U 9%eh@sddlmZddlZddlmZmZmZmZmZm Z ddl Z ddl m Z ddlmZmZmZmZmZdgZe jZedd d d d d d dZdd d ddddZdS)) annotationsN)CallableDictListSequenceTupleUnion)dim) _ellipsis AnonymousAxiscomma_separate parse_patternvalidate_rearrange_expressions rearrangeintstrz&Callable[[torch.Tensor], torch.Tensor]) tensor_ndimpattern axes_lengthsreturnc st||\}}t|||tdd|jD}|jr||t|jd}t|jd}||}|krtd|d|dn6d}t|j}t|j}|krtd|d|d|||} | dkrd d Std dt | Dig} d|jD]} t | t rr| D],} t | t st f| <d7q | std } f| <| | | | d7q| tkrt} tfd dt |D| <|7qtd| qdddfdd }||j}||j}tfdd| D}tfdd|D}d}d|dtd| d|rJddd|Dnddt|dt|d|rdt|gd nd!}t|t|S)"azTranslate an `einops`-style pattern into a callable that performs the rearrange using first-class dimensions. Since the an equivalent result is computed for tensors with the same number of dimensions, with the same pattern and specified axes lengths, this function can be memoized. Args: tensor_ndim (int): the number of dimensions in the tensor to rearrange pattern (str): the `einops`-style rearrangement pattern axes_lengths (int): any additional length specifications for dimensions Returns: Callable[[torch.Tensor], torch.Tensor]: a callable that performs the rearrangement css|] }| VqdSN).0r rrY/var/www/html/Darija-Ai-API/env/lib/python3.8/site-packages/functorch/einops/rearrange.py *sz-_create_rearrange_callable..r z!Number of dimensions in pattern (zH) must be less than or equal to the number of dimensions in the tensor ()rz;) must be equal to the number of dimensions in the tensor (cSs|Srr)tensorrrrAz,_create_rearrange_callable..css|]}d|VqdS)dNr)rirrrrCs1c3s|]}|VqdSrr)rj)dims_ifirst_class_dimsrrrYsUnexpected dimension: z5Sequence[Union[List[Union[str, AnonymousAxis]], str]]z!List[Union[str, Tuple[str, ...]]]) compositionrcs`g}|D]R}t|tr4|tfdd|Dq|tkrL|tqtd|q|S)z|Convert a `ParsedExpression.composition` into a `Tensor.__getitem__` index of strings representing first class dims.c3s |]}|D] }|VqqdSrr)r identifierr identifier_dim_maprrris zJ_create_rearrange_callable..composition_to_dims..r') isinstancelistappendtupler extend ValueError)r(Zdim_composition dimensionr*rrcomposition_to_dims`s  z7_create_rearrange_callable..composition_to_dimsc3s|]}|dVqdSrNr)raxisr*rrrwsc3s"|]\}}|d|fVqdSr4r)rr5lengthr*rrrxsZ do_rearrangezdef z(tensor): z = dims(z) css$|]\}}d|d|dVqdS)z z.size =  Nr)rr r6rrrrsz tensor = tensor[z].order(z return tensor.sum(z, keepdim=False) z return tensor )rrsumr(Z has_ellipsislenZ identifiersr1r/ranger,r-rAssertionErrorr r.r itemsr joinexeclocals)rrrleftrightZ n_anon_dimsZn_ellipsis_dimsZ n_named_dimsZ pattern_ndimZn_dimsZ anon_axesr2r)Z anon_axisr3Z left_dimsZ right_dimsZ anon_dimsZspecified_lengthsZcustom_rearrange_callable_nameZcustom_rearrange_callable_coder)r%r&r+r_create_rearrange_callables                rCzAUnion[torch.Tensor, List[torch.Tensor], Tuple[torch.Tensor, ...]]z torch.Tensor)rrrrcKs.t|tjst|}t|j|f|}||S)aA native implementation of `einops.rearrange`, a reader-friendly smart element reordering for multidimensional tensors. This operation includes functionality of transpose (axes permutation), reshape (view), squeeze, unsqueeze, stack, concatenate and other operations. See: https://einops.rocks/api/rearrange/ Args: tensor (Tensor or sequence of Tensor): the tensor(s) to rearrange pattern (str): the rearrangement pattern axes_lengths (int): any additional length specifications for dimensions Returns: Tensor: the rearranged tensor Examples: >>> # suppose we have a set of 32 images in "h w c" format (height-width-channel) >>> images = torch.randn((32, 30, 40, 3)) >>> # stack along first (batch) axis, output is a single array >>> rearrange(images, 'b h w c -> b h w c').shape torch.Size([32, 30, 40, 3]) >>> # concatenate images along height (vertical axis), 960 = 32 * 30 >>> rearrange(images, 'b h w c -> (b h) w c').shape torch.Size([960, 40, 3]) >>> # concatenated images along horizontal axis, 1280 = 32 * 40 >>> rearrange(images, 'b h w c -> h (b w) c').shape torch.Size([30, 1280, 3]) >>> # reordered axes to "b c h w" format for deep learning >>> rearrange(images, 'b h w c -> b c h w').shape torch.Size([32, 3, 30, 40]) >>> # flattened each image into a vector, 3600 = 30 * 40 * 3 >>> rearrange(images, 'b h w c -> b (c h w)').shape torch.Size([32, 3600]) >>> # split each image into 4 smaller (top-left, top-right, bottom-left, bottom-right), 128 = 32 * 2 * 2 >>> rearrange(images, 'b (h1 h) (w1 w) c -> (b h1 w1) h w c', h1=2, w1=2).shape torch.Size([128, 15, 20, 3]) >>> # space-to-depth operation >>> rearrange(images, 'b (h h1) (w w1) c -> b h w (c h1 w1)', h1=2, w1=2).shape torch.Size([32, 15, 20, 12]) )r,torchZTensorstackrCndim)rrrZrearrange_callablerrrrs3  ) __future__r functoolstypingrrrrrrrDZ functorch._Cr Z_CZ_parsingr r r rr__all__Zdims lru_cacherCrrrrrs   ~