+ NV-j6^RIt^RIHtHt^RIHt]P3RRlltRR]P3RRllt RR]P3R R llt ]P3R R llt R t ]P3RRllt ]P3RRllt]P3RRllt]P3RRlltRR]P3RRlltR]P3RRlltR#)N)CallableLiteralcV^8dQhR\R\PR\\P.\P3,/#)valuedtypereturnfloatmxDtyperarray)formats"\/Volumes/fast/ai/experiments/ui-tars-smoke/.venv/lib/python3.14/site-packages/mlx/nn/init.py __annotate__r s<  rxxj"(("#caaRVV3RllpV#)aAn initializer that returns an array filled with ``value``. Args: value (float): The value to fill the array with. dtype (Dtype, optional): The data type of the array. Default: ``float32``. Returns: Callable[[array], array]: An initializer that returns an array with the same shape as the input, filled with ``value``. Example: >>> init_fn = nn.init.constant(0.5) >>> init_fn(mx.zeros((2, 2))) array([[0.5, 0.5], [0.5, 0.5]], dtype=float32) cXV^8dQhR\PR\P/#rar r r)rs"rrconstant..__annotate__s"44rxx4BHH4rcJ<\P!VPSSR7#r)r fullshape)rrrs&r initializerconstant..initializerswwqwwU33r)rrrsff rconstantr! s,44 rg?c V^8dQhR\R\R\PR\\P.\P3,/#)rmeanstdrr r )rs"rrr%sD !02 rxxj"(("#rc"aaaRVVV3RllpV#)aAn initializer that returns samples from a normal distribution. Args: mean (float, optional): Mean of the normal distribution. Default: ``0.0``. std (float, optional): Standard deviation of the normal distribution. Default: ``1.0``. dtype (Dtype, optional): The data type of the array. Default: ``float32``. Returns: Callable[[array], array]: An initializer that returns an array with the same shape as the input, filled with samples from a normal distribution. Example: >>> init_fn = nn.init.normal() >>> init_fn(mx.zeros((2, 2))) array([[-0.982273, -0.534422], [0.380709, 0.0645099]], dtype=float32) cXV^8dQhR\PR\P/#rr)rs"rrnormal..__annotate__>s&QQrxxQBHHQrc^<\PPVPSSSR7#)rscalelocr)r randomnormalr)rrr$r%s&rrnormal..initializer>s%yyaggSd%PPrr )r$r%rrsfff rr.r.%s2QQ rc V^8dQhR\R\R\PR\\P.\P3,/#)rlowhighrr r )rs"rrrDsD !02 rxxj"(("#rc"aaaRVVV3RllpV#)aAn initializer that returns samples from a uniform distribution. Args: low (float, optional): The lower bound of the uniform distribution. Default: ``0.0``. high (float, optional): The upper bound of the uniform distribution. Default: ``1.0`` dtype (Dtype, optional): The data type of the array. Default: ``float32``. Returns: Callable[[array], array]: An initializer that returns an array with the same shape as the input, filled with samples from a uniform distribution Example: >>> init_fn = nn.init.uniform(low=0, high=1) >>> init_fn(mx.zeros((2, 2))) array([[0.883935, 0.863726], [0.617261, 0.417497]], dtype=float32) cXV^8dQhR\PR\P/#rr)rs"rruniform..__annotate__]s&BBrxxBBHHBrc^<\PPSSVPSR7#r)r r-uniformr)rrr2r1s&rruniform..initializer]s%yy  dAGG5 AArr )r1r2rrsfff rr7r7Ds2BB rcV^8dQhR\PR\\P.\P3,/#rrr )r r rr)rs"rrrcs2BHHhz2887K.LrcaRV3RllpV#)aAn initializer that returns an identity matrix. Args: dtype (Dtype, optional): The data type of the array. Default: ``float32``. Returns: Callable[[array], array]: An initializer that returns an identity matrix with the same shape as the input. Example: >>> init_fn = nn.init.identity() >>> init_fn(mx.zeros((2, 2))) array([[1, 0], [0, 1]], dtype=float32) cXV^8dQhR\PR\P/#)rarrr r)rs"rridentity..__annotate__vs"333bhh3rc<VP^8wg*VP^,VP^,8wd\RVP R24h\P!VP^,SR7#)rz6The input array must be a square matrix but got shape .)nr)ndimr ValueErrorr eye)r=rs&rridentity..initializervs^ 88q=CIIaLCIIaL8H STU vv ! E22rr rrsf ridentityrGcs&33 rc*VP^8d\RVP R24hVPR,pVP^,pVP^8d1^pVP^RF pW4,pK W,pW#,pW3#)rzPGlorot / He initialization requires at least 2 dimensional input but input with z dimensions.)rBrCr)xfan_infan_outreceptive_fieldds& r_calculate_fan_in_fan_outrOsvvz  vvhl 4  WWR[FggajGvvz2A  O)+ ?rcV^8dQhR\PR\\P\.