o
    h                 -   @   s6  d dl Z d dlZd dlmZ d dlmZ d dlmZmZ d dl	m
Z
mZmZmZ d dlmZ d dlZd dlmZ d dlmZmZmZ d dlmZmZmZmZ d d	lmZ d d
lmZm Z m!Z! d dlm"Z"m#Z#m$Z$m%Z%m&Z&m'Z'm(Z(m)Z)m*Z*m+Z+m,Z,m-Z-m.Z. d dl/m0Z0m1Z1m2Z2m3Z3m4Z4 d dl5m6Z6m7Z7 d dl8m9Z: d dl;m<Z= edZ>edZ?ej@jAZAejBCdddZDeEd\ZFZGZHde
e
e?e>f ge
e?e>f f fddZIde'fddZJdd ZKdd ZLeIeAjMeAjNge4 dddejOd d fd!d"ZPeIeAjQjReAjQjSge4 d#d$ ZTeIeAjUjReAjUjSge4 d%d&d'd(ZUdhd)d*ZVd+d, ZWdid.d/ZXeIeAjYjRd0d1 ZZeIeAj[e4 d2d3 Z[eIeAj\jReAj\jSeAj]jReAj]jSge4d4d5d6d7 Z^eIeAj_jReAj_jSge4 d8d9 Z_d:d; Z`dhd<ed=eaeb d>ecfd?d@ZdeIeAjejReAjejSge4 dAdB ZfdZgd=eaeb fdCdDZheIeAjijReAjijSge4 dEdF ZjeIeAjkjlddGdHdIZmeIeAjkjRejnddddJdKdLZoeIeAjpjReAjpjSge4 ejnddddJdMdNZqeIeAjpjreAjpjsge4 ejnddddJdOdPZteIeAjujReAjujSge4 dddddJdQdRZveIeAjwjReAjwjSge4 d<ed=eaeb dSebdTebfdUdVZxeIeAjyjRdhdWdXZzdYdZ Z{eIeAj|jRd[d\ Z}eIeAj~			djd]ed^ed_ed`ee daee dbeej fdcddZeIeAj	dkdeedfedgedbeej fdhdiZeIeAjdjdjddkd]edeedfedgedbeej f
dldmZeIeAj				 	 	j	%dldnejdoejd`ee dpee dbeej dqecdrebdsebdtebfdudvZeIeAjjRd-dwd<ed=ebdxedyejdzed{ecdefd|d}ZeIeAjjRd-dwd<ed=ebdxedyejdzed{ecdefd~dZe4 eIeAjjRdd ZeIeAjjRdddd d dddedzedee d5ee dee debdecdefddZeIeAjjReAjjge4 dd ZeIeAjjdhddZeIeAjjReAjjge4 dd ZeIeAjjdhddZeIeAjjRdd ZeIeAjjSdd ZeIeAjjRdd ZeIeAjjdd ZeIeAjjRdd ZeIeAjjRddddddddZeIeAjjRdmddZeIeAjjRdjddZeIeAjjRdmddZeIeAjjRdd ZeIeAjjdd Zd<edefddZd<ededefddZ	-didededecfddZdndededefddZdededecdefddZ	dodeded]edefddZdefddZeIeAjjReAjjge4ddădpdededecfddȄZeIeAjjReAjjSge4 d]edefddʄZeIeAjge4ddăd]efdd̄ZdedefddτZeIeAje4 d<ededecdefdd҄ZeIeAje4 dhd<ededecdefddԄZeIeAje4 dhd<edecdefddքZeIeAje4 dhd<edecdefdd؄ZeIeAjjRdqdedecdecfddۄZeIeAjjReAjjSge4 d]ededefddބZeIeAjjRdhdedecfddZeIeAjjReAjjSge4dddd d dd<edecdecdeeeef fddZeIeAjjReAjjSge4 d ddedededecdef
ddZeIeAjjReAjjSge4dddd-ddedecdeeeef fddZeIeAjjReAjjSge4dddd-d ddedecdecdeeeef fddZeIeAjjReAjjSge4 d-d ddedededecdecdefddZeIeAj΃e4ddd	-	-drdededecdecdeeeef f
ddZdedeececf fddZeIeAjjReAjjSge4d ddsdededeeef fddZeIeAjjReAjjge4dddddedeeeeef fddZeIeAjjR	 	-	dtded	ecd
ecdee fddZ֐dededeeaeb eaeb f fddZאdededee deeef fddZd]ededecfddZeIeAjڃd-d ddddddededecdecdee dee dee dee deeeeef fddZeIeAjjReAjjSgd-d dddededecdecdecdee defddZeIeAj݃e4dd d-d!	-	 	 dud<ededecd"ecdecdeeef fd#d$ZeIeAjjRd%d& ZeIeAje4 	-	 dvd]edededecd"ecdefd'd(Zd)d* Zd+d, ZeIeAje4 d-d. ZeIeAje4 d/d0 Zd1d2 ZeIeAje4d3d4d5 ZeIeAje4d3d6d7 Zd8d9 ZeIeAje4 d:d; ZeIeAje4 d<d= ZeIeAjjReAjjeAjjReAjjge4d3d>d? Zd@dA ZeIeAje4 dBdC ZeIeAje4 dDdE ZeIeAjjReAjjeAjjReAjjge4d3dFdG ZeIeAje4 dwd<edIedefdJdKZ eIeAje4 dLed<edIedMedef
dNdOZeIeAjjReAjjSge4d-d!djdjdPdQdRZeIeAjjReAjjSge4 ddGdSdTZeIeAjjdxdVdWZeIeAjj	dxdXdYZ
eIeAjjReAjjSge4 dkdZd[ZeIeAjjR	 	 dqd\d]ZeIeAje4d-d!d^d_ Zd`da ZdydcddZ	dkdeejd^ejdfeeaeb ebf dgeeaeb ebf dheeaeb ebf diecdjebdkeeeaeb ebf  fdldmZdndo ZeIeAjjRdeejd^ejd`eej dpeej dqeej drecdsedtefdudvZeIeAjjRdeejd^ejd`ejdfeaeb dgeaeb dheaeb diecdkeaeb djebfdwdxZejj
rejBCdyddZeIej@jjjRdzd{ ZeIej@jjjRd|d} Z ejj!	rejBCd~ddZ"eIej@j#j$dd Z%ejBCdddZ&eIej@j'j(jReIej@j'j)jRdd Z*eIej@j'j(j+dd Z,eIej@j'j-jReIej@j'j-j.dd Z/eIej@j'j-j+eIej@j'j-j0dd Z1eIej@j'j2jReIej@j'j3jRdd Z4ejBCdddZ5eIej@j6j7				 dzddZ8eIej@j6j9dd Z:dd Z;eIeAj<jR			 	-	d{ddZ=dd Z>eIeAj?jRdd Z@eIeAjAe4 			 	-	d{ddZBeIeAjCe4d3dd ZDeIeAjEjRdd ZFeIeAjGjRdd ZHeIeAjIjRdd ZJeIeAjKe4d3dd ZLdedefddZMeIeAjNe4dd5dd ZOeIeAjPe4d3dd ZQeIeAjRe4dd5dd ZSeIeAjTe4d3dd ZUeIeAjVjdkddZWeIeAjXjReAjXjSge4 dd ZYeIeAjZjReAjZjSge4 d%ddebfddZZeIej@jAj[jRej@jAj[jSge4 dd Z[eIeAj\jeAj]jgdd Z^eIeAj_jRgdd Z`eIeAjajReAjajSge4d-d!djdjdPddZbeIeAjcjgdÐdĄ ZdeIeAjejReAjfjRgdddŜdƐdǄZgeIeAjhjRgdddŜdȐdɄZieIeAjjge4 dʐd˄ ZkeIeAjlgd̐d̈́ ZmeIeAjngdΐdτ ZoeIeAjpgdАdф ZqeIeAjrgdҐdӄ ZseIeAjtgdԐdՄ ZtdebdebdebfdؐdلZudڐdۄ ZveIeAjwgd`ee fdܐd݄ZxeIeAjygdސd߄ ZzeIeAj{gdd Z|eIeAj}jRdd Z~eIeAje4 dd ZeIeAjjR	 	 	 		 	%d|ddZeIeAjjRdd ZdiddZeIeAjjReAjjSge4 d}ddddZeIeAjjReAjjRgdd ZeIeAjjeAjjeAjjeAjjeAjjReAjjge4d4d5d~ddZeIeAjjRdd ZeIeAjjRdd ZeIeAjjRdd ZeIeAjjeAjjeAjjeAjjeAjjReAjjReAjjRgdd ZeIeAjjeAjjeAjjeAjjgdddZeIeAjjReAjjgdd Zdd  ZeIeAjjeAjjgdd ZeIeAjjeAjjgdd ZeIeAjjRdd ZeIeAjjeAjjgdd ZeIeAjjeAjjgd	d
 ZeIeAjjRdd ZeIeAjje4 ddefddZeIeAjge4 	dddZeIeAjg	dddZeIeAjg	dddZeIeAjjReAjjRgdhddZeIeAjjdd ZeIeAjjRdd ZeIeAjdd ZeIeAje4 dd  ZeIeAjd!d" ZeIeAjjRdhd#d$ZeIeAjjRd%d& Zǐdmd'd(ZeIeAjjRd)d* ZeIeAjjd+d, Zːd-d. Z̐d/d0 Z͐d1d2 Zΐd3d4 Z	 dhd]ed5ebd6ebd7ebd8ebd9ebd:ebd;ebd<ebd=ebd>ebd?ebd@ebdAebdBebdCebdDebdEebdFebdGebdedHecf,dIdJZАdKdL Zd]eded5ebd6ebd7ebd8ebd9ebd:ebd;ebd<ebd=ebd>ebdBebdCebdDebdEebdFebdGebdef&dMdNZҐdOdP ZeIeAjjRdQdR ZeIeAjjR				 dzdSdTZeIeAjjRdUdV ZeIeAjڃe4dd5				 dzdWdXZeIeAj܃e4d3dYdZ Zd]ed[efd\d]ZG d^d_ d_eZd]ed[ed`ebfdadbZeIeAjjRdcdd ZeIeAje4 dedf ZeIeAje4d3dgdhdi ZeIeAjjRgdjdk ZeIeAjjR					ddldmZeIeAjjReAjjSge4 ddddd dndodpZeIeAjjReAjjSge4 ddddd dndqdrZeIeAjjbdsdt ZeIeAjjRdudv ZeIeAjjRddwdxZdid=ebdyebdzecfd{d|Zd}d~ Zdd ZeIeAjjRdhddZdhddZdkddZdd ZdkddZdddZeIeAjjRdd ZeIeAjdd ZeIeAjj eAjjeAjjeAjjge4 dkddZeIeAjj eAjjeAjjeAjjgdkddZeIeAjg		 	 	ddededededecdecdee fddZdedeebdf fddZeIeAj	g		 	 	ddedededee decdedecdecdee fddZ
eIeAjg			 	 	ddedededee dedecdecdee fddZeIeAjg	dkdededededededededebdebdedecdededee fddZeIeAjg		 		ddededededecdee dee fddZeIeAjg		dmdedededededededecdee dee fddZeIeAjg		 	ddedededee decdecdee fddZeIeAjg	 	ddededededee dedededededeaec decdee fddZeIeAjg	dkdedededededededededededebdebdedecdee f ddZeIeAjg					ddedededee dee debdebdedecdecdee deeb deeb dee dee fdÐdĄZeIeAjg			djdededededededededebdebdedecdededee deeb deeb f"dŐdƄZeIeAjg	 				ddededed`ee dee dee deeb deeb dedebdecdee dee dee deeb fdϐdЄZeIeAjg			 ddedededed`ee dee dee dej dej dededededebdecdee deeb decf$dԐdՄZ!eIeAj"jRg				 dd<ejdgejdejdejd`eej deej dbeej decfdڐdۄZ#eIeAj$j%eAj$j&ge4 didܐd݄Z'eIeAj(j%didސd߄Z)eIeAj*jReAj*jSge4 dhddGddZ+dd Z,dd Z-eIeAj.jReAj/jRgdkddZ.eIeAj0jReAj1jRgdmddZ0eIeAj2jReAj3jRg		dmdedeeebej f  deeebej f  dee dee f
ddZ2eIeAj4jReAj5jRgdjddZ4eIeAj6jReAj6j7eAj6jeAj6j8gdddZ9dd Z:eIeAj;jR		dmddZ<eIeAj=jRdd Z=eIeAj>jRdd Z>dd Z?dd Z@eIeAjAjReAjBjRgd}d dZCeIeAjDjRdddZDeIeAjEjRdddZFeIeAjGe4 	dddZHeIeAjIjReAjIjge4d4d5d~dd	ZJejKZLd
d ZMeIeAjNjRdd ZNeIeAjOjRdd ZOeIeAjPjRdd ZQeIeAjRjRdd ZReIeAjSjeAjSjTge4 d d dddZUeIeAjVge4 dddZWeIeAjXjReAjYjRg		dmddZZeIeAj[jRg		dmddZ\eIeAj]jRdd Z]eIeAj^jReAj^jSge4 djd d!Z^eIej@jAj_d"d# Z_eIej@jAj`d$d% Z`eIeAjae4 d d ddd&d'd(Zbd)d* ZceIeAjdd+d, ZeeIeAjf	%dd-d.ZgeIeAjh	%dd/d0ZieIeAjj	%dd1d2ZkeIeAjle4 d d d3d4d5ZmeIeAjne4 d6ebd<edefd7d8ZoeIeAjpd<efd9d:ZqeIeAjre4d-d!d<edefd;d<ZreIeAjse4 d<edefd=d>Zsd?d@ Zt					 ddAedBedeej deej dCee d`ee deej dbeej decfdDdEZueIeAjve4 			djdAedBedCee d`ee dbeej defdFdGZweIeAjxjRg					 ddAejdBejdejdejdCeej d`eej deej dbeej decfdHdIZyeIeAjze4 dJed=ebdKecdefdLdMZ{eIeAj|e4 ddNdOZ}eIeAj~e4 	%	 	 dd^ed5edPebdQecdRecdefdSdTZ~eIeAjjR	dd4edeae dUeaeb dVefdWdXZdYdZ Zd[d\ ZeeAj eeAj eeAj eeAj eeAj eeAj eeAj eeAj eeAj eeAj eeAj eeAj eeAj eeAj eeAj eeAj eeAj eeAj eeAj eeAj eeAj d]d^ ZeIeAje4 d_d` ZeIeAje4 djdadbdcZeIeAje4 djdadddeZeeAjZeeAjZeeAjZd dl5Zd dlZd dlZdfdg Ze  dS (      N)Sequence)Enum)reducewraps)CallableOptionalTypeVarUnion)	ParamSpec)SymBoolSymFloatTensor)_add_op_to_registry_convert_out_paramsglobal_decomposition_table
meta_table)
OpOverload)_prim_elementwise_meta$ELEMENTWISE_PRIM_TYPE_PROMOTION_KINDview_of)BoolLikecorresponding_complex_dtypecorresponding_real_dtypedefinitely_contiguouselementwise_dtypesELEMENTWISE_TYPE_PROMOTION_KIND	FloatLikeIntLikeis_contiguousmake_contiguous_strides_forNumbersuggest_memory_format
TensorLike)_maybe_convert_to_dtype_maybe_resize_out_resize_output_check_safe_copy_outout_wrapper)_broadcast_shapes_maybe_broadcast)_config)_pytree_T_PatenZIMPLMeta   returnc                    s    fdd}|S )Nc                    s$   t    fdd}t|  S )Nc                    s   t t|   d S N)r   r   opfn _/home/www/facesmatcher.com/frenv_anti/lib/python3.10/site-packages/torch/_meta_registrations.pyregisterA      z0register_meta.<locals>.wrapper.<locals>.register)r   pytreeZ	tree_map_)r6   r9   r3   r5   r8   wrapper>   s   zregister_meta.<locals>.wrapperr7   )r4   r<   r7   r3   r8   register_meta=   s   	r=   type_promotionc                    s>   t j|d| i\}  fdd|D }t| }t|dtjiS )Ntype_promotion_kindc                    s   g | ]}t | qS r7   )r#   .0xresult_dtyper7   r8   
<listcomp>S       z$elementwise_meta.<locals>.<listcomp>r>   )utilsr   r)   r   r   DEFAULT)r>   args_r7   rC   r8   elementwise_metaJ   s   
rK   c                 C   s(   t jt jt jt jt jt ji}|| | S r2   )torchZ	complex32halfcfloatfloatcdoubledoubleget)dtypeZfrom_complexr7   r7   r8   toRealValueType^   s
   rT   c                    s2   t tg|R   t k fdd d S )Nc                         d d  S )Nzoutput with shape z# doesn't match the broadcast shape r7   r7   Zbroadcasted_shape
self_shaper7   r8   <lambda>k       z)check_inplace_broadcast.<locals>.<lambda>)tupler(   rL   _check)rW   Z
args_shaper7   rV   r8   check_inplace_broadcastg   s
   r\   Fc	           	         s  t tjrt dkdd  t tjr$t dkdd  tdd fD rMtt  d u r> qStt	 fdd npRt t tj
s[J tt tfdd t tsqJ tdkd	d  tjf|d
||dS )Nr   c                   S      dS Nz:linspace only supports 0-dimensional start and end tensorsr7   r7   r7   r7   r8   rX          z(meta_linspace_logspace.<locals>.<lambda>c                   S   r]   r^   r7   r7   r7   r7   r8   rX      r_   c                 s   s    | ]}t |tV  qd S r2   )
isinstancecomplex)rA   argr7   r7   r8   	<genexpr>       z)meta_linspace_logspace.<locals>.<genexpr>c                         d  d S )Nzlinspace(): inferred dtype z& can't be safely cast to passed dtype r7   r7   )default_complex_dtyperS   r7   r8   rX      rY   c                      s*   dt j dt  j dt j dS )Nz4received an invalid combination of arguments - got (, ))type__name__r7   )endstartstepsr7   r8   rX      s    c                   S   r]   )Nz$number of steps must be non-negativer7   r7   r7   r7   r8   rX      r_   metarS   layoutdevice
pin_memoryrequires_grad)r`   rL   r   r[   dimanyrG   r   get_default_dtypeis_complex_dtyperS   _check_typer   empty)	rl   rk   rm   baserS   rq   rp   rr   rs   r7   )rf   rS   rk   rl   rm   r8   meta_linspace_logspaceo   sH   

