docs/build/doctrees/python/nn/_autosummary/mlx.nn.GroupNorm.doctree

<EFBFBD><05>E<00>sphinx.addnodes<65><73>document<6E><74><EFBFBD>)<29><>}<7D>(<28>	rawsource<63><65><00><>children<65>]<5D><>docutils.nodes<65><73>section<6F><6E><EFBFBD>)<29><>}<7D>(hhh]<5D>(h	<09>title<6C><65><EFBFBD>)<29><>}<7D>(h<05>mlx.nn.GroupNorm<72>h]<5D>h	<09>Text<78><74><EFBFBD><EFBFBD>mlx.nn.GroupNorm<72><6D><EFBFBD><EFBFBD><EFBFBD>}<7D>(<28>parent<6E>h<11>	_document<6E>h<03>source<63>N<EFBFBD>line<6E>Nuba<62>
attributes<EFBFBD>}<7D>(<28>ids<64>]<5D><>classes<65>]<5D><>names<65>]<5D><>dupnames<65>]<5D><>backrefs<66>]<5D>u<EFBFBD>tagname<6D>hhhhhh<1D>P/Users/awnihannun/repos/mlx/docs/src/python/nn/_autosummary/mlx.nn.GroupNorm.rst<73>hKubh<00>index<65><78><EFBFBD>)<29><>}<7D>(hhh]<5D>h}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>entries<65>]<5D>(<28>single<6C><65>GroupNorm (class in mlx.nn)<29><>mlx.nn.GroupNorm<72>hNt<4E>auh+h-hhhhhNhNubh<00>desc<73><63><EFBFBD>)<29><>}<7D>(hhh]<5D>(h<00>desc_signature<72><65><EFBFBD>)<29><>}<7D>(h<05>pGroupNorm(num_groups: int, dims: int, eps: float = 1e-05, affine: bool = True, pytorch_compatible: bool = False)<29>h]<5D>(h<00>desc_annotation<6F><6E><EFBFBD>)<29><>}<7D>(h<05>2[<#text: 'class'>, <desc_sig_space: <#text: ' '>>]<5D>h]<5D>(h<16>class<73><73><EFBFBD><EFBFBD><EFBFBD>}<7D>(hhKhhhNhNubh<00>desc_sig_space<63><65><EFBFBD>)<29><>}<7D>(h<05> <20>h]<5D>h<16> <20><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hhUhhhNhNubah}<7D>(h!]<5D>h#]<5D><>w<>ah%]<5D>h']<5D>h)]<5D>uh+hShhKubeh}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>	xml:space<63><65>preserve<76>uh+hIhhEhhh<1D>t/Users/awnihannun/repos/mlx/python/mlx/nn/layers/normalization.py:docstring of mlx.nn.layers.normalization.GroupNorm<72>hNubh<00>desc_addname<6D><65><EFBFBD>)<29><>}<7D>(h<05>mlx.nn.<2E>h]<5D>h<16>mlx.nn.<2E><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hhohhhNhNubah}<7D>(h!]<5D>h#]<5D>(<28>sig-prename<6D><65>
descclassname<6D>eh%]<5D>h']<5D>h)]<5D>hjhkuh+hmhhEhhhhlhNubh<00>	desc_name<6D><65><EFBFBD>)<29><>}<7D>(h<05>	GroupNorm<72>h]<5D>h<16>	GroupNorm<72><6D><EFBFBD><EFBFBD><EFBFBD>}<7D>(hh<>hhhNhNubah}<7D>(h!]<5D>h#]<5D>(<28>sig-name<6D><65>descname<6D>eh%]<5D>h']<5D>h)]<5D>hjhkuh+hhhEhhhhlhNubh<00>desc_parameterlist<73><74><EFBFBD>)<29><>}<7D>(h<05>enum_groups: int, dims: int, eps: float = 1e-05, affine: bool = True, pytorch_compatible: bool = False<73>h]<5D>(h<00>desc_parameter<65><72><EFBFBD>)<29><>}<7D>(h<05>num_groups: intint<6E>h]<5D>(h<00>
desc_sig_name<6D><65><EFBFBD>)<29><>}<7D>(h<05>
num_groups<EFBFBD>h]<5D>h<16>
num_groups<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>}<7D>(hh<>hhhNhNubah}<7D>(h!]<5D>h#]<5D><>n<>ah%]<5D>h']<5D>h)]<5D>uh+h<>hh<>ubh<00>desc_sig_punctuation<6F><6E><EFBFBD>)<29><>}<7D>(h<05>:<3A>h]<5D>h<16>:<3A><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hh<>hhhNhNubah}<7D>(h!]<5D>h#]<5D><>p<>ah%]<5D>h']<5D>h)]<5D>uh+h<>hh<>ubhT)<29><>}<7D>(h<05> <20>h]<5D>h<16> <20><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hh<>hhhNhNubah}<7D>(h!]