch of `str`s: >>> default_collate(['a', 'b', 'c']) ['a', 'b', 'c'] >>> # Example with `Map` inside the batch: >>> default_collate([{'A': 0, 'B': 1}, {'A': 100, 'B': 100}]) {'A': tensor([ 0, 100]), 'B': tensor([ 1, 100])} >>> # Example with `NamedTuple` inside the batch: >>> # xdoctest: +SKIP >>> Point = namedtuple('Point', ['x', 'y']) >>> default_collate([Point(0, 0), Point(1, 1)]) Point(x=tensor([0, 1]), y=tensor([0, 1])) >>> # Example with `Tuple` inside the batch: >>> default_collate([(0, 1), (2, 3)]) [tensor([0, 2]), tensor([1, 3])] >>> # Example with `List` inside the batch: >>> default_collate([[0, 1], [2, 3]]) [tensor([0, 2]), tensor([1, 3])] >>> # Two options to extend `default_collate` to handle specific type >>> # Option 1: Write custom collate function and invoke `default_collate` >>> def custom_collate(batch): ... elem = batch[0] ... if isinstance(elem, CustomType): # Some custom condition ... return ... ... else: # Fall back to `default_collate` ... return default_collate(batch) >>> # Option 2: In-place modify `default_collate_fn_map` >>> def collate_customtype_fn(batch, *, collate_fn_map=None): ... return ... >>> default_collate_fn_map.update(CustoType, collate_customtype_fn) >>> default_collate(batch) # Handle `CustomType` automatically r.)r3rX)r6r r rdefault_collates=rY)'__doc__r& contextlibrertypingrrrrrrZ torch._sixrcompiler!r rBr3rOrRrSrUrVrWrrX__annotations__suppress ImportErrorr npr Zbool_numberZobject_floatintrYr r r rs4    ?0B0 0 0000*