r{   c                    sN   t  jt jk fdd t |  dko  dk dd  |  jS )Nc                         d j  S )Nz2take(): Expected a long tensor for index, but got rS   r7   indexr7   r8   rX          zmeta_take.<locals>.<lambda>r   c                   S   r]   )Nz*take(): tried to take from an empty tensorr7   r7   r7   r7   r8   rX      r_   )rL   r[   rS   long_check_indexnumel	new_emptyshape)selfr   r7   r~   r8   	meta_take   s   

r   rt   c                   sh   j }j }t||kdd  t dko dk fdd tjj}|S )Nc                   S   r]   )Nz=linalg.cross: inputs must have the same number of dimensions.r7   r7   r7   r7   r8   rX      r_   zlinalg_cross.<locals>.<lambda>r0   c                      s"   d  d   d   S )Nzlinalg.cross: inputs dimension z must have length 3. Got  and sizer7   rt   otherr   r7   r8   rX      s
   )ndimrL   r[   r   r(   r   r   )r   r   rt   Zx_dZy_d	out_shaper7   r   r8   linalg_cross   s   
r   c                    s  ddl m mm}  fdd}fdd}t| dkr%dgt| S ttj| d}||dk}|r=||| |r=|S dgt| }	|rstt|d ddD ] }
|
t|d kr_d|	|
< qPt	||
d  d|	|
d   |	|
< qP|	S t|d }
|d }d}d}tt| d ddD ]m}|| | 9 }|dks|| |d  dkr|||d  || kr|
dkr|||k s|||
 dkr|| |	|
< |||
 9 }|
d8 }
|
dkr|||k s|||
 dks|||kr d S |dkr||d  }d}d}q|
dkrd S |	S )	Nr   )guard_or_falseguard_or_truesym_eqc                       r | S | S r2   r7   rB   )r   size_obliviousr7   r8   maybe_guard_or_false      z-_compute_stride.<locals>.maybe_guard_or_falsec                    r   r2   r7   r   )r   r   r7   r8   maybe_guard_or_true   r   z,_compute_stride.<locals>.maybe_guard_or_true   r   )
%torch.fx.experimental.symbolic_shapesr   r   r   lenr   operatormulrangemax)Z	old_shapeZ
old_stride	new_shaper   r   r   r   r   Z
zero_numel
new_strideZview_dZchunk_base_strideZtensor_numelZ
view_numelZtensor_dr7   )r   r   r   r8   _compute_stride   sj   


r   c                    sV   ddl m  t fdd|  D p*t fdd|  D p*t fdd|D S )Nr   has_hintc                 3       | ]} | V  qd S r2   r7   rA   sr   r7   r8   rc     rd   z+_view_has_unbacked_input.<locals>.<genexpr>c                 3   r   r2   r7   r   r   r7   r8   rc     rd   c                 3   r   r2   r7   r   r   r7   r8   rc     rd   )r   r   ru   r   stridear   r7   r   r8   _view_has_unbacked_input  s   r   Tc                    s  ddl m}m} tjddt   jdkr; }D ]}t	|dk tj
|d}q | u r9t S |S tdkra } jD ]}t	|dk tj
|d}qF| u r_t S |S ttjd}t	  |k fdd tt jkr|| jrt S |rt rnt rt} |S t    |d	}	|	d ur |	S |rtjjjjst rt dd
S d j d   d d}
t |
)Nr   )r   r   F)validater   r   c                         d j  d dS )Nz&Could not reshape a tensor with shape  as a tensor with shape !r   r7   r   r7   r8   rX   E      z%_view_unbacked_meta.<locals>.<lambda>)r   )size_oblivious_enabledz Cannot view a tensor with shape z and strides r   r   )!r   r   r   rG   Zextract_shape_from_varargsZ
infer_sizer   r   rL   r[   _refs	unsqueezer   r   r   Zsqueezer   r   r   r   r   r   
as_stridedr   r   r   fxexperimentalr*   backed_size_obliviousr   _view_unbacked_meta
ValueError)r   r   r   r   r   Z_alengthZshape_numelstridesZnew_stridesmsgr7   r   r8   r   !  sT   


"

r   c                 G   s:   t jjjjst| |rt| |S t jj| g|R ddiS )NZ
allow_copyF)	rL   r   r   r*   r   r   r   r   Z_reshape_view_helperr   r7   r7   r8   
_view_metae  s
   
r   c                 C   s$   t | d t| d tj| tjdS )Nzlinalg.matrix_expmemory_format)squareCheckInputscheckFloatingOrComplexrL   
empty_likecontiguous_formatr   r7   r7   r8   linalg_matrix_expo  s   

r   valuesindicesc                 C   sV   t j| j| j| jd}t j| j| jt jd}|  dkr'| jdkr't|| j ||fS )Nrq   rS   r   )	rL   ry   r   rq   rS   int64r   r   maybe_wrap_dim)r   rt   r   r   r7   r7   r8   	cummaxminw  s
   r   c                 C   s   t || j tj| tjdS Nr   )r   r   rL   r   r   )r   rt   r7   r7   r8   logcumsumexp  s   r   c                   s  |j }t|}|| }tt|}dd t|D }	|D ]}
d|	|
< qg g }}|D ]}
|	|
 s6||
 q*||
 q*|| }t|}|  |d | }|j fdddd |||d   }||}dgt|j|d   }|	|}|
d}||d< t|}tt|D ]}|||  ||d	 < q| j|tjd
 dd t|D }d	}|d	 }|dkr|| d ||| < ||||  9 }|d	8 }|dkst||D ]}| d	||  ||| < q| |||   | S )Nc                 S      g | ]}d qS Fr7   rA   rJ   r7   r7   r8   rE     rY   z_exec_fft.<locals>.<listcomp>Tc                        |  S r2   r7   r   Zself_stridesr7   r8   rX         z_exec_fft.<locals>.<lambda>keyreverser   r   r   r   c                 S   r   r   r7   r   r7   r7   r8   rE     rY   )r   r   listr   appendr   sortpermuter   Zreshaper   resize_rL   r   as_strided_storage_offset)outr   	out_sizesrt   forwardr   Zsignal_ndim
batch_dimsZdim_permuteZis_transformed_dimdleftrightZ	batch_endtmpinputZbatched_sizes
batch_sizeZbatched_out_sizesiZout_stridesZbatch_numelr7   r   r8   	_exec_fft  sN   




r   r   rt   exclude_lastc                    s<   t |}|   |d t|t|  j fddd |S )Nc                    r   r2   r7   r   r   r7   r8   rX     r   z_sort_dims.<locals>.<lambda>)r   )r   r   r   intr   )r   rt   r   sorted_dimsr7   r   r8   
_sort_dims  s   
r   c                 C   sH   t | jj |s|  S t| |}| |  }t|| |  ||dS )Nr   )	rL   r[   rS   
is_complexcloner   r   r   r   )r   rt   normalizationr   r   r   r7   r7   r8   meta_fft_c2c  s   
r   c                 C   s8   t | tkst | dkr| d dkr| d dkrdS dS )N   r   r   FT)r   cufft_max_ndimr   r7   r7   r8   use_optimized_cufft_path  s   0r   c                    s  t | jj t|  }t|}|d }|| d d }t|}|||< |r+|||< t| dks7t| dkr| j|t	| jd}	| }
t| dkrXt
|rXt|	|
||dd ngt|dkr`|n|}t|	|
||gdd t|dkr}| j|t	| jd}
|d d }|r|
|	}	}
|
  |j fd	d
dd ttt|}|t|| d  }t|	|
||dd |d t||  }|s|s|	||| kr|
j|t jd |
}	|	S | j|t	| jdS )Nr   r   r   cudaZxpur}   Tr   c                    r   r2   r7   r   r   r7   r8   rX     r   zmeta_fft_r2c.<locals>.<lambda>r   r   )rL   r[   rS   is_floating_pointr   r   device_hintr   rG   r   r   r   r   r   r   minr   r   r   )r   rt   r   Zonesidedinput_sizesr   Zlast_dimZlast_dim_halfsizeZonesided_sizesoutputZworking_tensorZtarget_sizesr   Zmax_dimsZ	last_dimsr7   r   r8   meta_fft_r2c  sX   

r  )	generatorc                C   s   t |t| gS r2   )r$   rL   Size)nr  r   r7   r7   r8   meta_randperm%  s   r  rS   rp   rq   rr   c                C      t j| ||||dS Nr  rL   ry   )r  rS   rp   rq   rr   r7   r7   r8   meta_randperm_default*  s   	
r  c                   s2   dt  k fdd t j|||||dS )Nr   c                      rU   Nz:random_ expects 'from' to be less than 'to', but got from=z >= to=r7   r7   highlowr7   r8   rX   F  rY   zmeta_randint.<locals>.<lambda>r  rL   r[   ry   )r  r   rS   rp   rq   rr   r7   r  r8   meta_randint8  s   
r  c                   s.   t  k fdd t j|||||dS )Nc                      rU   r  r7   r7   r  r7   r8   rX   [  rY   z"meta_randint_low.<locals>.<lambda>r  r  )r  r  r   rS   rp   rq   rr   r7   r  r8   meta_randint_lowM  s   
r  c                C   r  r	  r
  )r   rS   rp   rq   rr   r7   r7   r8   meta_rand_defaultb  s   
r  r   lastdimc           
      C   s*  t | jj t| dkrZt|  }|||d < | j|t| jd}t	|r5t
|| jt jd||ddS t|dkrGt| |d d d|}n| jt jd}t
||||d gddS | }t|dkrv|d d }t| ||dd}|dd  }t| }|||d < | j|t| jd}	t
|	|||ddS )	Nr   r   r}   r   Fr   r   r   )rL   r[   rS   r   r   r   r   r   rT   r   r   r   r   r   r   )
r   rt   r   r  r   r  tempr   Zc2c_dimsr   r7   r7   r8   meta_fft_c2rj  s4   	r  c                 C   sf   ddl m} || st| dkrtdt|tr1|| |}|  | kr1t	j
||   | S )Nr   )free_unbacked_symbolsr   zQmore than one element of the written-to tensor refers to a single memory location)r   r  rL   Z_debug_has_internal_overlapRuntimeErrorr`   r   tor   r.   Zexpand_copydefault)r   srcZnon_blockingr  Zintermediater7   r7   r8   
meta_copy_  s   
r  c                 C   sX   t |  }t |  }||  krdn|| ||  }||d ||| ||fS Nr   )r   r   r   rt   insert)tensorrt   Zresult_sizesZresult_stridesr   r7   r7   r8   inferUnsqueezeGeometry  s    r   c                 C   s0   t ||  d }t| |\}}| || | S r  )r   rt   r   r   )r   rt   Zg_sizesZ	g_stridesr7   r7   r8   meta_unsqueeze_  s   r!  r   weight_metabias_activation_opt	out_dtypec           	      C   s   t | j}|d ur|d|dksJ d|d| dd ks%J |d|d< t| jdks7J dd| df}|d urQ| jtjkrM|tjksQJ d| j||d u r[| jn|d	||}|S )	Nr   zoutput size mismatchr   r   r   z*we can only handle the squashed input case9out_dtype is only supported for i8i8->i32 linear operatorr}   )
r   r   r   r   rS   rL   int8int32r   r   )	r   r"  r#  r$  r%  r&  output_sizesZtransposed_stridesr  r7   r7   r8   meta_sparse_structured_linear  s$   
	r+  mat1	mat1_metamat2c                 C   s   t | jdks	J t |jdksJ t |jdksJ | d|dd ks)J | d|dg}|d urF|jtjkrB|tjksFJ d|j||d u rP|jn|d}|S )Nr   r   r   r'  r}   r   r   r   rS   rL   r(  r)  r   )r,  r-  r.  r&  r*  r  r7   r7   r8   meta_sparse_structured_mm  s   r0  r   )alphabetar&  c          	      C   s   t | jdksJ dt |jdksJ t |jdksJ t |jdks&J | d|dks4J d|d|dd ksBJ |d|dg}|d ur_|jtjkr[|tjks_J d|j||d u ri|jn|d}|S )Nr   zEonly input broadcasted to columns of mat1 * mat2 product is supportedr   r   r'  r}   r/  )	r   r,  r-  r.  r1  r2  r&  r*  r  r7   r7   r8   meta_sparse_structured_addmm  s(   r3  compressed_Adense_Br1  transpose_resultalg_idsplit_ksplit_k_modec	                 C   s  |j tjtjtjtjtjhv sJ d| j |j ksJ dt|jdks(J d| j tjtjfv }	|	r5dnd}
|	rA|	 rAJ d|
d}|
d	}|  d
 |
|  }|d urb||
dksbJ |d urx|	rt|tjtjtjtjhv sxJ d|r~||fn||f}|j||dS )Nz;_cslt_sparse_mm only supports fp16, bf16, int8, and fp8e4m3zinputs must have the same dtyper   z'_cslt_sparse_mm only supports 2d inputs
   	   z.dense input must be transposed for 8bit dtypesr   r      z\out_dtype is not supported for {compressed_A.dtype} x {dense_B.dtype} -> {out_dtype} matmul!r}   )rS   rL   float32float16bfloat16r(  float8_e4m3fnr   r   r   r   r   r)  r   )r4  r5  r$  r1  r&  r6  r7  r8  r9  Zis_8bit_input_typeZcompression_factorkr  moutput_shaper7   r7   r8   meta__cslt_sparse_mm  sB   