<5D>h#]<5D>h`ah%]<5D>h']<5D>h)]<5D>uh+hShh<>ubh<62>)<29><>}<7D>(h<05>intint<6E>h]<5D>h<00>pending_xref<65><66><EFBFBD>)<29><>}<7D>(hhh]<5D>(h<00>pending_xref_condition<6F><6E><EFBFBD>)<29><>}<7D>(h<05>int<6E>h]<5D>h<16>int<6E><74><EFBFBD><EFBFBD><EFBFBD>}<7D>(hh<>hhhNhNubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>	condition<6F><6E>resolved<65>uh+h<>hh<>ubh<62>)<29><>}<7D>(h<05>int<6E>h]<5D>h<16>int<6E><74><EFBFBD><EFBFBD><EFBFBD>}<7D>(hh<>hhhNhNubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>	condition<6F><6E>*<2A>uh+h<>hh<>ubeh}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>	refdomain<69><6E>py<70><79>reftype<70><65>class<73><73>	reftarget<65><74>int<6E><74>refspecific<69><63><EFBFBD>	py:module<6C><65>mlx.nn<6E><6E>py:class<73>Nuh+h<>hh<>ubah}<7D>(h!]<5D>h#]<5D>h<EFBFBD>ah%]<5D>h']<5D>h)]<5D>uh+h<>hh<>ubeh}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>hjhkuh+h<>hh<>ubh<62>)<29><>}<7D>(h<05>dims: intint<6E>h]<5D>(h<>)<29><>}<7D>(h<05>dims<6D>h]<5D>h<16>dims<6D><73><EFBFBD><EFBFBD><EFBFBD>}<7D>(hjhhhNhNubah}<7D>(h!]<5D>h#]<5D>h<EFBFBD>ah%]<5D>h']<5D>h)]<5D>uh+h<>hjubh<62>)<29><>}<7D>(h<05>:<3A>h]<5D>h<16>:<3A><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj&hhhNhNubah}<7D>(h!]<5D>h#]<5D>h<EFBFBD>ah%]<5D>h']<5D>h)]<5D>uh+h<>hjubhT)<29><>}<7D>(h<05> <20>h]<5D>h<16> <20><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj4hhhNhNubah}<7D>(h!]<5D>h#]<5D>h`ah%]<5D>h']<5D>h)]<5D>uh+hShjubh<62>)<29><>}<7D>(h<05>intint<6E>h]<5D>h<EFBFBD>)<29><>}<7D>(hhh]<5D>(h<>)<29><>}<7D>(h<05>int<6E>h]<5D>h<16>int<6E><74><EFBFBD><EFBFBD><EFBFBD>}<7D>(hjIhhhNhNubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>	condition<6F>h<EFBFBD>uh+h<>hjFubh<62>)<29><>}<7D>(h<05>int<6E>h]<5D>h<16>int<6E><74><EFBFBD><EFBFBD><EFBFBD>}<7D>(hjXhhhNhNubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>	condition<6F>h<EFBFBD>uh+h<>hjFubeh}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>	refdomain<69>h<EFBFBD><68>reftype<70>j<00>	reftarget<65><74>int<6E><74>refspecific<69><63><EFBFBD>	py:module<6C>j<00>py:class<73>Nuh+h<>hjBubah}<7D>(h!]<5D>h#]<5D>h<EFBFBD>ah%]<5D>h']<5D>h)]<5D>uh+h<>hjubeh}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>hjhkuh+h<>hh<>ubh<62>)<29><>}<7D>(h<05>eps: floatfloat = 1e-05<30>h]<5D>(h<>)<29><>}<7D>(h<05>eps<70>h]<5D>h<16>eps<70><73><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubah}<7D>(h!]<5D>h#]<5D>h<EFBFBD>ah%]<5D>h']<5D>h)]<5D>uh+h<>hj<>ubh<62>)<29><>}<7D>(h<05>:<3A>h]<5D>h<16>:<3A><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubah}<7D>(h!]<5D>h#]<5D>h<EFBFBD>ah%]<5D>h']<5D>h)]<5D>uh+h<>hj<>ubhT)<29><>}<7D>(h<05> <20>h]<5D>h<16> <20><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubah}<7D>(h!]<5D>h#]<5D>h`ah%]<5D>h']<5D>h)]<5D>uh+hShj<>ubh<62>)<29><>}<7D>(h<05>
floatfloat<EFBFBD>h]<5D>h<EFBFBD>)<29><>}<7D>(hhh]<5D>(h<>)<29><>}<7D>(h<05>float<61>h]<5D>h<16>float<61><74><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>	condition<6F>h<EFBFBD>uh+h<>hj<>ubh<62>)<29><>}<7D>(h<05>float<61>h]<5D>h<16>float<61><74><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>	condition<6F>h<EFBFBD>uh+h<>hj<>ubeh}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>	refdomain<69>h<EFBFBD><68>reftype<70>j<00>	reftarget<65><74>float<61><74>refspecific<69><63><EFBFBD>	py:module<6C>j<00>py:class<73>Nuh+h<>hj<>ubah}<7D>(h!]<5D>h#]<5D>h<EFBFBD>ah%]<5D>h']<5D>h)]<5D>uh+h<>hj<>ubhT)<29><>}<7D>(h<05> <20>h]<5D>h<16> <20><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubah}<7D>(h!]