\P3,/#r:r r rrr )rs"rrr7(( 88( rxx)*(rcaRRV3RlllpV#)aA Glorot normal initializer. This initializer samples from a normal distribution with a standard deviation computed from the number of input (``fan_in``) and output (``fan_out``) units according to: .. math:: \sigma = \gamma \sqrt{\frac{2.0}{\text{fan\_in} + \text{fan\_out}}} For more details see the original reference: `Understanding the difficulty of training deep feedforward neural networks `_ Args: dtype (Dtype, optional): The data type of the array. Default: ``float32``. Returns: Callable[[array, float], array]: An initializer that returns an array with the same shape as the input, filled with samples from the Glorot normal distribution. Example: >>> init_fn = nn.init.glorot_normal() >>> init_fn(mx.zeros((2, 2))) array([[0.191107, 1.61278], [-0.150594, -0.363207]], dtype=float32) >>> init_fn(mx.zeros((2, 2)), gain=4.0) array([[1.89613, -4.53947], [4.48095, 0.995016]], dtype=float32) cdV^8dQhR\PR\R\P/#rrgainr r rr )rs"rr#glorot_normal..__annotate__s.GGrxxGuGrxxGrc<\V4wr#V\P!RW#,, 4,p\PP VP VSR7#)g@rr+r)rOmathsqrtr r-r.r)rrVrKrLr%rs&& rr"glorot_normal..initializersJ3A6TYYsf&6788yyaggSFFrr"r rFsf r glorot_normalr_sFGG rcV^8dQhR\PR\\P\.\P3,/#r:rQ)rs"rrrrRrcaRRV3RlllpV#)aA Glorot uniform initializer. This initializer samples from a uniform distribution with a range computed from the number of input (``fan_in``) and output (``fan_out``) units according to: .. math:: \sigma = \gamma \sqrt{\frac{6.0}{\text{fan\_in} + \text{fan\_out}}} For more details see the original reference: `Understanding the difficulty of training deep feedforward neural networks `_ Args: dtype (Dtype, optional): The data type of the array. Default: ``float32``. Returns: Callable[[array, float], array]: An initializer that returns an array with the same shape as the input, filled with samples from the Glorot uniform distribution. Example: >>> init_fn = nn.init.glorot_uniform() >>> init_fn(mx.zeros((2, 2))) array([[0.223404, -0.890597], [-0.379159, -0.776856]], dtype=float32) >>> init_fn(mx.zeros((2, 2)), gain=4.0) array([[-1.90041, 3.02264], [-0.912766, 4.12451]], dtype=float32) cdV^8dQhR\PR\R\P/#rUrW)rs"rr$glorot_uniform..__annotate__s.FFrxxFuFrxxFrc<\V4wr#V\P!RW#,, 4,p\PP V)W@P SR7#)g@r)rOr[r\r r-r7r)rrVrKrLlimitrs&& rr#glorot_uniform..initializersL3A6tyy(8!9::yy  %u EErr^r rFsf rglorot_uniformrgsFFF rcV^8dQhR\PR\\P\R,\ .\P3,/#rrr rKrLr r rrrr )rs"rrrs?77 887 rxx!45u=rxxGH7rcaRRV3RlllpV#)aBuild a He normal initializer. This initializer samples from a normal distribution with a standard deviation computed from the number of input (``fan_in``) or output (``fan_out``) units according to: .. math:: \sigma = \gamma \frac{1}{\sqrt{\text{fan}}} where :math:`\text{fan}` is either the number of input units when the ``mode`` is ``"fan_in"`` or output units when the ``mode`` is ``"fan_out"``. For more details see the original reference: `Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification `_ Args: dtype (Dtype, optional): The data type of the array. Default: ``float32``. Returns: Callable[[array, str, float], array]: An initializer that returns an array with the same shape as the input, filled with samples from the He normal distribution. Example: >>> init_fn = nn.init.he_normal() >>> init_fn(mx.zeros((2, 2))) # uses fan_in array([[-1.25211, 0.458835], [-0.177208, -0.0137595]], dtype=float32) >>> init_fn(mx.zeros((2, 2)), mode="fan_out", gain=5) array([[5.6967, 4.02765], [-4.15268, -2.75787]], dtype=float32) c~V^8dQhR\PR\R,R\R\P/#rrmoderVr rjr rrr )rs"rrhe_normal..