rD  )include_selfr   sourcer   rE  c                C      t j| t jdS r   rL   r   r   r   rt   r   rF  r   rE  r7   r7   r8   meta_index_reduceL  s   
rJ  c                C      | S r2   r7   rI  r7   r7   r8   meta_index_reduce_Y  s   
rL  c                 C   s.   t |  }|  dkr| ||< | |S Nr   )r   r   rt   r   r   )r   rt   r   result_sizer7   r7   r8   meta_index_selectg  s   
rO  )lengthsr   offsetsaxisunsafeinitialdatarP  rQ  rR  rS  c          
         sf   |d urt d fdd}|d ur||jS |d ur/|jd d |jd d f }	||	S td)Nz?segment_reduce(): indices based reduction is not supported yet.c                    s(   t j| j d d   jdt jdS )Nr   rn   rS   rq   r   )rL   ry   r   rS   r   )lengths_shaperR  rU  r7   r8   segment_reduce_lengths_tensor  s   z:meta_segment_reduce.<locals>.segment_reduce_lengths_tensorr   r   z<segment_reduce(): Either lengths or offsets must be defined.)NotImplementedErrorr   r  )
rU  r   rP  r   rQ  rR  rS  rT  rY  rW  r7   rX  r8   meta_segment_reducep  s   
r[  c                 C   
   |  dS Nr7   r   r   r7   r7   r8   meta_max     
r_  c                 C   6   t | j|f}t| ||}| || j|tjdfS Nr}   rG   reduction_dimsr   _compute_reduction_shaper   rL   r   r   rt   keepdimrC  r7   r7   r8   meta_max_dim  
   rh  c                 C   r\  r]  r^  r   r7   r7   r8   meta_min  r`  rj  c                 C   ra  rb  rc  rf  r7   r7   r8   meta_min_dim  ri  rk  c                 C   s4   |   r
t| j}n	t| tjd\}}tj| |dS Nr?   r}   )r   r   rS   r   r   INT_TO_FLOATrL   r   )r   rD   rJ   r7   r7   r8   
meta_angle  s   
ro  c                 C   s$   t ||  | j |t | S r2   )rL   Z_resize_output_r   rq   copy_angle)r   r   r7   r7   r8   meta_angle_out  s   rr  c                 C      d S r2   r7   )valr7   r7   r8   assert_async     ru  c                 C   rs  r2   r7   )rt  
assert_msgr7   r7   r8   assert_async_meta  rv  rx  c                 C   rs  r2   r7   )r   r7   r7   r8   
print_meta  rv  ry  rS   rp   rq   rr   r   c                 C   s   t jdddS )Nr   rn   rq   r
  rz  r7   r7   r8   make_dep_token  s   	r|  c                 C   s4   ddl m} t| ttfrtd|| ||d d S )Nr   )constrain_range'Constraining SymFloat or Symbool is nyir   r   )r   r}  r`   r   r   r   )r   r   r   r}  r7   r7   r8   sym_constrain_range  s   r  c                 C      t j| ||d |S Nr  )r.   r  r   r   r   	dep_tokenr7   r7   r8   functional_sym_constrain_range     r  c                 C   s   ddl m} |d u r|d u rt|  d S t| ttfr tdt| t	u r>|d ur1t
| |k |d ur<t
| |k d S || ||d d S )Nr   )_constrain_range_for_sizer~  r  )r   r  rL   _check_is_sizer`   r   r   r   ri   r   r[   )r   r   r   r  r7   r7   r8   sym_constrain_range_for_size  s   
r  c                 C   r  r  )r.   r  r  r7   r7   r8   'functional_sym_constrain_range_for_size  r  r  c                 C   s   |S r2   r7   )rt  rw  r  r7   r7   r8   functional_assert_async_meta  rv  r  f_namec                 C   sX   |   dksJ | d| d| dks*J | d| d d| d dd S )Nr   z3: The input tensor must have at least 2 dimensions.r   z5: A must be batches of square matrices, but they are  by 	 matrices)rt   r   )r   r  r7   r7   r8   r     s    r   Anamec                    s   t j jk fdd t j jk fdd t  d dk fdd t  ddk fdd d S )Nc                         dj  d j  dS )Nz:Expected b and A to be on the same device, but found b on z
 and A on 	 instead.r{  r7   r  r   r7   r8   rX      
   z(linearSolveCheckInputs.<locals>.<lambda>c                      r  )Nz=Expected b and A to have the same dtype, but found b of type z and A of type r  r}   r7   r  r7   r8   rX   (  r  r   r  c                      s   d  d d  d dS )Nz3A must be batches of square matrices, but they are r  r  r   r  r   r7   r  r7   r8   rX   0  s
   c                      s:   d d  d d  d d d d d 
S )NzIncompatible matrix sizes for z: each A matrix is r   r  z but each b matrix is r  r   r7   r  r  r   r7   r8   rX   8  s   )rL   r[   rq   rS   r   )r   r  r  r7   r  r8   linearSolveCheckInputs  s    


r  tallow_low_precision_dtypesc                    s^   | j  t|  p|   fdd |s-t tjtjtjtjfv  fdd d S d S )Nc                          d  S )Nz<: Expected a floating point or complex tensor as input. Got r7   r7   rS   r  r7   r8   rX   I      z(checkFloatingOrComplex.<locals>.<lambda>c                      r  )Nz*: Low precision dtypes not supported. Got r7   r7   r  r7   r8   rX   N  r  )	rS   rL   r[   r   r   rO   rQ   rN   rP   )r  r  r  r7   r  r8   r   A  s   r   arg_namec                    s"   t |  dk fdd d S )Nr   c                          d  dS )Nz: The input tensor z! must have at least 2 dimensions.r7   r7   r  r  r7   r8   rX   V  rY   zcheckIsMatrix.<locals>.<lambda>)rL   r[   rt   )r  r  r  r7   r  r8   checkIsMatrixS  s   
r  Br   c                    sZ   t   t tr ddkn	 ddk fdd d S )Nr  r   c                      sH    drdnd d  d d  d d d d d d	S )
Nz2: Incompatible shapes of A and B for the equation zAX = BzXA = Bz (r  rB   r   r   rh   r   r7   r  r  r  r   r7   r8   rX   _  s   
z#checkInputsSolver.<locals>.<lambda>)r   r  rL   r[   r   )r  r  r   r  r7   r  r8   checkInputsSolverZ  s   

*r  resultfn_nameresult_namec                    s&   t jjk fdd d S )Nc                	      s$     d d dj  dj  	S )Nz: Expected z5 and input tensors to be on the same device, but got z on z and input on r{  r7   r  r   r  r  r7   r8   rX   o  s   z!checkSameDevice.<locals>.<lambda>)rL   r[   rq   )r  r  r   r  r7   r  r8   checkSameDeviceg  s   
r  UPLOc                    s8      }tt dko|dkp|dk fdd d S )Nr   ULc                      
   d  S )Nz1Expected UPLO argument to be 'L' or 'U', but got r7   r7   r  r7   r8   rX   z     
 zcheckUplo.<locals>.<lambda>)upperrL   r[   r   )r  ZUPLO_uppercaser7   r  r8   	checkUplov  s
   
r  eigenvaluesZeigenvectorsr  	compute_vc                 C   sp   t | d t| t| j}|r | |}||t|dd n| dg}|  | j|t| j	d}||fS )Nzlinalg.eighFZ	row_majorr   r}   )
r   r  r   r   r   r   r   poprT   rS   )r  r  r  r   Zvecsvalsr7   r7   r8   meta__linalg_eigh~  s   


r  c                 C   s@   t | d t| jr| jnt| j}| j| jd d |dS )Nzlinalg.eigvalsr   r}   r   rG   rw   rS   r   r   r   )r   complex_dtyper7   r7   r8   meta__linalg_eigvals  s   


r  c                 C   sX   t | d t| jr| jnt| j}| j| jd d |d}| j| j|d}||fS )Nz
linalg.eigr   r}   r  )r   r  r   Zvectorsr7   r7   r8   meta_linalg_eig  s   


r  r  c                 C   s   | j jtjdddS )Nr   r  r   )ZmTr   rL   r   	transpose)r  r7   r7   r8   cloneBatchedColumnMajor     r  r  c                 C   s   t | S r2   )r  )r   r  r  r7   r7   r8   _cholesky_solve_helper  s   r  c                    sP   t jdkfdd t  jdk fdd t d\}}t|||S )Nr   c                         d j  dS )Nz-b should have at least 2 dimensions, but has  dimensions insteadr   r7   r   r7   r8   rX     r  z cholesky_solve.<locals>.<lambda>c                      r  )Nz-u should have at least 2 dimensions, but has r  r  r7   r  r7   r8   rX     r  cholesky_solve)rL   r[   r   !_linalg_broadcast_batch_dims_namer  )r   r  r  Zself_broadcastedZA_broadcastedr7   r  r8   r    s   

r  c                 C   s.   |   dkrtj| tjdS t| d t| S )Nr   r   cholesky)r   rL   r   legacy_contiguous_formatr   r  r   r  r7   r7   r8   r    s   
r  c                 C   s   t | d t| S )Ncholesky_inverse)r   r  r  r7   r7   r8   r    s   
r  check_errorsc                 C   sf   t | d t| d | j}t|}t|d}| |}||| | j|d|d  tjd}||fS )Nzlinalg.choleskyFr   r   r}   )	r   r   r   r   r   r   r   rL   r)  )r  r  r  ZA_shaper   Z	L_stridesr  infosr7   r7   r8   linalg_cholesky_ex  s   



r  tauc                    s  t jdkdd  t ddkdd  t ddkdd  t jj dkfd	d jdkr[jd d }jd d  t  |k fd
d t jjkfdd tdd t jjtjddjj	dS )Nr   c                   S   r]   )NzHtorch.linalg.householder_product: input must have at least 2 dimensions.r7   r7   r7   r7   r8   rX     r_   z,linalg_householder_product.<locals>.<lambda>r  r   c                   S   r]   )Nzbtorch.linalg.householder_product: input.shape[-2] must be greater than or equal to input.shape[-1]r7   r7   r7   r7   r8   rX     r_   c                   S   r]   )Nz`torch.linalg.householder_product: input.shape[-1] must be greater than or equal to tau.shape[-1]r7   r7   r7   r7   r8   rX     r_   r   c                         dj  d j  S )Nzptorch.linalg.householder_product: Expected tau to have one dimension less than input, but got tau.ndim equal to  and input.ndim is equal to r  r7   r   r  r7   r8   rX     
   c                      r  )Nzltorch.linalg.householder_product: Expected batch dimensions of tau to be equal to input.shape[:-2], but got r7   r7   actual_batch_tau_shaper7   r8   rX        c                      r  )Nz,torch.linalg.householder_product: tau dtype z does not match input dtype r}   r7   r  r7   r8   rX     s   
z torch.linalg.householder_productr  Fr  r   r   rS   rq   )
rL   r[   r   r   r   rS   r  empty_stridedr   rq   )r   r  Zexpected_batch_tau_shaper7   )r  r   r  r8   linalg_householder_product  sD   


r  c                 C   s^   t | d t| ddd | | j}|| jt| jdd | j| jd d tjd}||fS )Nzlinalg.inv_exF)r  r  r  r}   r   r   r   r   r   r   rL   r)  )r  r  r  r  r7   r7   r8   linalg_inv_ex_meta  s   
r  LDpivotsinfo)	hermitianr  r  c                C   st   t | d t| d tj| jt| jdd| j| jd}| j| jd d tj	d}| j| jd d tj	d}|||fS )Nztorch.linalg.ldl_factor_exFr  r  r   r}   r  )
r   r   rL   r  r   r   rS   rq   r   r   )r   r  r  r  r  r  r7   r7   r8   linalg_ldl_factor_ex_meta+  s   


r  )r  c                   s   t d td t d t jdk fdd jd d }t|jkfdd ttj	fdd tj	 j	k fdd t
 \}}tj|t|d	d
 j	 jdS )Nztorch.linalg.ldl_solver   c                      r  )NzMtorch.linalg.ldl_solve: Expected B to have at least 2 dimensions, but it has r  r  r7   )r  r7   r8   rX   N     z'linalg_ldl_solve_meta.<locals>.<lambda>r   c                      r  )Nzjtorch.linalg.ldl_solve: Expected LD.shape[:-1] and pivots.shape to be the same, but got pivots with shape  insteadr   r7   r  r7   r8   rX   V  r  c                      r|   )Nz<torch.linalg.ldl_solve: Expected pivots to be integers. Got r}   r7   r  r7   r8   rX   ]  r   c                      r  )Nz!torch.linalg.ldl_solve: LD dtype z does not match b dtype r}   r7   )r  r  r7   r8   rX   a  r   Fr  r  )r   r   r  rL   r[   r   r   rG   is_integer_dtyperS   _linalg_broadcast_batch_dimsr  r   rq   )r  r  r  r  Zexpected_pivots_shapeB_broadcast_sizerJ   r7   )r  r  r  r8   linalg_ldl_solve_meta@  s6   
	






r  Pr  )pivotr  c          	         s   t  jdk fdd t j}|d }|d }t||}||d< |r+ |}n dg}||d<  |}||d< ||d<  |}|||fS )Nr   c                      r  )Nz@linalg.lu: Expected tensor with 2 or more dimensions. Got size: r  r   r7   r  r7   r8   rX   q  r  z linalg_lu_meta.<locals>.<lambda>r  r   r   )rL   r[   r   r   r   r   r   )	r  r  sizesrB  r  rA  r  r  r  r7   r  r8   linalg_lu_metal  s$   





r  LU)r  r  c          	         s   t  jdk fdd t j}|d }|d }t j|t|dd j jd}|	  t
|||d<  j|t jd	}|	   j|t jd	}|||fS )
Nr   c                      r  )NzFtorch.lu_factor: Expected tensor with 2 or more dimensions. Got size: r  r   r7   r  r7   r8   rX     r  z*linalg_lu_factor_ex_meta.<locals>.<lambda>r  r   Fr  r  r}   )rL   r[   r   r   r   r  r   rS   rq   r  r   r   r   )	r  r  r  r  rB  r  r  r  r  r7   r  r8   linalg_lu_factor_ex_meta  s&   