<5D>h#]<5D>h`ah%]<5D>h']<5D>h)]<5D>uh+hShj<>ubh<00>desc_sig_operator<6F><72><EFBFBD>)<29><>}<7D>(h<05>=<3D>h]<5D>h<16>=<3D><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubah}<7D>(h!]<5D>h#]<5D><>o<>ah%]<5D>h']<5D>h)]<5D>uh+j<>hj<>ubhT)<29><>}<7D>(h<05> <20>h]<5D>h<16> <20><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hjhhhNhNubah}<7D>(h!]<5D>h#]<5D>h`ah%]<5D>h']<5D>h)]<5D>uh+hShj<>ubh	<09>inline<6E><65><EFBFBD>)<29><>}<7D>(h<05>1e-05<30>h]<5D>h<16>1e-05<30><35><EFBFBD><EFBFBD><EFBFBD>}<7D>(hjhhhNhNubah}<7D>(h!]<5D>h#]<5D><>
default_value<75>ah%]<5D>h']<5D>h)]<5D><>support_smartquotes<65><73>uh+jhj<>ubeh}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>hjhkuh+h<>hh<>ubh<62>)<29><>}<7D>(h<05>affine: boolbool = True<75>h]<5D>(h<>)<29><>}<7D>(h<05>affine<6E>h]<5D>h<16>affine<6E><65><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj/hhhNhNubah}<7D>(h!]<5D>h#]<5D>h<EFBFBD>ah%]<5D>h']<5D>h)]<5D>uh+h<>hj+ubh<62>)<29><>}<7D>(h<05>:<3A>h]<5D>h<16>:<3A><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj=hhhNhNubah}<7D>(h!]<5D>h#]<5D>h<EFBFBD>ah%]<5D>h']<5D>h)]<5D>uh+h<>hj+ubhT)<29><>}<7D>(h<05> <20>h]<5D>h<16> <20><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hjKhhhNhNubah}<7D>(h!]<5D>h#]<5D>h`ah%]<5D>h']<5D>h)]<5D>uh+hShj+ubh<62>)<29><>}<7D>(h<05>boolbool<6F>h]<5D>h<EFBFBD>)<29><>}<7D>(hhh]<5D>(h<>)<29><>}<7D>(h<05>bool<6F>h]<5D>h<16>bool<6F><6C><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj`hhhNhNubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>	condition<6F>h<EFBFBD>uh+h<>hj]ubh<62>)<29><>}<7D>(h<05>bool<6F>h]<5D>h<16>bool<6F><6C><EFBFBD><EFBFBD><EFBFBD>}<7D>(hjohhhNhNubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>	condition<6F>h<EFBFBD>uh+h<>hj]ubeh}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>	refdomain<69>h<EFBFBD><68>reftype<70>j<00>	reftarget<65><74>bool<6F><6C>refspecific<69><63><EFBFBD>	py:module<6C>j<00>py:class<73>Nuh+h<>hjYubah}<7D>(h!]<5D>h#]<5D>h<EFBFBD>ah%]<5D>h']<5D>h)]<5D>uh+h<>hj+ubhT)<29><>}<7D>(h<05> <20>h]<5D>h<16> <20><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubah}<7D>(h!]<5D>h#]<5D>h`ah%]<5D>h']<5D>h)]<5D>uh+hShj+ubj<62>)<29><>}<7D>(h<05>=<3D>h]<5D>h<16>=<3D><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubah}<7D>(h!]<5D>h#]<5D>jah%]<5D>h']<5D>h)]<5D>uh+j<>hj+ubhT)<29><>}<7D>(h<05> <20>h]<5D>h<16> <20><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubah}<7D>(h!]<5D>h#]<5D>h`ah%]<5D>h']<5D>h)]<5D>uh+hShj+ubj)<29><>}<7D>(h<05>True<75>h]<5D>h<16>True<75><65><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubah}<7D>(h!]<5D>h#]<5D>j ah%]<5D>h']<5D>h)]<5D><>support_smartquotes<65><73>uh+jhj+ubeh}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>hjhkuh+h<>hh<>ubh<62>)<29><>}<7D>(h<05>$pytorch_compatible: boolbool = False<73>h]<5D>(h<>)<29><>}<7D>(h<05>pytorch_compatible<6C>h]<5D>h<16>pytorch_compatible<6C><65><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubah}<7D>(h!]<5D>h#]<5D>h<EFBFBD>ah%]<5D>h']<5D>h)]<5D>uh+h<>hj<>ubh<62>)<29><>}<7D>(h<05>:<3A>h]<5D>h<16>:<3A><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubah}<7D>(h!]<5D>h#]<5D>h<EFBFBD>ah%]<5D>h']<5D>h)]<5D>uh+h<>hj<>ubhT)<29><>}<7D>(h<05> <20>h]<5D>h<16> <20><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubah}<7D>(h!]