+ "ݳj^RIHt^RIt]!RR7!RR44t]!RR7!RR]44t]!RR7!R R ]44t]!RR7!R R ]44t]!RR7!R R]44t]!RR7!RR]44t]!RR7!RR]44t ]PP] ]].4R#)) dataclassNT)frozenc]tRt^ tRtRtR#) Granularityz Base class for representing the granularity of quantization. This class serves as a parent for specific granularity types used in quantization operations, such as per-tensor or per-axis quantization. N__name__ __module__ __qualname____firstlineno____doc____static_attributes__ra/home/luca/screen-checking/.venv/lib/python3.14/site-packages/torchao/quantization/granularity.pyrr  rrc]tRt^tRtRtR#) PerTensorz Represents per-tensor granularity in quantization. This granularity type calculates the quantization parameters based off the entire tensor. rNrrrrrrrrrc0a]tRt^$toRtV3RltRtVtR#)PerAxisa Represents per-axis granularity in quantization. This granularity type calculates different quantization parameters along a specified axis of the tensor. Examples: * input_tensor shape [A, B], axis 0 -> scale_shape [A, 1] * input_tensor shape [A, B], axis 1 -> scale_shape [1, B] * input_tensor shape [A, B, C], axis 1 -> scale_shape [1, B, 1] Attributes: axis (int): The axis which is kept, reduction is performed across all the other axes c&<V^8dQh/S[;R&#)axisint)format __classdict__s"r __annotate__PerAxis.__annotate__$s$ I%rrNr r r r r __annotate_func__r__classdictcell__rs@rrr$rrc0a]tRt^9toRtV3RltRtVtR#)PerGroupa Represents per-channel group granularity in quantization. This granularity type calculates different quantization parameters for each group of elements. For example if the input tensor is shape [8, 16], and the group size is 4, then the input tensor is reshaped to [64, 4] quantization parameters are calculated for each group of 4 elements, giving a total of 64 quantization parameters. Attributes: group_size (int): The size of each quantization group c&<V^8dQh/S[;R&#)r group_sizer)rrs"rrPerGroup.__annotate__9s$O%rrNrr"s@rr%r%9r#rr%c4a]tRt^NtoRtRtV3RltRtVtR#)PerRowa} Represents row-wise granularity in quantization. Examples: * input_tensor shape [A, B], dim 0 -> scale_shape [1, B] * input_tensor shape [A, B], dim 1 -> scale_shape [A, 1] * input_tensor shape [A, B], dim -1 -> scale_shape [A, 1] * input_tensor shape [A, B, C], dim 1 -> scale_shape [A, 1, C] Attributes: dim (int): The dim which is reduced across, all other dims are kept c&<V^8dQh/S[;R&#)rdimr)rrs"rrPerRow.__annotate__Ns MrrN) r r r r r r,r rr!r"s@rr*r*Ns Crr*c]tRt^`tRtRtR#)PerTokena Represents per-token granularity in quantization. This granularity type calculates a different set of quantization parameters for each token, which is represented as the last dimension of the tensor. For example, if the input tensor has shape [2, 3, 4], then there are 6 tokens with 4 elements each, and we will calculate 6 sets of quantization parameters, one for each token. If the input tensor has only two dimensions, e.g. [8, 16], then this is equivalent to `PerAxis(axis=0)`, which yields 8 sets of quantization parameters. rNrrrrr0r0`s   rr0c0a]tRt^stoRtV3RltRtVtR#)PerBlocka Represents multidimensional per-block granularity in quantization. Example: * block_size has shape [X, Y] * input_tensor shape [A] -> scaling undefined * input_tensor shape [A, B] -> scale shape [A // X, B // Y] * input_tensor shape [A, B, C] -> scale shape [A, B // X, C // Y] * input_tensor shape [A, B, C, D] -> scale shape [A, B, C // X, D // Y], and so on Note that `PerBlock((1, Y))` is equivalent to `PerGroup(Y)` Attributes: block_size (tuple[int, ...]): The size of each quantization group c:<V^8dQh/S[S[R3,;R&#)r. block_size)tupler)rrs"rrPerBlock.__annotate__ss6c3h7rrNrr"s@rr2r2sr#rr2) dataclassesrtorchrrrr%r*r0r2 serializationadd_safe_globalsrrrr;s"  $    $    $k( ${( $[" $ {  $ $ {  :$$h %BCr