r  )r   adjointr  c                   s   t d tj jk fdd tjtjkdd  td t |d tddkdd  tjd d jkfdd t	 \}}tj
|t|| d	 j jd
}| dkru|su| ru| }|S )Nztorch.linalg.lu_solvec                      r  )NzPlinalg.lu_solve: Expected LU and B to have the same dtype, but found LU of type  and B of type r  r}   r7   )r  r  r7   r8   rX     r  z&linalg_lu_solve_meta.<locals>.<lambda>c                   S   r]   )NzElinalg.lu_solve: pivots should be a Tensor of scalar type torch.int32r7   r7   r7   r7   r8   rX     r_   zlinalg.lu_solver   c                   S   r]   )NzYlinalg.lu_solve: Number of pivots per batch should be same as the dimension of the matrixr7   r7   r7   r7   r8   rX     r_   c                      r  )Nzclinalg.lu_solve: Expected LU.shape[:-1] and pivots.shape to be the same, but got pivots with shape r  r   r7   r  r7   r8   rX     r  r  r  r   )r   rL   r[   rS   r   r   r  r   r   r  r  r   rq   r   r   Zconj)r  r  r  r   r  r  rJ   r  r7   )r  r  r  r8   linalg_lu_solve_meta  s<   




r  unpack_dataunpack_pivotsc                    s   t  jdk fdd |rt |jt jkdd  t j}|d }|d }t||}||d< |r9 |}n dg}|rX||d<  |}	||d< ||d<  |}
n dg}	 dg}
||	|
fS )Nr   c                      r  )NzFtorch.lu_unpack: Expected tensor with 2 or more dimensions. Got size: r  r   r7   r  r7   r8   rX     r  z lu_unpack_meta.<locals>.<lambda>c                   S   r]   )Nztorch.lu_unpack: LU_pivots is expected to be a contiguous tensor of torch.int32 dtype.
Note: this function is intended to be used with the output produced by torch.linalg.lu_factorr7   r7   r7   r7   r8   rX        r  r   r   )	rL   r[   r   rS   r)  r   r   r   r   )r  r  r  r  r  rB  r  rA  r  r  r  r7   r  r8   lu_unpack_meta  s4   





r  modec                    sd    dkrd}d}||fS  dkrd}d}||fS  dkr$d}d}||fS t d fdd ||fS )NreducedTZcompleteFrc                         d  dS )Nzqr received unrecognized mode 'z=' but expected one of 'reduced' (default), 'r', or 'complete'r7   r7   r  r7   r8   rX     s   z _parse_qr_mode.<locals>.<lambda>rL   r[   )r  	compute_qr  r7   r  r8   _parse_qr_mode  s"   	
r  QRr  c                 C   s   t | d t| d t|\}}| jd }| jd }t||}|r>t| j}|r*|n||d< | |}||t|dd n| dg}t| j}	|sM|sO|n||	d< | |	}
|
|	t|	dd ||
fS )Nz	linalg.qrr  r   Fr  r   )	r  r   r  r   r   r   r   r   r   )r  r  r  Zreduced_moderB  r  rA  ZQ_shaper  ZR_shaper  r7   r7   r8   linalg_qr_meta$  s"   








r  sign	logabsdetc                 C   s   t | d t| dd | j}| |d d }| j|d d t| jd}tj|t|d| j| j	d}| j|d d tj
d}||||fS )Nzlinalg.slogdetFr  r}   r  r   )r   r   r   r   rT   rS   rL   r  r   rq   r)  )r  r   r  r  r  r  r7   r7   r8   _linalg_slogdet@  s   
r   full_matrices
compute_uvdriverc                 C   s   t | d t| d t| jd d }| jd }| jd }t||}|r]|||r*|n|g }| |}	|	|t|dd ||rB|n||g }
| |
}t| dk}||
t|
|d n| dg}	| dg}| j||g t	| j
d}|	||fS )	Nz
linalg.svdr  r   Fr  r   r   r}   )r  r   r   r   r   r   r   r   r   rT   rS   )r  r  r  r  r   rB  r  rA  ZU_shaper  ZV_shapeVZis_cudaSr7   r7   r8   _linalg_svd_metaT  s$   







r  arg1arg2c                 C   sn   | j d d }|j d d }t||}t|}|| d| dg7 }t|}||d|dg7 }||fS )Nr  r   )r   r(   r   r   )r  r  Zarg1_batch_sizesZarg2_batch_sizesexpand_batch_portionarg1_expand_sizearg2_expand_sizer7   r7   r8   r  z  s   
r  c                 C   sV   |rt | || t| |\}}|| jkr| n| |}||jkr"|n||}||fS r2   )r  r  r   expand)r  r  r  r
  r  Zarg1_broadcastedZarg2_broadcastedr7   r7   r8   r    s   r  r   c                 C   s6   | j d d }|jdkp| jd |jko|j |k}|S )Nr   r   )r   r   )r   r   Zexpected_batched_rhs_shapevector_caser7   r7   r8   linalg_solve_is_vector_rhs  s
   
r  )r   r  r  r  r  r  c                   sh  t  d t jjk fdd t }|r dn}	t |	|d t|	 \}
}t|p6| dd  |rC|
d d n|
}tj|t	|| jj
d} j}tj|t	|d j j
d} j|d d tjd} j|d d	 tjd}||||f}||||f}td
d |D rt||D ]\}}t||j ||j|  t||dd q|S )Nzlinalg.solvec                         d j  dj  dS )NzKlinalg.solve: Expected A and B to have the same dtype, but found A of type r  r  r}   r7   r  r  r7   r8   rX     r  z"_linalg_solve_ex.<locals>.<lambda>r   c                   S   r]   )Nzlinalg.solve: Vector broadcasting of the left hand side is not supported for left=False. In this case linalg.solve is equivalent to B / A.squeeze(-1)r7   r7   r7   r7   r8   rX     r  r  Fr}   r  c                 s   s    | ]}|d uV  qd S r2   r7   r@   r7   r7   r8   rc         z#_linalg_solve_ex.<locals>.<genexpr>)	copy_fromcopy_toexact_dtype)r   rL   r[   rS   r  r   r  r  r  r   rq   r   r   r)  allzipr$   r   r   r&   )r  r  r   r  r  r  r  r  r  B_ZB_broad_shaperJ   Zresult_shapeZresult_r   ZLU_Zpivots_Zinfo_r   resr  or7   r  r8   _linalg_solve_ex  sJ   



r  )r   unitriangularr   r  r   c          	      C   s   |d u r
|  dg}t|tsJ t| ||d t|| d \}}|dd o+| }|r6t||j	}|S t
||j	rL||ddj	 |dd |S )Nr   zlinalg.solve_triangularr  r   )r   r`   r"   r  r  r  r   Zis_conjr$   r   r%   r   
transpose_)	r  r  r  r   r  r   r  ZA_Zavoid_copy_Ar7   r7   r8   linalg_solve_triangular_meta  s   
r  XM)r  r  c           	         s   t jdkfdd t  jdk fdd t d  jt jkrOt \}}t j|t|ddj	j
d}t j|t|dd j	 j
d}||fS  jt jks[ jt jkrjt }d	g}||fS t dd
d  ||fS )Nr   c                      r  )NzMtorch.triangular_solve: Expected b to have at least 2 dimensions, but it has r  r  r7   r   r7   r8   rX     r  z'triangular_solve_meta.<locals>.<lambda>c                      r  )NzMtorch.triangular_solve: Expected A to have at least 2 dimensions, but it has r  r  r7   r  r7   r8   rX     r  triangular_solveFr  r  r   c                   S   r]   )Nz+triangular_solve: Got an unexpected layout.r7   r7   r7   r7   r8   rX   (  r_   )rL   r[   r   r  rp   stridedr  r  r   rS   rq   
sparse_csr
sparse_bsrr   r   )	r   r  r  r  r  Zself_broadcast_sizeZA_broadcast_sizeZsolutionZcloned_coefficientr7   r  r8   triangular_solve_meta  s<   	




r$  c                 C   sp   t | d t| d | | jd d }| | j}|| jt| jdd | j| jd d tjd}|||fS )Nz
linalg.detr  Fr  r   r}   r  )r  Zdetr  r  r7   r7   r8   _linalg_det_meta-  s   


r%  c                    s  t jdkdd  t jdkdd  |rdndt j jd kfdd t j jd kfdd t jd jd kd	d  t jj d
kfdd t jjkfdd jdkrjd d }jd d t |kfdd jd d  t  |k fdd t jjkfdd t jjkfdd tdd tdd t jjtjddjjdS )Nr   c                   S   r]   )Nz3torch.ormqr: input must have at least 2 dimensions.r7   r7   r7   r7   r8   rX   E  r_   zormqr.<locals>.<lambda>c                   S   r]   )Nz3torch.ormqr: other must have at least 2 dimensions.r7   r7   r7   r7   r8   rX   H  r_   r  r   c                      r  )Ntorch.ormqr: other.shape[z0] must be greater than or equal to tau.shape[-1]r7   r7   left_size_conditionr7   r8   rX   N  r   c                      r  )Nr&  z"] must be equal to input.shape[-2]r7   r7   r'  r7   r8   rX   R  r   c                   S   r]   )NzHtorch.ormqr: tau.shape[-1] must be less than or equal to input.shape[-1]r7   r7   r7   r7   r8   rX   W  r_   r   c                      r  )Nz[torch.ormqr: Expected tau to have one dimension less than input, but got tau.ndim equal to r  r  r7   r  r7   r8   rX   \  r  c                      r  )Nzhtorch.ormqr: Expected other to have the same number of dimensions as input, but got other.ndim equal to r  r  r7   r   r   r7   r8   rX   c  r  c                      r  )NzWtorch.ormqr: Expected batch dimensions of tau to be equal to input.shape[:-2], but got r7   r7   r  r7   r8   rX   n  r  c                      r  )NzYtorch.ormqr: Expected batch dimensions of other to be equal to input.shape[:-2], but got r7   r7   )actual_batch_other_shaper7   r8   rX   w  r  c                         d j  dj  S )NzPtorch.ormqr: Expected input and tau to have the same dtype, but input has dtype z and tau has dtype r}   r7   r  r7   r8   rX     r  c                      r+  )NzRtorch.ormqr: Expected input and other to have the same dtype, but input has dtype z and other has dtype r}   r7   r)  r7   r8   rX     r  ztorch.ormqrr  r   Fr  r  )	rL   r[   r   r   rS   r  r  r   rq   )r   r  r   r   r  Zexpected_batch_shaper7   )r*  r  r   r(  r   r  r8   ormqr;  sn   	







r,  c                   s   t td  k fdd j}| d k}|}| }|r3td|D ]}|o0|dk}q&ntd|D ]}|oB|dk}q8t |pI| fdd d S )Nr   c                      s   dd   dt  S )Nzpadding size is expected to be r   z, but got: r   r7   )rt   paddingr7   r8   rX         z,_padding_check_valid_input.<locals>.<lambda>r   r   c                      s    d d  d d  dj  S )N	Expected r   zD or r   zcD (batch mode) tensor with possibly 0 batch size and other non-zero dimensions for input, but got: r   r7   )rt   r   r7   r8   rX     s   )rL   r[   r   r   r   r   )r   r.  rt   Z	input_dimZis_batch_modeZvalid_batch_modeZvalid_non_batch_moder   r7   )rt   r   r.  r8   _padding_check_valid_input  s$   r1  c                   s   d}d d}j dkrd} d7  |d7 }t|dd |\|}   |rHtk o>k  fdd tdkfdd j dkra|fS ||fS )	Nr   r   r0   r   c                         d d d  dj  S NzcArgument #4: Padding size should be less than the corresponding input dimension, but got: padding (rg   ) at dimension 
 of input r   r7   dim_wr   pad_lpad_rr7   r8   rX        z_pad1d_common.<locals>.<lambda>c                      re   )Nz
input (W: z%) is too small. Calculated output W: r7   r7   )input_woutput_wr7   r8   rX     rY   r   )r   r   r1  rL   r[   r   )r   r.  is_reflection	dim_planenbatchnplaner7   )r7  r   r;  r<  r8  r9  r8   _pad1d_common  s0   




rA  c                 C      t | |ddS NTr=  )rA  r   r.  r7   r7   r8   meta_reflection_pad1d     rF  c                    *   t  jt jk fdd t |ddS )Nc                         d j   dS )Nz)"replication_pad1d" not implemented for ''rS   __str__r7   r   r7   r8   rX         z(meta_replication_pad1d.<locals>.<lambda>FrD  )rL   r[   rS   boolrA  rE  r7   rM  r8   meta_replication_pad1d  
   

rP  c                   s   d |st t|dkdd  jdkr d7  |\ }|  |r=t |k o3|k  fdd t  k fdd jS )Nr   r   c                   S   r]   )Nz padding size is expected to be 2r7   r7   r7   r7   r8   rX     r_   z(_pad1d_backward_common.<locals>.<lambda>r0   c                      r2  r3  r   r7   r6  r7   r8   rX     r:  c                         d d   S Nz(grad_output width unexpected. Expected: , Got: r   r7   r7  grad_outputr<  r7   r8   rX     rF   rL   r[   r   r   r   r   r   )rV  r   r.  r=  r;  r7   )r7  rV  r   r<  r8  r9  r8   _pad1d_backward_common  s$   

rX  
grad_inputc                 C      t | ||ddS rC  rX  rV  r   r.  r7   r7   r8   meta_reflection_pad1d_backward
     r]  c                 C   rZ  )NFrD  r[  r\  r7   r7   r8   meta_replication_pad1d_backward  r^  r_  c                   s2  dd d}d}t |dd j}|dkr'd}d7  d7  |d7 }|\	
|} 
   	 |rptk oS	k 	fdd t
k ofk  
fdd tdkpydkfd	d jd
kr|fS ||fS )Nr   r   r   r      c                      r2  r3  r   r7   r6  r7   r8   rX   0  r:  z_pad2d_common.<locals>.<lambda>c                         d d d  dj  S NzcArgument #6: Padding size should be less than the corresponding input dimension, but got: padding (rg   r4  r5  r   r7   dim_hr   pad_bpad_tr7   r8   rX   7  r:  c                      s   d  d d d S )Nz
input (H:  W: z%) is too small. Calculated output H: r7   r7   )input_hr;  output_hr<  r7   r8   rX   ?  s
   r0   r1  r   r   rL   r[   r   )r   r.  r=  Z
dim_slicesr?  r   r@  r7   )rd  r7  r   rh  r;  ri  r<  re  r8  r9  rf  r8   _pad2d_common  sB   