<5D>h#]<5D>h`ah%]<5D>h']<5D>h)]<5D>uh+hShj<>ubh<62>)<29><>}<7D>(h<05>boolbool<6F>h]<5D>h<EFBFBD>)<29><>}<7D>(hhh]<5D>(h<>)<29><>}<7D>(h<05>bool<6F>h]<5D>h<16>bool<6F><6C><EFBFBD><EFBFBD><EFBFBD>}<7D>(hjhhhNhNubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>	condition<6F>h<EFBFBD>uh+h<>hjubh<62>)<29><>}<7D>(h<05>bool<6F>h]<5D>h<16>bool<6F><6C><EFBFBD><EFBFBD><EFBFBD>}<7D>(hjhhhNhNubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>	condition<6F>h<EFBFBD>uh+h<>hjubeh}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>	refdomain<69>h<EFBFBD><68>reftype<70>j<00>	reftarget<65><74>bool<6F><6C>refspecific<69><63><EFBFBD>	py:module<6C>j<00>py:class<73>Nuh+h<>hj<>ubah}<7D>(h!]<5D>h#]<5D>h<EFBFBD>ah%]<5D>h']<5D>h)]<5D>uh+h<>hj<>ubhT)<29><>}<7D>(h<05> <20>h]<5D>h<16> <20><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj6hhhNhNubah}<7D>(h!]<5D>h#]<5D>h`ah%]<5D>h']<5D>h)]<5D>uh+hShj<>ubj<62>)<29><>}<7D>(h<05>=<3D>h]<5D>h<16>=<3D><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hjDhhhNhNubah}<7D>(h!]<5D>h#]<5D>jah%]<5D>h']<5D>h)]<5D>uh+j<>hj<>ubhT)<29><>}<7D>(h<05> <20>h]<5D>h<16> <20><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hjRhhhNhNubah}<7D>(h!]<5D>h#]<5D>h`ah%]<5D>h']<5D>h)]<5D>uh+hShj<>ubj)<29><>}<7D>(h<05>False<73>h]<5D>h<16>False<73><65><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj`hhhNhNubah}<7D>(h!]<5D>h#]<5D>j ah%]<5D>h']<5D>h)]<5D><>support_smartquotes<65><73>uh+jhj<>ubeh}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>hjhkuh+h<>hh<>ubeh}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>hjhkuh+h<>hhEhhhhlhNubeh}<7D>(h!]<5D>h<ah#]<5D>(<28>sig<69><67>
sig-object<63>eh%]<5D>h']<5D>h)]<5D><>module<6C><65>mlx.nn<6E>jh<06>fullname<6D>h<EFBFBD>uh+hCh<1D>t/Users/awnihannun/repos/mlx/python/mlx/nn/layers/normalization.py:docstring of mlx.nn.layers.normalization.GroupNorm<72>hKhh@hhubh<00>desc_content<6E><74><EFBFBD>)<29><>}<7D>(hhh]<5D>(h	<09>	paragraph<70><68><EFBFBD>)<29><>}<7D>(h<05>.Applies Group Normalization [1] to the inputs.<2E>h]<5D>h<16>.Applies Group Normalization [1] to the inputs.<2E><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+j<>h<1D>t/Users/awnihannun/repos/mlx/python/mlx/nn/layers/normalization.py:docstring of mlx.nn.layers.normalization.GroupNorm<72>hKhj<>hhubj<62>)<29><>}<7D>(h<05>5Computes the same normalization as layer norm, namely<6C>h]<5D>h<16>5Computes the same normalization as layer norm, namely<6C><79><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+j<>hj<>hKhj<>hhubh	<09>
math_block<EFBFBD><EFBFBD><EFBFBD>)<29><>}<7D>(h<05>=y = \frac{x - E[x]}{\sqrt{Var[x]} + \epsilon} \gamma + \beta,<2C>h]<5D>h<16>=y = \frac{x - E[x]}{\sqrt{Var[x]} + \epsilon} \gamma + \beta,<2C><><EFBFBD><EFBFBD><EFBFBD>}<7D>hj<>sbah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>docname<6D><65>'python/nn/_autosummary/mlx.nn.GroupNorm<72><6D>number<65>N<EFBFBD>label<65>N<EFBFBD>nowrap<61><70>hjhkuh+j<>hj<>hKhj<>hhubj<62>)<29><>}<7D>(hX,where :math:`\gamma` and :math:`\beta` are learned per feature dimension
parameters initialized at 1 and 0 respectively. However, the mean and
variance are computed over the spatial dimensions and each group of
features. In particular, the input is split into num_groups across the