__annotate__sDGG 88G)*GG  Grc<\V4wr4VR8XdTpMVR8XdTpM\RV R24hV\P!V4, p\P P VPVSR7#)rKrLInvalid mode: ". Valid modes are: fan_in, fan_outrZ)rOrCr[r\r r-r.r)rrorVrKrLfanr%rs&&& rrhe_normal..initializersr 4A6 8 C Y C~dV3UVW WTYYs^#yyaggSFFrrKr"r rFsf r he_normalrxsNGG rcV^8dQhR\PR\\P\R,\ .\P3,/#rirk)rs"rrr%s?99 889 rxx!45u=rxxGH9rcaRRV3RlllpV#)aA He uniform (Kaiming uniform) initializer. This initializer samples from a uniform distribution with a range computed from the number of input (``fan_in``) or output (``fan_out``) units according to: .. math:: \sigma = \gamma \sqrt{\frac{3.0}{\text{fan}}} where :math:`\text{fan}` is either the number of input units when the ``mode`` is ``"fan_in"`` or output units when the ``mode`` is ``"fan_out"``. For more details see the original reference: `Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification `_ Args: dtype (Dtype, optional): The data type of the array. Default: ``float32``. Returns: Callable[[array, str, float], array]: An initializer that returns an array with the same shape as the input, filled with samples from the He uniform distribution. Example: >>> init_fn = nn.init.he_uniform() >>> init_fn(mx.zeros((2, 2))) # uses fan_in array([[0.0300242, -0.0184009], [0.793615, 0.666329]], dtype=float32) >>> init_fn(mx.zeros((2, 2)), mode="fan_out", gain=5) array([[-1.64331, -2.16506], [1.08619, 5.79854]], dtype=float32) c~V^8dQhR\PR\R,R\R\P/#rnrp)rs"rr he_uniform..__annotate__NsDFF 88F)*FF  Frc<\V4wr4VR8XdTpMVR8XdTpM\RV R24hV\P!RV, 4,p\P P V)W`PSR7#)rKrLrsrtg@r)rOrCr[r\r r-r7r)rrorVrKrLrurers&&& rrhe_uniform..initializerNsy 4A6 8 C Y C~dV3UVW Wtyys++yy  %u EErrwr rFsf r he_uniformr%sRFF rc V^8dQhR\R\R\R\PR\\P.\P3,/#)rsparsityr$r%rr r )rs"rrrasR,,, , , 88 , rxxj"(("# ,rc&aaaaRVVVV3RllpV#)aAn initializer that returns a sparse matrix. Args: sparsity (float): The fraction of elements in each column to be set to zero. mean (float, optional): Mean of the normal distribution. Default: ``0.0``. std (float, optional): Standard deviation of the normal distribution. Default: ``1.0``. dtype (Dtype, optional): The data type of the array. Default: ``float32``. Returns: Callable[[array], array]: An initializer that returns an array with the same shape as the input, filled with samples from a normal distribution. Example: >>> init_fn = nn.init.sparse(sparsity=0.5) >>> init_fn(mx.zeros((2, 2))) array([[-1.91187, -0.117483], [0, 0]], dtype=float32) cXV^8dQhR\PR\P/#rr)rs"rrsparse..__annotate__s"  rxx BHH rc<VP^8wd \R4hVPwr\\P !SV,44p\ P!\ PPVPR7^R7p\ PPVPSSSR7p^V\ P!V4PV^4VRRV13,3&V#)rz,Only tensors with 2 dimensions are supportedr)axisr*:NNNN) rBrCrintr[ceilr argsortr-r7r.arangereshape) rrowscols num_zerosorderrr$rr%s & rrsparse..initializers 66Q;KL LWW  (T/23  299,,177,;!D II  177#4u  MDE"))D/ ! !$ *E!ZiZ-,@ @Arr )rr$r%rrsffff rsparseras<   rcV^8dQhR\R\PR\\P.\P3,/#)rrVrr r )rs"rrrs<)) ) hh) rxxj"(("#)rcaaRVV3RllpV#)acAn initializer that returns an orthogonal matrix. Args: gain (float, optional): Scaling factor for the orthogonal matrix. Default: ``1.0``. dtype (Dtype, optional): Data type of the array. Default: ``float32``. Returns: Callable[[array], array]: An initializer that returns an orthogonal matrix with the same shape as the input. cXV^8dQhR\PR\P/#rr)rs"rr orthogonal..__annotate__s"rxxBHHrc<VP^8wd \R4hVPwr\W4p\P P W33R7p\PPV\PR7wrV\P!V4pV\P!V4,pVRV1RV13,pVS ,pVPS4#)rzHOrthogonal initialization requires a 2D array but got a {a.ndim}D array.r)streamN) rBrCrmaxr r-r.linalgqrcpudiagsignastype) rrrrArmatqrrNrrVs & rrorthogonal..initializers 66Q;&  WW  Oyyqf-yy||D|0 GGAJ  N eteUdUlO Hxxrr )rVrrsff r orthogonalrs4 r)r[typingrrmlx.corecorer float32r!r.r7rGrOr_rgrxrrrr rrrs $%'JJ:C2::@C2::> "zz:,jj(Xjj(Xjj7vjj9|jj ,`))r