rk  c                 C   rB  rC  )rk  rE  r7   r7   r8   meta_reflection_pad2dK  rG  rl  c                    rH  )Nc                      rI  )Nz)"replication_pad2d" not implemented for 'rJ  rK  r7   rM  r7   r8   rX   V  rN  z(meta_replication_pad2d.<locals>.<lambda>FrD  )rL   r[   rS   rO  rk  rE  r7   rM  r8   meta_replication_pad2dQ  rQ  rm  c                    s   dd d}|j }| dkrd7  d7  |d7 }|\}}}}|  }	| }
|	| | |
| | tkfdd t k fdd ||j S )Nr   r   r   r`  c                      rR  rS  r   r7   rU  r7   r8   rX   x  rF   z%meta_pad2d_backward.<locals>.<lambda>c                      rR  Nz)grad_output height unexpected. Expected: rT  r   r7   rd  rV  ri  r7   r8   rX   |  rF   )r   rt   rL   r[   r   r   )rV  r   r.  r>  rW   r8  r9  rf  re  rh  r;  r7   )rd  r7  rV  ri  r<  r8   meta_pad2d_backward[  s,   
rp  c             	      s  ddd d}t |dd jdk}|r+d}d7 d7  d7  |d7 }|\
|}    
   	|rtk odk fdd tk ow
k 
fd	d tk ok  fd
d t	dkpdkpdk	fdd |r||	fS |	fS )Nr0   r   r   r   r      c                      r2  r3  r   r7   r6  r7   r8   rX     r:  z_pad3d_common.<locals>.<lambda>c                      ra  rb  r   r7   rc  r7   r8   rX     r:  c                      ra  )NzcArgument #8: Padding size should be less than the corresponding input dimension, but got: padding (rg   r4  r5  r   r7   )dim_dr   pad_bkpad_fr7   r8   rX     r:  c                      s(   d  d d d d d S )Nz
input (D:  H: rg  z%) is too small. Calculated output D: r7   r7   )input_drh  r;  output_dri  r<  r7   r8   rX     s   rj  )r   r.  r=  r>  Z
batch_moder?  r@  r7   )rr  rd  r7  r   rv  rh  r;  rw  ri  r<  re  rs  rt  r8  r9  rf  r8   _pad3d_common  sP   





rx  c                 C   rB  rC  )rx  rE  r7   r7   r8   meta_reflection_pad3d  rG  ry  c                    rH  )Nc                      rI  )Nz)"replication_pad3d" not implemented for 'rJ  rK  r7   rM  r7   r8   rX     rN  z(meta_replication_pad3d.<locals>.<lambda>FrD  )rL   r[   rS   rO  rx  rE  r7   rM  r8   meta_replication_pad3d  rQ  rz  c                    s(  t t|dkdd  |jdksJ j|jksJ ddd |jdkr2d7 d7  d7  |\}}}}}}| }	|}
|}|	| | |
| | || | t kfdd t kfd	d t  k fd
d ||jS )N   c                   S   r]   )Nz padding size is expected to be 6r7   r7   r7   r7   r8   rX     r_   z%meta_pad3d_backward.<locals>.<lambda>r0   r   r   rq  c                      rR  rS  r   r7   rU  r7   r8   rX     rF   c                      rR  rn  r   r7   ro  r7   r8   rX     rF   c                      rR  )Nz(grad_output depth unexpected. Expected: rT  r   r7   )rr  rV  rw  r7   r8   rX     rF   rW  )rV  r   r.  r8  r9  rf  re  rt  rs  rv  rh  r;  r7   )rr  rd  r7  rV  rw  ri  r<  r8   meta_pad3d_backward  s<   




r|  r   pc                 C   s^   t |  dd  | d}|dkr| dgjt jdS | ||d  d fjt jdS )Nc                   S   r]   )Nz(_pdist_forward requires contiguous inputr7   r7   r7   r7   r8   rX   	  r_   z%meta__pdist_forward.<locals>.<lambda>r   r   r   r   )rL   r[   r   r   r   r  r  )r   r}  r  r7   r7   r8   meta__pdist_forward 	  s   
r~  gradpdistc                 C   s8   t | dd  t | dd  t j|t jdS )Nc                   S   r]   )Nz._pdist_backward requires self to be contiguousr7   r7   r7   r7   r8   rX   	  r_   z&meta__pdist_backward.<locals>.<lambda>c                   S   r]   )Nz/_pdist_backward requires pdist to be contiguousr7   r7   r7   r7   r8   rX   	  r_   r   )rL   r[   r   r   r  )r  r   r}  r  r7   r7   r8   meta__pdist_backward	  s   r  )r2  r1  c                   s  ddl m}m}  d} d}d}	|t|j|||	fr-|||	ft 	 dkdd  t	 dkdd  t
jsatj j  koVjkn   fd	d  j}
j|
d |
d td ko|d kfd
d  S )Nr   )r   r   r   r   r0   c                   S   r]   Nzbatch1 must be a 3D tensorr7   r7   r7   r7   r8   rX   %	  r_   zmeta_baddbmm.<locals>.<lambda>c                   S   r]   Nzbatch2 must be a 3D tensorr7   r7   r7   r7   r8   rX   &	  r_   c                      s   dj  d j  dj  S )Nz+Input dtypes must be the same, got: input: z
, batch1: z
, batch2: r}   r7   )batch1batch2r   r7   r8   rX   *	      c                	      &   d d d d  d d  d	S Nz@Expected size for first two dimensions of batch2 tensor to be: [rg   z] but got: [r   r   ].r7   r7   batch2_sizesbscontraction_sizer7   r8   rX   2	  s   )r   r   r   r   rL   Zsym_notr   r  r[   rt   
exp_configZ&skip_dtype_check_in_meta_registrationsrS   r   )r   r  r  r2  r1  r   r   dim1dim2Zdim3batch1_sizesr7   )r  r  r  r  r  r   r8   meta_baddbmm	  s,   


r  c                C   rG  r   rH  r   r  r7   r7   r8   meta_bernoulli:	  s   r        ?c                 C   rK  r2   r7   r   r}  r  r7   r7   r8   meta_bernoulli_A	  rv  r  c                 C   rG  r   rH  r  r7   r7   r8   meta_bernoulli_pF	  r^  r  c                 C   
   t | S r2   rL   r   r  r7   r7   r8   meta_poissonL	  r`  r  c                 C   s6   t |
|  k dd  t j| t jd}t | |fS )Nc                   S   r]   )NzJError in fused_moving_avg_obs_fake_quant_cpu: ch_axis must be < self.dim()r7   r7   r7   r7   r8   rX   d	  r_   z6meta__fused_moving_avg_obs_fq_helper.<locals>.<lambda>r}   )rL   r[   rt   r   rO  )r   Zobserver_onZfake_quant_onZrunning_minZrunning_maxscaleZ
zero_pointZaveraging_constZ	quant_minZ	quant_maxZch_axisZper_row_fake_quantZsymmetric_quantmaskr7   r7   r8   $meta__fused_moving_avg_obs_fq_helperR	  s   
r  c                    sn   t |  dkdd  t | dkdd  | j\ |j\t  k fdd | S )Nr   c                   S   r]   )Nza must be 2Dr7   r7   r7   r7   r8   rX   m	  r_   zmeta_mm.<locals>.<lambda>c                   S   r]   )Nzb must be 2Dr7   r7   r7   r7   r8   rX   n	  r_   c                	      s   d d  d d d	S )Nz/a and b must have same reduction dim, but got [rg   z] X [r  r7   r7   ZM1ZM2Nr  r7   r8   rX   s	  s    )rL   r[   rt   r   r   r   br7   r  r8   meta_mmj	  s   

r  c                    s0   |rt  fddtjD S tj S )Nc                 3   s&    | ]}| vrj | nd V  qdS )r   Nr   rA   r   dimsr   r7   r8   rc   z	  s   $ z+_compute_reduction_shape.<locals>.<genexpr>)rZ   r   r   rG   compute_reduction_output_shaper   )r   r  rg  r7   r  r8   re  x	  s   re  strc                 C   sD   t | tjjr| jjS t| dr t| jdr | jjdkr | jjS dS )Nrq   ri   rn   r   )r`   rL   Z_subclassesZ
FakeTensorZfake_deviceri   hasattrrq   )r  r7   r7   r8   r   	  s   

r   input_tensorr   r.  dilationis_transposedgroupsoutput_paddingc                    s@  dt dt dt dt dt dt fdd}dt dt dt dt dt d	t dt fd
d}	|jdd  }
| jdd   |r<||jd  }n|jd }|jd | | jd krQtd| jd |gt|tre|gt  }nt|dkrt|d gt  }t|tr|gt  }nt|dkr|d gt  }t|tr|gt  }nt|dkr|d gt  }d }|rt|tr|gt  }nt|dkr|d gt  }n|}tt D ]2}|r|	 | || || |
| || ||  qՈ| | || || |
| ||  qt	t
dd dd  D  fdd S )Nlnr}  r   rA  r   r1   c                 S   s$   | d|  ||d   d | d S )a  
        Formula to apply to calculate the length of some dimension of the output

        See: https://pytorch.org/docs/stable/generated/torch.nn.Conv2d.html

        Args:
            ln: length of the dimension
            p: padding in that dim
            d: dilation in that dim
            k: kernel size in that dim
            s: stride in that dim
        Returns:
            The output length
        r   r   r7   )r  r}  r   rA  r   r7   r7   r8   _formula	  s   $z+calc_conv_nd_return_shape.<locals>._formular4   c                 S   s(   | d | d|  ||d   | d S )a  
        Formula to apply to calculate the length of some dimension of the output
        if transposed convolution is used.
        See: https://pytorch.org/docs/stable/generated/torch.nn.ConvTranspose2d.html

        Args:
            ln: length of the dimension
            p: padding in that dim
            d: dilation in that dim
            k: kernel size in that dim
            s: stride in that dim
            op: output padding in that dim

        Returns:
            The output length
        r   r   r7   )r  r}  r   rA  r   r4   r7   r7   r8   _formula_transposed	  s   (z6calc_conv_nd_return_shape.<locals>._formula_transposedr   r   r   zInvalid channel dimensionsc                 s       | ]}|d kV  qdS r   Nr7   r@   r7   r7   r8   rc   	  r  z,calc_conv_nd_return_shape.<locals>.<genexpr>c                      s   dt   ddd   dS )NzGiven input size per channel: z&. Calculated output size per channel: r   z. Output size is too small)r   r7   r  Z	ret_shaper7   r8   rX   	  s    
z+calc_conv_nd_return_shape.<locals>.<lambda>)r   r   r  r`   r   r   r   r   rL   r[   ru   )r  r"  r   r.  r  r  r  r  r  r  kernel_sizeZout_channelsZoutput_padding_listr   r7   r  r8   calc_conv_nd_return_shape	  sb   "
&




"r  c                 C      t j| t jkS r2   rL   _prims_commonr!   channels_lasttenr7   r7   r8   is_channels_last	     r  running_meanrunning_vartrainingexponential_average_factorepsilonc                    s    j }|d ur
|j n|j }	|d ur|j n|j }
 fdd} |j| d}|r4 |	} |
}n
 d} d}|||fS )Nc                      s(   t  rtjS  jtjdrtjS tjS r   )r  rL   r  r   r   r7   r  r7   r8   pick_memory_format
  s
   z2meta_miopen_batch_norm.<locals>.pick_memory_formatr   r   )r   r   r  )r  r"  r$  r  r  r  r  r  r   Zsave_mean_shapeZsave_var_shaper  r   Z	save_meanZsave_varr7   r  r8   meta_miopen_batch_norm 
  s   



r  c	              	      sf    fdd}	t  ||||||r|nd }
d}d} |dkr%d|
|<  |
}|j|	 d}|S )Nc                      s^   t  dkrt strtjS nt rtjS  jtjdr#tjS  jtjdr-tjS d S Nr   r   )r   r  rL   r  r   r   preserve_formatr7   r  r"  r7   r8   r  2
  s   z%meta_conv.<locals>.pick_memory_formatr   r   r   )r  r   r   r  )r  r"  r$  r   r.  r  r  r  r  r  	shape_outZinput_channels_dimZoutput_channels_dimr   r7   r  r8   	meta_conv&
  s$   

r  mkldnnc
              	   C   sH   t | ||||d|g }
| |
}tj}|  dkrtj}|j|d}|S )NFrq  r   )r  r   rL   r  rt   channels_last_3dr  )r  r"  r$  r.  r   r  r  attrscalars	algorithmr  r   Zout_memory_formatr7   r7   r8   meta_mkldnn_convolution_defaultX
  s   
r  c                 C   s$   |  g | jd d |jd R S Nr   r   r   r   )r  r"  r$  r  r  r  r7   r7   r8   meta_linear_pointwise_defaulto
  s   $r  mklc                 C   s$   |  g | jd d |jd R S r  r  )r  Zpacked_weightZorig_weightr$  r   r7   r7   r8   meta_mkl_linearz
  s   r  onednnc              	   C   s|   t | ||||	d|
d }|tjtjtjtjfv sJ | j||d}t|dv s*J dt|dkr3tjntj	}|j
|d}|S )NFr}   r0   r`  zonly conv1d/2d are supportedr`  r   )r  rL   r=  r?  uint8r(  r   r   r  r   r  )rB   x_scalex_zpww_scalew_zpr$  r   r.  r  r  output_scaleoutput_zero_pointoutput_dtyper  r  r  r  r   formatr7   r7   r8   meta_qconv_pointwise
  s    
r  c                 C   s   |dksJ |S )Nsumr7   )rB   r  r  r  r  r  accumr$  r   r.  r  r  r  r  r  Zaccum_scaleZaccum_zero_pointbinary_op_namer1  unary_op_nameunary_op_argsunary_op_algorithmr7   r7   r8   meta_qconv2d_pointwise_binary
  s   r  c                 C   sF   t | j}|jd |d< |	tjtjtjtjfv sJ | j||	d}|S )Nr   r   r}   )r   r   rL   r=  r?  r(  r  r   )rB   r  r  r  r  r  r$  r  r  r  Zpost_op_nameZpost_op_argsZpost_op_algorithmrC  r   r7   r7   r8   meta_qlinear_pointwise
  s
   
r  c                 C   sR   |dkr|S t | j}|jd |d< |
tjtjtjtjfv s J | j||
d}|S )Nr  r   r   r}   )r   r   rL   r=  r?  r  r(  r   )rB   r  r  r  r  r  Zx_2r$  r  r  r  Zx2_scaleZx2_zpr  r1  r  r  r  rC  r   r7   r7   r8   meta_qlinear_pointwise_binary
  s   
r  c                 C   s&   t | j}|jd |d< | |}|S )Nr   r   )r   r   r   )rB   r  r$  rC  r   r7   r7   r8   meta_linear_dynamic_fp16
  s   

r  	quantizedr7   r   r   c                 C   sr   t | |||||\}}}|  dkr| dnd}	tj}
|  dkr(|||g}n|	|||g}tj|| j| j|
dS Nr`  r   r0   rV  )#max_pool2d_checks_and_compute_shapert   r   rL   r  ry   rS   rq   r   r  r   r.  r  	ceil_modenInputPlaneoutputHeightoutputWidthr?  r   r   r7   r7   r8   meta_quantized_max_pool2d  s$   r  c                 C   s   t |  dkd|   d t | dkd|  d t | jt jt jt jfv d| j  t |jt jkd|j  t |jt jkd|j  t |j| jkd|j  | j	| 
d	|
d	| jd
S )Nr   zx must be a 2D tensor, got Dzw must be a 2D tensor, got #expected x to be f32/f16/bf16, got expected w to be uint8, got z q_group_size must be int64, got z5q_scale_and_zeros must have the same dtype as x, got r   r}   )rL   r[   rt   rS   r=  r>  r?  r  r   r   r   rB   r  q_group_sizeZq_scale_and_zerosr7   r7   r8   meta_int4mm_packed_weight_cpu+  s      