feature dimension.<2E>h]<5D>(h<16>where <20><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubh	<09>math<74><68><EFBFBD>)<29><>}<7D>(h<05>:math:`\gamma`<60>h]<5D>h<16>\gamma<6D><61><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+j<>hj<>ubh<16> and <20><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubj<62>)<29><>}<7D>(h<05>
:math:`\beta`<60>h]<5D>h<16>\beta<74><61><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+j<>hj<>ubhX are learned per feature dimension
parameters initialized at 1 and 0 respectively. However, the mean and
variance are computed over the spatial dimensions and each group of
features. In particular, the input is split into num_groups across the
feature dimension.<2E><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubeh}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+j<>hj<>hK	hj<>hhubj<62>)<29><>}<7D>(h<05><>The feature dimension is assumed to be the last dimension and the dimensions
that precede it (except the first) are considered the spatial dimensions.<2E>h]<5D>h<16><>The feature dimension is assumed to be the last dimension and the dimensions
that precede it (except the first) are considered the spatial dimensions.<2E><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+j<>hj<>hKhj<>hhubj<62>)<29><>}<7D>(h<05>%[1]: https://arxiv.org/abs/1803.08494<EFBFBD>h]<5D>(h<16>[1]: <20><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hjhhhNhNubh	<09>	reference<63><65><EFBFBD>)<29><>}<7D>(h<05> https://arxiv.org/abs/1803.08494<EFBFBD>h]<5D>h<16> https://arxiv.org/abs/1803.08494<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>}<7D>(hjhhhNhNubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>refuri<72>juh+j
hjubeh}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+j<>hj<>hKhj<>hhubh	<09>
field_list<EFBFBD><EFBFBD><EFBFBD>)<29><>}<7D>(hhh]<5D>h	<09>field<6C><64><EFBFBD>)<29><>}<7D>(hhh]<5D>(h	<09>
field_name<EFBFBD><EFBFBD><EFBFBD>)<29><>}<7D>(h<05>
Parameters<EFBFBD>h]<5D>h<16>
Parameters<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj-hhhNhNubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+j+hj(hhlhKubh	<09>
field_body<EFBFBD><EFBFBD><EFBFBD>)<29><>}<7D>(hhh]<5D>h	<09>bullet_list<73><74><EFBFBD>)<29><>}<7D>(hhh]<5D>(h	<09>	list_item<65><6D><EFBFBD>)<29><>}<7D>(hhh]<5D>j<EFBFBD>)<29><>}<7D>(h<05>Enum_groups (intint) -- Number of groups to separate the features into<74>h]<5D>(h<00>literal_strong<6E><67><EFBFBD>)<29><>}<7D>(h<05>
num_groups<EFBFBD>h]<5D>h<16>
num_groups<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>}<7D>(hjPhhhNhNubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+jNhjJubh<16> (<28><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hjJhhhNhNubh<62>)<29><>}<7D>(hhh]<5D>(h<>)<29><>}<7D>(h<05>int<6E>h]<5D>h<00>literal_emphasis<69><73><EFBFBD>)<29><>}<7D>(h<05>int<6E>h]<5D>h<16>int<6E><74><EFBFBD><EFBFBD><EFBFBD>}<7D>(hjkhhhNhNubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+jihjeubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>	condition<6F>h<EFBFBD>uh+h<>hjbubh<62>)<29><>}<7D>(h<05>int<6E>h]<5D>jj)<29><>}<7D>(h<05>int<6E>h]<5D>h<16>int<6E><74><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+jihj<>ubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>	condition<6F>h<EFBFBD>uh+h<>hjbubeh}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>	