r  c                    s4   t   koj k fdd d S )Nc                      s8   d  d d dd   d dj   S )NzExpected a tensor of dimension z and tensor.size[z] == rg   zbut got : dimension z] = rt   r   r7   rt   dim_sizer   r  r7   r8   rX   C  s    z check_dim_size.<locals>.<lambda>)rL   r[   rt   r   )r  rt   r  r   r7   r   r8   check_dim_size@  s   r  c                    s  dd }|d|\}}	t t|dv dd  t  jt jt jt jt jfv fdd t|dkr8||	}
}nt|d	krH|d |d }
}n|d
|\}
}|d|\}}t |d u p_|dkdd    dkro 	dnd	} 	d} 	d} 	d}t
||||
d	|}t
||	||d	|}t }t ||	|
|||d	d	||||||   dkr|||g}n||||g}t j| j j|dS )Nc                    D   t t|dv  fdd |d }t|dkr|n|d }||fS )Nr   r   c                      r  )Nzavg_pool2d: 4 must either be a single int, or a tuple of two intsr7   r7   r  r7   r8   rX   U  r   z1meta_avg_pool2d.<locals>.unpack.<locals>.<lambda>r   r   rL   r[   r   r  rt  HWr7   r  r8   unpackR     

zmeta_avg_pool2d.<locals>.unpackr  r   r   r   c                   S   r]   NzOavg_pool2d: stride must either be omitted, a single int, or a tuple of two intsr7   r7   r7   r7   r8   rX   ^  r_   z!meta_avg_pool2d.<locals>.<lambda>c                      rI  )Nz""avg_pool2d" not implemented for 'rJ  rK  r7   rM  r7   r8   rX   b  rN  r   r   r   r.  c                   S   r]   Nzdivisor must be not zeror7   r7   r7   r7   r8   rX   o  r_   r`  r  r  r   r0   rV  )rL   r[   r   rS   r  uint16uint32uint64rt   r   pooling_output_shaperG   r!   pool2d_shape_checkry   rq   )r   r  r   r.  r  count_include_paddivisor_overrider  kHkWdHdWpadHpadWr?  r  inputHeight
inputWidthr  r  r   r   r7   rM  r8   meta_avg_pool2dH  sj   
	





r   c                 C   sj   t | ||||||dd|	|
|||| |  }|	}t|||d | t|||d | t|||d | d S )Nr   r0   r   )r  rt   r  )r   Z
gradOutputr?  r  r  r  r  r  r  r  r  r  r  r  
mem_formatr   nOutputPlaner7   r7   r8   avg_pool2d_backward_shape_check  s,   r#  c                 C   s  t t|dkpt|dkdd  |d }t|dkr|n|d }	t t|dkp5t|dkp5t|dkdd  t|dkrB|n|d }
t|dkrN|	nt|dkrV|
n|d }t t|dkpgt|dkdd  |d }t|dkrx|n|d }t |d u p|dkdd  |j}| d	kr|d
 nd}|d }|d }|d }t||||
d|}t||	||d|}t|}t|| |||	|
||||||||| t j	||j
|j|dS )Nr   r   c                   S   r]   )NzKavg_pool2d: kernel_size must either be a single int, or a tuple of two intsr7   r7   r7   r7   r8   rX     r_   z*meta_avg_pool2d_backward.<locals>.<lambda>r   c                   S   r]   r  r7   r7   r7   r7   r8   rX     r_   c                   S   r]   )NzGavg_pool2d: padding must either be a single int, or a tuple of two intsr7   r7   r7   r7   r8   rX     r_   c                   S   r]   r  r7   r7   r7   r7   r8   rX     r_   r`  r  r  r  r   rV  )rL   r[   r   r   rt   r  rG   r!   r#  ry   rS   rq   )ZgradOutput_r   r  r   r.  r  r  r  r  r  r  r  r  r  
input_sizer?  r  r  r  r  r  r!  r7   r7   r8   meta_avg_pool2d_backward  sj   "(
r%  c                    s6  t t|dv dd  |d }t|dkr|n|d }t|dkr$|n|d }	t | p2t|dv dd  t  jt jt jt jt jfv fdd |sP|n|d }
|sX|nt|dkr`|
n|d }|sh|	nt|dkrp|
n|d }t t|dv d	d  |d }t|dkr|n|d }t|dkr|n|d }t  jd
v dd  t | p|dkdd   	d} 	d} 	d} 	d} 	d}t
||||
d|}t
||||d|}t
||	||d|}t ||||	|
|||||ddd||||||ddd  jdkr ||||fS  |||||fS )Nr   r0   c                   S   r]   NzFavg_pool3d: kernel_size must be a single int, or a tuple of three intsr7   r7   r7   r7   r8   rX     r_   z!meta_avg_pool3d.<locals>.<lambda>r   r   r   c                   S   r]   NzJavg_pool3d: stride must be omitted, a single int, or a tuple of three intsr7   r7   r7   r7   r8   rX   $  r_   c                      rI  )Nz""avg_pool3d" not implemented for 'rJ  rK  r7   rM  r7   r8   rX   (  rN  c                   S   r]   NzBavg_pool3d: padding must be a single int, or a tuple of three intsr7   r7   r7   r7   r8   rX   0  r_   r`  rq  c                   S   r]   Nz9non-empty 4D or 5D (batch mode) tensor expected for inputr7   r7   r7   r7   r8   rX   8  r_   c                   S   r]   r  r7   r7   r7   r7   r8   rX   =  r_   r  r  r  r   zavg_pool3d()T)check_input_sizer`  )rL   r[   r   rS   r  r  r  r  r   r   r  pool3d_shape_checkr   )r   r  r   r.  r  r  r  kTr  r  dTr  r  padTr  r  r?  nslicesitimeiheightiwidthotimeoheightowidthr7   rM  r8   meta_avg_pool3d  s   

  





r8  c                 C   s  t t|dv dd  |d }t|dkr|n|d }	t|dkr$|n|d }
t | p2t|dv dd  |s;|n|d }|sC|	nt|dkrK|n|d }|sS|
nt|dkr[|n|d }t t|dv dd  |d }t|dkrw|n|d }t|dkr|n|d }t |jd	v d
d  t | p|dkdd  |d}|d}|d}|d}t||||d|}t||	||d|}t||
||d|}t|| |||	|
||||||||||||d ||jS )Nr&  c                   S   r]   r'  r7   r7   r7   r7   r8   rX   w  r_   z*meta_avg_pool3d_backward.<locals>.<lambda>r   r   r   c                   S   r]   r(  r7   r7   r7   r7   r8   rX     r_   c                   S   r]   r)  r7   r7   r7   r7   r8   rX     r_   r*  c                   S   r]   r+  r7   r7   r7   r7   r8   rX     r_   c                   S   r]   r  r7   r7   r7   r7   r8   rX     r_   r  r  r  r   zavg_pool3d_backward())	rL   r[   r   r   r   r  avg_pool3d_backward_shape_checkr   r   )rV  r   r  r   r.  r  r  r  r.  r  r  r/  r  r  r0  r  r  r1  r2  r3  r4  Zotime_for_shape_checkZoheight_for_shape_checkZowidth_for_shape_checkr7   r7   r8   meta_avg_pool3d_backwardi  st   
  




r:  c                    sZ   t  jdkp jdk fdd  jd d t| }t }t j| j j	|dS )Nr0   r`  c                      r|   )Nz"Expected 3D or 4D tensor, but got r   r7   r   r7   r8   rX     r   z*meta_adaptive_avg_pool2d.<locals>.<lambda>r  rV  )
rL   r[   r   r   rZ   rG   r!   ry   rS   rq   )r   output_sizerC  r   r7   r   r8   meta_adaptive_avg_pool2d  s   

r<  c                    s@   t  jdkp jdk fdd   jd d t| S )Nr`  rq  c                      r|   )Nz"Expected 4D or 5D tensor, but got r   r7   r   r7   r8   rX     r   z*meta_adaptive_avg_pool3d.<locals>.<lambda>r  )rL   r[   r   r   r   rZ   )r   r;  r7   r   r8   meta_adaptive_avg_pool3d  s
   
r=  c                    s    j }td|D ]t dk fdd qt|dkp$|dkfdd tj jk fdd tj}trDtj}	j
j|d	S )
Nr   r   c                      r   )Nz{adaptive_avg_pool2d_backward(): Expected grad_output to have non-zero                       size for non-batch dimensions,  with dimension  being emptyr   r7   )grad_outr   r7   r8   rX     s
    z4meta__adaptive_avg_pool2d_backward.<locals>.<lambda>r0   r`  c                      r|   )NzBadaptive_avg_pool2d_backward(): Expected 3D or 4D tensor, but got r   r7   r   r7   r8   rX     r   c                      r  Nexpected dtype z! for `grad_output` but got dtype r}   r7   )r@  r   r7   r8   rX     r   r   )r   r   rL   r[   r   rS   r   r  r  r   r   r  )r@  r   r   r   r7   )r@  r   r   r8   "meta__adaptive_avg_pool2d_backward  s$   

rC  c                 C   s   t | d tj|tjdS )NZadaptive_avg_pool3d_backwardr   )!_adaptive_pool_empty_output_checkrL   r   r  rV  r   r7   r7   r8   "meta__adaptive_avg_pool3d_backward  s   
rF  rV  c                    s<   j }td|D ]tdk fdd qd S )Nr   r   c                      s     dj  d dS )Nzc(): Expected grad_output to have non-zero size for non-batch dimensions, but grad_output has sizes r>  r?  r   r7   r  rV  r   r7   r8   rX     s
   z3_adaptive_pool_empty_output_check.<locals>.<lambda>)r   r   rL   r[   r   )rV  r  r   r7   rG  r8   rD    s   rD  c                    s"  j }t|dv fdd td|D ] t dk fdd qtt|dkdd  d}d}d}j d	krGd}|d7 }|d }|\}}j d
krm|||f}|}	j|tjd}
|	|
fS ||||f}t	}|j
|d}	j|tjdj
|d}
|	|
fS )Nr  c                      r|   )Nz:adaptive_max_pool2d(): Expected 3D or 4D tensor, but got: r   r7   rM  r7   r8   rX     r   z*meta_adaptive_max_pool2d.<locals>.<lambda>r   r   c                         dj  d  dS )Nzjadaptive_max_pool2d(): Expected input to have non-zero size for non-batch dimensions, but input has sizes r>  r?  r   r7   r   r   r7   r8   rX   	  
   r   c                   S   r]   )NzCadaptive_max_pool2d(): internal error: output_size.size() must be 2r7   r7   r7   r7   r8   rX     r_   r`  r0   r}   r   )r   rL   r[   r   r   r   r   r   rG   r!   r  )r   r;  r   ZdimHsizeBsizeDosizeHosizeWr   r   r   r   r7   rI  r8   meta_adaptive_max_pool2d  sD   







rO  c                    sd    j }t|dv  fdd t d tj jk fdd t}jj	|dS )Nr  c                      r|   )NzKadaptive_max_pooling2d_backward(): Expected 3D or 4D grad_output, but got: r   r7   rV  r7   r8   rX   4  r   z3meta_adaptive_max_pool2d_backward.<locals>.<lambda>adaptive_max_pool2d_backwardc                      r  rA  r}   r7   rV  r   r7   r8   rX   ;  r   r   )
r   rL   r[   rD  rS   rG   r!   r   r   r  )rV  r   r   r   r   r7   rR  r8   !meta_adaptive_max_pool2d_backward.  s   



rS  c                    s   j }t|dv fdd td|D ] t dk fdd qtt|dkdd  d}d}d}|d	krFd}|d7 }|}|\}}}|d
kr[||||f}	n|||||f}	|	}
j|	tjd}|
|fS )Nr*  c                      r|   )Nz:adaptive_max_pool3d(): Expected 4D or 5D tensor, but got: r   r7   rM  r7   r8   rX   H  r   z*meta_adaptive_max_pool3d.<locals>.<lambda>r   r   c                      rH  )Nzjadaptive_max_pool3d(): Expected input to have non-zero size for non-batch dimensions, but input has sizes r>  r?  r   r7   rI  r7   r8   rX   M  rJ  r0   c                   S   r]   )NzCadaptive_max_pool3d(): internal error: output_size.size() must be 3r7   r7   r7   r7   r8   rX   U  r_   rq  r`  r}   )r   rL   r[   r   r   r   r   r   )r   r;  r   ZdimDrK  rL  ZosizeTrM  rN  r   r   r   r7   rI  r8   meta_adaptive_max_pool3dB  s8   





rT  c                 C   s   t | d ||jS )Nadaptive_max_pool3d_backward)rD  r   r   )rV  r   r   r7   r7   r8   !meta_adaptive_max_pool3d_backwardn  s   
rV  c                 C   s   |d u rt d| |S )Nz:cannot repeat_interleave a meta tensor without output_size)r  r   )repeatsr;  r7   r7   r8   meta_repeat_interleave_Tensoru  s   
rX  c                 C   s:   | j jsJ |j jsJ t| j|j}| j|t| j dS rb  )rS   r   r(   r   r   r   )realimagr   r7   r7   r8   meta_complex|  s   r[  )
fill_valuer\  c                C   s   | j ||  ftjdS rb  )r   rt   rL   r   )r   r   r\  r7   r7   r8   nonzero_static  s   r]  c                 C   s<   t tjdd  t j|  |  fd|  ft j| jdS )Nc                   S   r]   )NaY  The register_meta function for torch.nonzero() raises unimplemented by default, as a correct data-independent implementation does not exist. This implementation returns a fake value, assuming all elements of the tensor are non-zero. To enable this registration, please set 'torch.fx.experimental._config.meta_nonzero_assume_all_nonzero' to True.r7   r7   r7   r7   r8   rX     r_   znonzero.<locals>.<lambda>r   rS   rq   )	rL   Z_check_not_implementedr  Zmeta_nonzero_assume_all_nonzeror  r   rt   r   rq   r   r7   r7   r8   nonzero  s   
r_  c              
      s@  t tdd  g }tD ]q\d ur|t jt jt jt jt jfv dd  jt jt jfv rv }t	|t 
j jkfdd tjD ]#t 
j j  kfdd ||d qQq| q| q|t t	jkfdd dd lm} t|j t	jk rd  t	jk sd}d	}D ]|dkrǈd urd}q|dkr҈d u rd
}qd ur qqd}|sg }g }tD ]\d ur| | qtD ]\d u r| | q||g g  g tD ]&\}	d u rBr8 j|	  q"j|	  q"tjq" fdd}
   }ddlm} | dkrk|S |
}t|}t|ttt	|krt|j|}t|}t|t|}|| |}|S )Nc                   S   r]   )Nz#at least one index must be providedr7   r7   r7   r7   r8   rX     r_   z#meta_index_Tensor.<locals>.<lambda>c                   S   r]   )Nz?tensors used as indices must be long, int, byte or bool tensorsr7   r7   r7   r7   r8   rX     r_   c                      r|   )N)too many indices for tensor of dimension r  r7   r   r7   r8   rX     r   c                	      s$   dj  d  dj  d  S )NzThe shape of the mask 
 at index z0 does not match the shape of the indexed tensor r   r7   )r   r   jrA  r   r7   r8   rX     s
    r   c                      s   dj  dt  dS )Nr`  z (got rh   )r   r   r7   )r   r   r7   r8   rX     r/  r   Fr   Tc                    sL      }t |  }dgt |tt| jt  < | ||S )zI
        This follows restride_src in TensorAdvancedIndexing.cpp
        r   )r   r   r   r   r   )r   r   r   )after_shapebefore_shapereplacement_shaper7   r8   _restride_src   s    z(meta_index_Tensor.<locals>._restride_srcguard_size_oblivious) rL   r[   rO  	enumeraterS   r   r   r(  r_  r   r   r   r   r   r   selecttorch._refsr   r   r)   r   r   r   rh  r   rG   Z3compute_elementwise_output_logical_to_physical_permZ
apply_permr   Zinvert_permr   r   )r   r   r  r_  refsstateZhas_contiguous_subspacer  Ztransposed_indicesrt   rf  r   rh  Zrestrided_selfpermZ
perm_shaper   r7   )	rc  rd  r   r   r   rb  rA  re  r   r8   meta_index_Tensor  s   