refdomain<69><6E>py<70><79>refexplicit<69><74><EFBFBD>reftype<70>j<00>	reftarget<65>j<EFBFBD><00>refspecific<69><63><EFBFBD>	py:module<6C>j<EFBFBD><00>py:class<73>h<EFBFBD>uh+h<>hjJubh<16>)<29><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hjJhhhNhNubh<16> – <20><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hjJhhhNhNubh<16>.Number of groups to separate the features into<74><6F><EFBFBD><EFBFBD><EFBFBD>}<7D>(hjJhhhNhNubeh}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+j<>hjGubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+jEhjBubjF)<29><>}<7D>(hhh]<5D>j<EFBFBD>)<29><>}<7D>(h<05>Fdims (intint) -- The feature dimensions of the input to normalize over<65>h]<5D>(jO)<29><>}<7D>(h<05>dims<6D>h]<5D>h<16>dims<6D><73><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+jNhj<>ubh<16> (<28><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubh<62>)<29><>}<7D>(hhh]<5D>(h<>)<29><>}<7D>(h<05>int<6E>h]<5D>jj)<29><>}<7D>(h<05>int<6E>h]<5D>h<16>int<6E><74><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+jihj<>ubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>	condition<6F>h<EFBFBD>uh+h<>hj<>ubh<62>)<29><>}<7D>(h<05>int<6E>h]<5D>jj)<29><>}<7D>(h<05>int<6E>h]<5D>h<16>int<6E><74><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+jihj<>ubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>	condition<6F>h<EFBFBD>uh+h<>hj<>ubeh}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>	refdomain<69>j<EFBFBD><00>refexplicit<69><74><EFBFBD>reftype<70>j<00>	reftarget<65>j<EFBFBD>j<><00>j<EFBFBD>j<>j<>h<>uh+h<>hj<>ubh<16>)<29><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubh<16> – <20><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubh<16>5The feature dimensions of the input to normalize over<65><72><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubeh}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+j<>hj<>ubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+jEhjBubjF)<29><>}<7D>(hhh]<5D>j<EFBFBD>)<29><>}<7D>(h<05>Eeps (floatfloat) -- A small additive constant for numerical stability<74>h]<5D>(jO)<29><>}<7D>(h<05>eps<70>h]<5D>h<16>eps<70><73><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj6hhhNhNubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+jNhj2ubh<16> (<28><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj2hhhNhNubh<62>)<29><>}<7D>(hhh]<5D>(h<>)<29><>}<7D>(h<05>float<61>h]<5D>jj)<29><>}<7D>(h<05>float<61>h]<5D>h<16>float<61><74><EFBFBD><EFBFBD><EFBFBD>}<7D>(hjOhhhNhNubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+jihjKubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>	condition<6F>h<EFBFBD>uh+h<>hjHubh<62>)<29><>}<7D>(h<05>float<61>h]<5D>jj)<29><>}<7D>(h<05>float<61>h]<5D>h<16>float<61><74><EFBFBD><EFBFBD><EFBFBD>}<7D>(hjhhhhNhNubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+jihjdubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>	condition<6F>h<EFBFBD>uh+h<>hjHubeh}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>	refdomain<69>j<EFBFBD><00>refexplicit<69><74><EFBFBD>reftype<70>j<00>	reftarget<65>jjj<><00>j<EFBFBD>j<>j<>h<>uh+h<>hj2ubh<16>)<29><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj2hhhNhNubh<16> – <20><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj2hhhNhNubh<16>1A small additive constant for numerical stability<74><79><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj2hhhNhNubeh}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+j<>hj/ubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+jEhjBubjF)<29><>}<7D>(hhh]<5D>j<EFBFBD>)<29><>}<7D>(h<05>Xaffine (boolbool) -- If True learn an affine transform to apply after the