ro  c                 C   sT   d }d }d }|
d r|  | }|
d r|  | }|
d r%|  |}|||fS )Nr   r   r   r   r   )grad_output_input_weight_Zbias_sizes_optr   r.  r  Z
transposedr  r  output_maskZbackend_grad_inputZbackend_grad_weightZbackend_grad_biasr7   r7   r8   meta_convolution_backward!  s   

ru  c                   s     d} d}| ||f} t  dkdd  t dkdd  t  d dk fdd t  d dk fd	d t|  d|ko^|  d|kd
d  | |   S )Nr   r   r0   c                   S   r]   r  r7   r7   r7   r7   r8   rX   E  r_   zmeta_addbmm.<locals>.<lambda>c                   S   r]   r  r7   r7   r7   r7   r8   rX   F  r_   r   c                         d  d d d S )Nz8batch1 and batch2 must have same number of batches, got r   r   r   r7   r  r  r7   r8   rX   I  r  c                
      6   d  d d  d d d d d d	S )Nz#Incompatible matrix sizes for bmm (r   rB   r   r   rh   r   r7   rw  r7   r8   rX   M  
   c                   S   r]   )Nz.self tensor does not match matmul output shaper7   r7   r7   r7   r8   rX   T  r_   )r   r  rL   r[   rt   r   )r   r  r  r2  r1  r  r  r7   rw  r8   meta_addbmm?  s$   

rz  c                 K   s   |  |  S r2   rp  )r   r  kwargsr7   r7   r8   meta_randint_likeY  s   r|  )
grad_scale	found_infc       	            s4   | |||||fD ] t t t fdd qd S )Nc                         dt   S Nz'exponent must be a tensor list but got ri   r7   lr7   r8   rX   t  r  z#meta__fused_adam_.<locals>.<lambda>rL   r[   r`   r   )r   gradsexp_avgsexp_avg_sqsmax_exp_avg_sqsstate_stepslrbeta1beta2weight_decayepsamsgradmaximizer}  r~  r7   r  r8   meta__fused_adam_^  s   
r  c       	            sZ   | |||||fD ] t t t fdd qdd }|| ||||||||fS )Nc                      r  r  r  r7   r  r7   r8   rX     r  z"meta__fused_adam.<locals>.<lambda>c                 S   s   dd | D S )Nc                 S   s   g | ]}t |qS r7   r  )rA   r  r7   r7   r8   rE     rF   z=meta__fused_adam.<locals>.empty_like_list.<locals>.<listcomp>r7   )Ztensor_listr7   r7   r8   empty_like_list  s   z)meta__fused_adam.<locals>.empty_like_listr  )r   r  r  r  r  r  r  r  r  r  r  r  r  r}  r~  r  r7   r  r8   meta__fused_adamx  s   
r  c                    s   t   dkdd  t  dkdd  t  jt ju  fdd t jt ju fdd t  ddk fd	d  j ddft jd
S )Nr   c                   S   r]   )Nza must be a 2D tensorr7   r7   r7   r7   r8   rX     r_   zmeta__int_mm.<locals>.<lambda>c                   S   r]   )Nzb must be a 2D tensorr7   r7   r7   r7   r8   rX     r_   c                      r|   )Nzexpected self to be int8, got r}   r7   )r   r7   r8   rX     r   c                      r|   )Nzexpected mat2 to be int8, got r}   r7   )r  r7   r8   rX     r   r   r   c                
      rx  )Nz'Incompatible matrix sizes for _int_mm (r   rB   r   r   rh   r   r7   r  r7   r8   rX     ry  r}   )rL   r[   rt   rS   r(  r   r   r)  r  r7   r  r8   meta__int_mm  s   



 r  c                    st   t   dkdd  t  jt ju  fdd  d} dd } j|d ||d  d	|d ft jd
S )Nr   c                   S   r]   Nzw must be a 2D tensorr7   r7   r7   r7   r8   rX     r_   z2meta__convert_weight_to_int4pack.<locals>.<lambda>c                      r|   Nr  r}   r7   r  r7   r8   rX     r   r   r      r<      r}   )rL   r[   rt   rS   r  r   r   r)  r  Zinner_k_tilesr  rA  r7   r  r8    meta__convert_weight_to_int4pack  s   



r  c                    s`   t   dkdd  t  jt ju  fdd  d} d} j||d ft jdS )Nr   c                   S   r]   r  r7   r7   r7   r7   r8   rX     r_   z:meta__convert_weight_to_int4pack_for_cpu.<locals>.<lambda>c                      r|   Nzexpected w to be int32, got r}   r7   r  r7   r8   rX     r   r   r   r}   )rL   r[   rt   rS   r)  r   r   r  r  r7   r  r8   (meta__convert_weight_to_int4pack_for_cpu  s   




r  c                    s   t  dkdd  t   dkdd  t jt jt jt jfv fdd t  jt ju  fdd j	d 	dd	 jd
S )Nr   c                   S   r]   Nzx must be a 2D tensorr7   r7   r7   r7   r8   rX     r_   z*meta__weight_int4pack_mm.<locals>.<lambda>r`  c                   S   r]   )Nzw must be a 4D tensorr7   r7   r7   r7   r8   rX     r_   c                      r|   Nr  r}   r7   r   r7   r8   rX     r   c                      r|   r  r}   r7   r  r7   r8   rX     r   r   r  r}   
rL   r[   rt   rS   r=  r>  r?  r)  r   r   r  r7   r  rB   r8   meta__weight_int4pack_mm  s   


"r  c                       t  dkdd  t   dkdd  t jt jt jt jfv fdd t  jt ju  fdd j	d 	djdS )	Nr   c                   S   r]   r  r7   r7   r7   r7   r8   rX     r_   z2meta__weight_int4pack_mm_for_cpu.<locals>.<lambda>c                   S   r]   r  r7   r7   r7   r7   r8   rX     r_   c                      r|   r  r}   r7   r   r7   r8   rX     r   c                      r|   r  r}   r7   r  r7   r8   rX     r   r   r}   )
rL   r[   rt   rS   r=  r>  r?  r  r   r   r  r7   r  r8    meta__weight_int4pack_mm_for_cpu     


r  c                    r  )	Nr   c                   S   r]   r  r7   r7   r7   r7   r8   rX     r_   z;_weight_int4pack_mm_with_scales_and_zeros.<locals>.<lambda>c                   S   r]   r  r7   r7   r7   r7   r8   rX     r_   c                      r|   r  r}   r7   r   r7   r8   rX     r   c                      r|   r  r}   r7   r  r7   r8   rX     r   r   r}   r  )rB   r  r  ZqScaleZqZerosr7   r  r8   )_weight_int4pack_mm_with_scales_and_zeros  r  r  r   r  c                 C   s   | | d | | S r  r7   r  r7   r7   r8   kai_roundup  s   r  c           	         s   | dkrv||kr/d}d}d}d
dddd 
fddfd	d
}||||||S |d dkrx|| dkrzd}d}d}d
ddd  fdd} 	
fdddd  fdd fdd	|||||||S d S d S d S )Nr`  r  r<  r   c                 S   s   t || d}t | |S )Nr`  r  )rA  krsrZkr_sr_roundedup4r7   r7   r8   kai_k_roundedup  s   
z3get_kai_packed_weight_size.<locals>.kai_k_roundedupc                    s8    | ||}|d dksJ d||d     S )Nr   r   zk_internal must be evenr7   )rA  nrr  r  Z
k_internal)r  kai_num_bytes_biaskai_num_bytes_multiplier_rhskai_num_bytes_sum_rhsr7   r8   9kai_get_rhs_packed_stride_rhs_pack_nxk_qsi4cxp_qsu4cxs1s0  s   z]get_kai_packed_weight_size.<locals>.kai_get_rhs_packed_stride_rhs_pack_nxk_qsi4cxp_qsu4cxs1s0c                    s    t | || }| |||| S r2   r  )r  rA  r  r  r  num_rows)r  r7   r8   7kai_get_rhs_packed_size_rhs_pack_nxk_qsi4cxp_qsu4cxs1s0'  s   z[get_kai_packed_weight_size.<locals>.kai_get_rhs_packed_size_rhs_pack_nxk_qsi4cxp_qsu4cxs1s0r  r   c                    sR   || dksJ | dksJ |  dksJ t | || }|||||| S rM  r  )r  rA  r  r  r  blr  )kai_bl_multiple_of;kai_get_rhs_packed_stride_rhs_pack_nxk_qsi4c32p_qsu4c32s1s0kai_nr_multiple_ofr7   r8   9kai_get_rhs_packed_size_rhs_pack_nxk_qsi4c32p_qsu4c32s1s0?  s   
z]get_kai_packed_weight_size.<locals>.kai_get_rhs_packed_size_rhs_pack_nxk_qsi4c32p_qsu4c32s1s0c                    s^   || dksJ | dksJ |  dksJ  }| |}||}|||    S rM  r7   )rA  r  r  r  r  num_bytes_multiplier_rhsZnum_blocks_per_rowZnum_bytes_per_block)r  #kai_get_bf16_datatype_size_in_bytesr  kai_num_blocks_per_rowr  kai_num_bytes_per_blockr  r7   r8   r  O  s   
z_get_kai_packed_weight_size.<locals>.kai_get_rhs_packed_stride_rhs_pack_nxk_qsi4c32p_qsu4c32s1s0c                   S   r]   )Nr   r7   r7   r7   r7   r8   r  e  r  zGget_kai_packed_weight_size.<locals>.kai_get_bf16_datatype_size_in_bytesc                    s   |  dksJ t | || S rM  r  )rA  r  r  r7   r8   r  h  s   z:get_kai_packed_weight_size.<locals>.kai_num_blocks_per_rowc                    s   |   dksJ | d | S )Nr   r   r7   )r  r  r  r7   r8   r  l  s   z;get_kai_packed_weight_size.<locals>.kai_num_bytes_per_blockr7   )	Zn_bitsr  KZ	groupsizeZkai_nrZkai_krZkai_srr  r  r7   )r  r  r  r  r  r  r  r  r  r  r  r8   get_kai_packed_weight_size  s@   
-r  c                    s   t  jt ju  fdd t jj rE||kr|jt jks4||k rE|d dkrE|| dkrE|jt jkrEt	d|||} j
t|t jdS   |  } j
|t jdS )Nc                      r|   r  r}   r7   weightsr7   r8   rX   {  r   z2meta__dyn_quant_pack_4bit_weight.<locals>.<lambda>r  r   r`  r}   )rL   r[   rS   r  backendsZkleidiaiis_availablerO   r?  r  r   r   r   )r  Zscales_zerosr$  
block_sizein_featuresout_featuresZpacked_weight_sizer7   r  r8    meta__dyn_quant_pack_4bit_weightu  s.   





	r  c                    sR   t   dkdd  t  jt jfv  fdd  d} j|| jdS )Nr   c                   S   r]   )Nzinput must be a 2D tensorr7   r7   r7   r7   r8   rX     r_   z-meta__dyn_quant_matmul_4bit.<locals>.<lambda>c                      r|   )Nzexpected input to be f32, got r}   r7   inpr7   r8   rX     r   r   r}   )rL   r[   rt   rS   r=  r   r   )r  Zpacked_weightsr  r  r  r  r7   r  r8   meta__dyn_quant_matmul_4bit  s   

r  c                    s   t  dkdd  t jt jt jt jfv fdd t   dkdd  t  jt ju  fdd j	d 	djdS )	Nr   c                   S   r]   r  r7   r7   r7   r7   r8   rX     r_   z*meta__weight_int8pack_mm.<locals>.<lambda>c                      r|   r  r}   r7   r   r7   r8   rX     r   c                   S   r]   r  r7   r7   r7   r7   r8   rX     r_   c                      r|   )Nzexpected w to be int8, got r}   r7   r  r7   r8   rX     r   r   r}   )
rL   r[   rt   rS   r=  r>  r?  r(  r   r   )rB   r  Zq_scalesr7   r  r8   meta__weight_int8pack_mm  s   


r  c           	         s  t  dkfdd t  dkfdd t ddkfdd t tjdd  t tjdd  t |d	kd
d  t  dv  fdd d}d}jd d }jd d }tt 	||}|
||g |S )Nr   c                         d    dS )Nz1cdist only supports at least 2D tensors, X1 got: r  r   r7   )x1r7   r8   rX     rY   z$meta_cdist_forward.<locals>.<lambda>c                      r  )Nz1cdist only supports at least 2D tensors, X2 got: r  r   r7   )x2r7   r8   rX     rY   r   c                      rv  )Nz4X1 and X2 must have the same number of columns. X1: r   z X2: r   r7   )r  r  r7   r8   rX     r  c                   S   r]   )Nz=cdist only supports floating-point dtypes, X1 got: {x1.dtype}r7   r7   r7   r7   r8   rX     r_   c                   S   r]   )Nz=cdist only supports floating-point dtypes, X2 got: {x2.dtype}r7   r7   r7   r7   r8   rX     r_   r   c                   S   r]   )Nz)cdist only supports non-negative p valuesr7   r7   r7   r7   r8   rX     r_   Nr   r   c                      r  )Nz%possible modes: None, 1, 2, but was: r7   r7   )compute_moder7   r8   rX     r  r  )rL   r[   rt   r   rG   is_float_dtyperS   r   r   broadcast_shapesextendr   )	r  r  r}  r  r1r2batch_tensor1batch_tensor2rC  r7   )r  r  r  r8   meta_cdist_forward  s@   









r  c                 C   s   |j d }|j d }|j d }|j d d }|j d d }	tt||	}
|
 }|||g t|
}|dksE|dksE|dksE|dkrJt|S |t|j krV|	|}tj
|tjdS )Nr   r  r   r   )r   r   rL   r  copyr  mathprod
zeros_liker  r   r   )r  r  r  r}  Zcdistc1r  r  r  r  r	  Ztensor1_expand_sizeZbatch_productr7   r7   r8   meta_cdist_backward  s   