normalization.<2E>h]<5D>(jO)<29><>}<7D>(h<05>affine<6E>h]<5D>h<16>affine<6E><65><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+jNhj<>ubh<16> (<28><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubh<62>)<29><>}<7D>(hhh]<5D>(h<>)<29><>}<7D>(h<05>bool<6F>h]<5D>jj)<29><>}<7D>(h<05>bool<6F>h]<5D>h<16>bool<6F><6C><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+jihj<>ubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>	condition<6F>h<EFBFBD>uh+h<>hj<>ubh<62>)<29><>}<7D>(h<05>bool<6F>h]<5D>jj)<29><>}<7D>(h<05>bool<6F>h]<5D>h<16>bool<6F><6C><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+jihj<>ubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>	condition<6F>h<EFBFBD>uh+h<>hj<>ubeh}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>	refdomain<69>j<EFBFBD><00>refexplicit<69><74><EFBFBD>reftype<70>j<00>	reftarget<65>j<EFBFBD>j<><00>j<EFBFBD>j<>j<>h<>uh+h<>hj<>ubh<16>)<29><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubh<16> – <20><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubh<16>CIf True learn an affine transform to apply after the
normalization.<2E><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj<>hhhNhNubeh}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+j<>hj<>ubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+jEhjBubjF)<29><>}<7D>(hhh]<5D>j<EFBFBD>)<29><>}<7D>(h<05>opytorch_compatible (boolbool) -- If True perform the group normalization in
the same order/grouping as PyTorch.<2E>h]<5D>(jO)<29><>}<7D>(h<05>pytorch_compatible<6C>h]<5D>h<16>pytorch_compatible<6C><65><EFBFBD><EFBFBD><EFBFBD>}<7D>(hjhhhNhNubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+jNhjubh<16> (<28><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hjhhhNhNubh<62>)<29><>}<7D>(hhh]<5D>(h<>)<29><>}<7D>(h<05>bool<6F>h]<5D>jj)<29><>}<7D>(h<05>bool<6F>h]<5D>h<16>bool<6F><6C><EFBFBD><EFBFBD><EFBFBD>}<7D>(hj/hhhNhNubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+jihj+ubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>	condition<6F>h<EFBFBD>uh+h<>hj(ubh<62>)<29><>}<7D>(h<05>bool<6F>h]<5D>jj)<29><>}<7D>(h<05>bool<6F>h]<5D>h<16>bool<6F><6C><EFBFBD><EFBFBD><EFBFBD>}<7D>(hjHhhhNhNubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+jihjDubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>	condition<6F>h<EFBFBD>uh+h<>hj(ubeh}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>	refdomain<69>j<EFBFBD><00>refexplicit<69><74><EFBFBD>reftype<70>j<00>	reftarget<65>jJj<><00>j<EFBFBD>j<>j<>h<>uh+h<>hjubh<16>)<29><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hjhhhNhNubh<16> – <20><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hjhhhNhNubh<16>NIf True perform the group normalization in
the same order/grouping as PyTorch.<2E><><EFBFBD><EFBFBD><EFBFBD>}<7D>(hjhhhNhNubeh}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+j<>hjubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+jEhjBubeh}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+j@hj=ubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+j;hj(ubeh}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+j&hj#ubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+j!hj<>hhhNhNubeh}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D>uh+j<>hh@hhhhlhNubeh}<7D>(h!]<5D>h#]<5D>(j<><00>class<73>eh%]<5D>h']<5D>h)]<5D><>domain<69>j<EFBFBD><00>objtype<70>j<EFBFBD><00>desctype<70>j<EFBFBD><00>noindex<65><78>uh+h>hhhhhNhNubeh}<7D>(h!]<5D><>mlx-nn-groupnorm<72>ah#]<5D>h%]<5D><>mlx.nn.groupnorm<72>ah']<5D>h)]<5D>uh+h
hhhhhh,hKubah}<7D>(h!]<5D>h#]<5D>h%]<5D>h']<5D>h)]<5D><>source<63>h,uh+h<01>current_source<63>N<EFBFBD>current_line<6E>N<EFBFBD>settings<67><73>docutils.frontend<6E><64>Values<65><73><EFBFBD>)<29><>}<7D>(hN<>	generator<6F>N<EFBFBD>	datestamp<6D>N<EFBFBD>source_link<6E>N<EFBFBD>
source_url<EFBFBD>N<EFBFBD>
toc_backlinks<6B><73>entry<72><79>footnote_backlinks<6B>K<01>
sectnum_xform<72>K<01>strip_comments<74>N<EFBFBD>strip_elements_with_classes<65>N<EFBFBD>
strip_classes<65>N<EFBFBD>report_level<65>K<02>
halt_level<EFBFBD>K<05>exit_status_level<65>K<05>debug<75>N<EFBFBD>warning_stream<61>N<EFBFBD>	traceback<63><6B><EFBFBD>input_encoding<6E><67>	utf-8-sig<69><67>input_encoding_error_handler<65><72>strict<63><74>output_encoding<6E><67>utf-8<><38>output_encoding_error_handler<65>j<EFBFBD><00>error_encoding<6E><67>utf-8<><38>error_encoding_error_handler<65><72>backslashreplace<63><65>
language_code<64><65>en<65><6E>record_dependencies<65>N<EFBFBD>config<69>N<EFBFBD>	id_prefix<69>h<06>auto_id_prefix<69><78>id<69><64>
dump_settings<67>N<EFBFBD>dump_internals<6C>N<EFBFBD>dump_transforms<6D>N<EFBFBD>dump_pseudo_xml<6D>N<EFBFBD>expose_internals<6C>N<EFBFBD>strict_visitor<6F>N<EFBFBD>_disable_config<69>N<EFBFBD>_source<63>h,<2C>_destination<6F>N<EFBFBD>
_config_files<65>]<5D><>file_insertion_enabled<65><64><EFBFBD>raw_enabled<65>K<01>line_length_limit<69>M'<27>pep_references<65>N<EFBFBD>pep_base_url<72><6C>https://peps.python.org/<2F><>pep_file_url_template<74><65>pep-%04d<34><64>rfc_references<65>N<EFBFBD>rfc_base_url<72><6C>&https://datatracker.ietf.org/doc/html/<2F><>	tab_width<74>K<08>trim_footnote_reference_space<63><65><EFBFBD>syntax_highlight<68><74>long<6E><67>smart_quotes<65><73><EFBFBD>smartquotes_locales<65>]<5D><>character_level_inline_markup<75><70><EFBFBD>doctitle_xform<72><6D><EFBFBD>
docinfo_xform<72>K<01>sectsubtitle_xform<72><6D><EFBFBD>
image_loading<6E><67>link<6E><6B>embed_stylesheet<65><74><EFBFBD>cloak_email_addresses<65><73><EFBFBD>section_self_link<6E><6B><EFBFBD>env<6E>Nub<75>reporter<65>N<EFBFBD>indirect_targets<74>]<5D><>substitution_defs<66>}<7D><>substitution_names<65>}<7D><>refnames<65>}<7D><>refids<64>}<7D><>nameids<64>}<7D>j<EFBFBD>j<>s<>	nametypes<65>}<7D>j<EFBFBD><00>sh!}<7D>(j<>hh<hEu<45>
footnote_refs<66>}<7D><>
citation_refs<66>}<7D><>
autofootnotes<65>]<5D><>autofootnote_refs<66>]<5D><>symbol_footnotes<65>]<5D><>symbol_footnote_refs<66>]<5D><>	footnotes<65>]<5D><>	citations<6E>]<5D><>autofootnote_start<72>K<01>symbol_footnote_start<72>K<00>
id_counter<EFBFBD><EFBFBD>collections<6E><73>Counter<65><72><EFBFBD>}<7D><><EFBFBD>R<EFBFBD><52>parse_messages<65>]<5D><>transform_messages<65>]<5D><>transformer<65>N<EFBFBD>include_log<6F>]<5D><>
decoration<EFBFBD>Nhhub.