 

r  c	                    s  t  jt jt jfv  fdd t jt jt jfv fdd t tjfdd d}	|rEt |	dkdd  |	d8 }	|	d}
d urzt |t	kdd  t j
dkfd	d t    k fd
d fdddd fdd}tdkr  d}  }|tkr |	d}q d}nL||
|}|ttfv s|s͈ d}nd}|	}jd }|tkr|rt |dkdd  |d8 }|jd }n| }|
|||fS )Nc                      r|   )Nz(expected indices to be long or int, got r}   r7   )r   r7   r8   rX     r   z$meta_embedding_bag.<locals>.<lambda>c                      r|   )Nz(expected offsets to be long or int, got r}   r7   )rQ  r7   r8   rX     r   c                      r|   )Nz/expected weight to be floating point type, got r}   r7   )r"  r7   r8   rX     r   r   r   c                   S   r]   Nz1include_last_offset: numBags should be at least 1r7   r7   r7   r7   r8   rX     r_   c                   S   r]   )Nz@embedding_bag: per_sample_weights only supported with mode='sum'r7   r7   r7   r7   r8   rX     r_   c                      r  )Nz1expected per_sample_weights to be 1D tensor, got r  r  r7   )per_sample_weightsr7   r8   rX     r  c                      s   d   d    dS )Nz%expected per_sample_weights.numel() (z$ to be the same as indices.numel() (rh   r   r7   )r   r  r7   r8   rX     s   c                    s    | ||o| ddkS Nr   r   r   r  r  r  padding_idx)is_fast_path_index_selectr7   r8   is_fast_path_index_select_scale  s   z;meta_embedding_bag.<locals>.is_fast_path_index_select_scalec                 S   s<   | j tjks| j tjko| ddko|ddko|dk S Nr   r   )rS   rL   rO   rM   r   )r  r  r  r7   r7   r8   r  "  s   z5meta_embedding_bag.<locals>.is_fast_path_index_selectc                    s"   |d ur| |||S  | ||S r2   r7   r  )r  r  r7   r8   is_fast_path*  s   z(meta_embedding_bag.<locals>.is_fast_pathcpuc                   S   r]   r  r7   r7   r7   r7   r8   rX   D  r_   )rL   r[   rS   r   r   rG   r  r   r   MODE_SUMr   r   r   MODE_MAX	MODE_MEANr   )r"  r   rQ  scale_grad_by_freqr  sparser  Zinclude_last_offsetr  Znum_bagsr  r  
offset2bagbag_sizemax_indicesZfast_path_sumZnumBagsr7   )r   r  r  rQ  r  r"  r8   meta_embedding_bag  st   








r  c                 G   sB   t | ||g|R  \}}}}t|dkr|| }||||fS )Nr  )r  r   r   r   )r"  r   rQ  rI   r  r  r  r  r7   r7   r8   meta_embedding_bag_forward_onlyM  s   r  c                 C   s.   |r|S | j js| j jr| j S |rtjS | j S r2   )rS   r   r   rL   r   )r   rS   promote_int_to_longr7   r7   r8   _get_reduction_dtypeW  s   r  r}   c                C   s6   t | |dd}t| j|}t| ||}| j||dS )NT)r  r}   )r  rG   rd  r   re  r   )r   r  rg  rS   r  rC  r7   r7   r8   meta_nansumd  s   r  c                 C   s$   t | jtt|  }| |S r2   )rG   r  r   rZ   r   rt   r   )r   rC  r7   r7   r8   meta_medianm  s   
r  c                 C   sL   t | dkrtd t| j|f}t| ||}| || j|tjdfS )Nr   zmedian CUDA with indices outputr}   )	r   rG   alert_not_deterministicrd  r   re  r   rL   r   )r   rt   rg  rC  r7   r7   r8   meta_median_mode_dimu  s   
r  c                 C   rK  r2   r7   r   r7   r7   r8   meta_logical_not_  rv  r  c                    s   t t|  kdd  tD ]\ t dk fdd qt|   }d| t| j fddttD }| |S )Nc                   S   r]   )NzZNumber of dimensions of repeat dims can not be smaller than number of dimensions of tensorr7   r7   r7   r7   r8   rX     r_   zmeta_repeat.<locals>.<lambda>r   c                      rU   )Nz"Repeats cannot be negative, found ra  r7   r7   )r   repr7   r8   rX     rY   r  c                    s   g | ]
} | |  qS r7   r7   r  )padded_sizerW  r7   r8   rE     r  zmeta_repeat.<locals>.<listcomp>)	rL   r[   r   rt   ri  rZ   r   r   r   )r   rW  Znum_new_dimensionsZtarget_sizer7   )r   r  r  rW  r8   meta_repeat  s   
r  c                 C   rK  r2   r7   r   r7   r7   r8   
meta_zero_  rv  r  c                 C   s   t |tjrt| j|j | S r2   )r`   rL   r   r\   r   r   r   r7   r7   r8   meta_binop_inplace  s   r   c                 C   sf   dd }dd }dd }|| r||rt d|| r$||s$t dt|tjr1t| j|j | S )	a*  
    Some checks for inplace ops.
    Checks for promotion rules for some dtypes.
    int.add/sub_(float) and bool.add/sub_(others) are rejected.
    Promoting in these in-place operations would require reallocating
    and copying over elements, hence not allowed.
    Checks for alpha param.
    c                 S       t | trt| jS t | tS r2   )r`   r"   rG   r  rS   r   rb   r7   r7   r8   is_integeric     

z.meta_binop_inplace_alpha.<locals>.is_integericc                 S   r  r2   )r`   r"   rG   r  rS   r   r  r7   r7   r8   
is_floatic  r  z,meta_binop_inplace_alpha.<locals>.is_floaticc                 S   r  r2   )r`   r"   rG   Zis_boolean_dtyperS   r   r  r7   r7   r8   is_booleanic  r  z.meta_binop_inplace_alpha.<locals>.is_booleanicz]Promotion of int.add/sub_(float) in in-place ops are not possible due to element size change.z_Promotion of book.add/sub_(others) in in-place ops are not possible due to element size change.)r  r`   rL   r   r\   r   )r   r   r1  r  r  r  r7   r7   r8   meta_binop_inplace_alpha  s   r  c                 K      t | tjdS Nr>   rK   r   rH   )r   r{  r7   r7   r8   
meta_round  s   r  c                    sl   t tj fdd tt jr&t tj fdd d S t tt fdd d S )Nc                           dj  S )Nz7: Expected input tensor to have an integral dtype. Got r}   r7   )r  r   r7   r8   rX     rY   z#shift_dtype_check.<locals>.<lambda>c                      r  )Nz6: Expected shift value to have an integral dtype. Got r}   r7   r  rt  r7   r8   rX     rY   c                      s     d S )Nz): Expected shift value to be an int. Got r7   r7   r  r7   r8   rX     r  )rL   r[   rG   r  rS   r`   r   r   r  r   rt  r7   r  r8   shift_dtype_check  s   

r  c                 C      t d| | t| |tjdS )Nrshiftr
  r  rK   r   rH   r  r7   r7   r8   meta_rshifts     r  c                 C   r  )Nlshiftr
  r  r  r7   r7   r8   meta_lshifts  r  r  c                 C      |  | jS r2   r  r   r7   r7   r8   	meta_zero     r  c                 C   rK  r2   r7   r   rt  r7   r7   r8   
meta_fill_  rv  r  c                 C   r  r2   r  r  r7   r7   r8   	meta_fill!     
r  c                 C   rK  r2   r7   r   r7   r7   r8   
meta_relu_&  rv  r  c                 C      t | |tjdS r	  r  )r   r   r1  r7   r7   r8   meta__add_relu+     r!        ?UUUUUU?c                 C   r  r2   r  r   noiselowerr  r  r  r7   r7   r8   meta_rrelu_with_noise3  s   
r(  c                 C   s   t | t |fS r2   r  r%  r7   r7   r8    meta_rrelu_with_noise_functional;  s   r)  c                 C   rK  r2   r7   )r   r'  r  r  r  r7   r7   r8   meta_rrelu_with_noise_B  s   r*  c                 C   r  r2   r  r   r   r   
accumulater7   r7   r8   meta_index_putI  r  r-  c                 C   s   t | j|j | S r2   r\   r   )r   r  valuer7   r7   r8   meta_masked_fill_N  s   r0  c                 C   s    |  |  jt| d}|S r   )r   r   r  rG   r!   )r   r  r  Zmasked_scaler7   r7   r8   meta__masked_scaleT  s   r1  c                    s@   t |jt jt jfv dd  t  jjk fdd  S )Nc                   S   r]   )NzMask must be bool or uint8r7   r7   r7   r7   r8   rX   _  r_   z&meta_masked_scatter_.<locals>.<lambda>c                      r+  )NzEmasked_scatter: expected self and source to have same dtypes but got r   r}   r7   r   rF  r7   r8   rX   c  s
    )rL   r[   rS   rO  r  )r   r  rF  r7   r2  r8   meta_masked_scatter_\  s   
r3  c                 C   s*   t | |\} }tj| tjd}t|||S r   )r)   rL   r   r   r3  )r   r  rF  r  r7   r7   r8   meta_masked_scatteri  s   r4  c                 C   s
   |  |S r2   r^  )r   r  r  r7   r7   r8   meta_masked_scatter_backwardq  r  r5  c                 C   rK  r2   r7   r+  r7   r7   r8   meta_index_put_v  rv  r6  c                 C   r  r2   )viewr   r   r7   r7   r8   
meta_alias{  r  r8  c           
         s8  t |  dkdd  t | dkdd  |  }|  |d |d |d } d }||ft  d koB d k fdd |rt| jt jkpX| jt jko]|t jk}t || jkpf|d	d  |	|}	n|}	|sd urt  dkd
d  t  kfdd |	S )Nr0   c                   S   r]   r  r7   r7   r7   r7   r8   rX     r_   z)common_meta_baddbmm_bmm.<locals>.<lambda>c                   S   r]   r  r7   r7   r7   r7   r8   rX     r_   r   r   r   c                	      r  r  r7   r7   r  r7   r8   rX     s    c                   S   r]   )Nzfout_dtype only supported for torch.float32 output with float16/bfloat16 inputs or same as input dtypesr7   r7   r7   r7   r8   rX     r_   c                   S   r]   )Nzself must be a 3D tensorr7   r7   r7   r7   r8   rX     r_   c                      s   d  d   S )Nz*Expected an input tensor shape with shape z but got shape: r   r7   )r;  self_baddbmmr7   r8   rX     r   )
rL   r[   rt   r   rS   r>  r?  r=  r   r  )
r  r  Zis_bmmr9  r&  r  Zres_rowsZres_colsZsupported_out_dtyper  r7   )r  r  r  r;  r9  r8   common_meta_baddbmm_bmm  s>   


r:  c                 C   s   t | |dS )NTr:  )r   r.  r7   r7   r8   meta_bmm  r  r<  c                 C   s   t | |d|dS )NT)r&  r;  )r   r.  r&  r7   r7   r8   meta_bmm_dtype  s   r=  c                 C   s<   | | }| | }|dkrt |dk t |dk kr|d8 }|S r  )rO  )rB   yqr  r7   r7   r8   div_rtn  s
    r@  c                 C   sZ   t | | | ||d   d |r|d nd |d }|r+|d | | | kr+|d8 }|S r  )r@  )	inputSize
kernelSizer8  r9  r   r  r  Z
outputSizer7   r7   r8   pooling_output_shape_pad_lr  s*   

	rC  c                    sl   t |dkdd  t dkfdd t d   d d k fdd t| | |S )Nr   c                   S   r]   )Nzstride should not be zeror7   r7   r7   r7   r8   rX     r_   z&pooling_output_shape.<locals>.<lambda>c                      r  )Nz'pad must be non-negative, but got pad: r7   r7   padr7   r8   rX     r  r   r   c                      s   d d d  S )NzApad should be at most half of effective kernel size, but got pad=z, kernel_size=z and dilation=r7   r7   r  rB  rE  r7   r8   rX     s
   )rL   r[   rC  )rA  rB  rE  r   r  r  r7   rF  r8   r    s   r  c              	      sN     }tdkodkdd  t|dko|dkdd  t|dko+|dkdd   ddko= ddk}|tjkrWt|dkoQ|oQ d	dkd
d  n"t|d	krf ddkrf|pr|dkor|or d	dk fdd td 
kod 	k	
fdd tdkodkfdd d S )Nr   c                   S   r]   )NzCkernel size should be greater than zero, but got kH: {kH}, kW: {kW}r7   r7   r7   r7   r8   rX     r_   z$pool2d_shape_check.<locals>.<lambda>c                   S   r]   )Nz>stride should be greater than zero, but got dH: {dH}, dW: {dW}r7   r7   r7   r7   r8   rX     r_   c                   S   r]   )Nz\dilation should be greater than zero, but got dilationH: {dilationH}, dilationW: {dilationW}r7   r7   r7   r7   r8   rX     r_   r   r   r`  r0   c                   S   r]   )NzExpected 4D (batch mode) tensor expected for input with channels_last layout with optional 0 dim batch size for input, but got: {input.size()}r7   r7   r7   r7   r8   rX     r_   c                         d    S )NzYExpected 3D or 4D (batch mode) tensor with optional 0 dim batch size for input, but got: r   r7   rM  r7   r8   rX     r  c                      s   d d d d  S )NzKpad should be smaller than or equal to half of kernel size, but got padW = z	, padH = z, kW = z, kH = r7   r7   )r  r  r  r  r7   r8   rX     s    c                      s*   d d  d d d d dS NzGiven input size: (rB   z). Calculated output size: (z). Output size is too smallr7   r7   )r  r  r  r"  r  r  r7   r8   rX     s    )rt   rL   r[   r   r  )r   r  r  r  r  r  r  	dilationH	dilationWr  r  r  r  r  r   r   Z
valid_dimsr7   )r   r  r  r  r  r  r"  r  r  r  r  r8   r    sB   

r  r1  r.  r  r  r/  r  r  pTpHpW	dilationTrI  rJ  r2  r3  r4  r5  r6  r7  r,  c              
      s  	j }tdkodkodkfdd tdko&dko& dk fdd tdko<dko<dkfdd t|dv 	fdd t|D ]|dkradkraqVt	dk	fd	d qV|rt
kokok
fd
d td kod kod kfdd tdkodkodk
fdd d S )Nr   c                         d d  d S )Nz5kernel size should be greater than zero, but got kT: z, kH: z, kW: r7   r7   )r  r.  r  r7   r8   rX   A     z$pool3d_shape_check.<locals>.<lambda>c                      rO  )Nz0stride should be greater than zero, but got dT: z, dH: z, dW: r7   r7   )r  r/  r  r7   r8   rX   H  s   c                      rO  )Nz9dilation should be greater than zero, but got dilationT: z, dilationH: z, dilationW: r7   r7   )rI  rN  rJ  r7   r8   rX   N  rP  r*  c                      r  )Nz/: Expected 4D or 5D tensor for input, but got: r   r7   )r  r   r7   r8   rX   V  rY   rq  c                      s     dj  d dS )NzZ: Expected input's non-batch dimensions to have positive length, but input has a shape of z and non-batch dimension z has length zero!)r   r   r7   )r  r   r   r7   r8   rX   _  s
   c                      s*   d d  d d d d dS )Nzinput image (T: ru  rg  z ) smaller than kernel size (kT:  kH:  kW: rh   r7   r7   )r3  r2  r4  r  r.  r  r7   r8   rX   i  s   r   c                      s(   d d d  d d d S )NzHpad should be smaller than or equal to half of kernel size, but got kT: rR  rQ  z padT: z padW: z padH: r7   r7   )r  r.  r  rL  rK  rM  r7   r8   rX   q  s   r   c                      s6   d d d  d d d d d dS rH  r7   r7   )r3  r2  r4  r1  r6  r5  r7  r7   r8   rX   y  s   )r   rL   r[   r   r   )r   r1  r.  r  r  r/  r  r  rK  rL  rM  rN  rI  rJ  r2  r3  r4  r5  r6  r7  r  r,  r   r7   )r  r/  r  rI  rN  rJ  r  r   r3  r   r2  r4  r  r.  r  r1  r6  r5  r7  rL  rK  rM  r8   r-  %  sJ   	"r-  c                 C   s   | j }t| |||||||	|
|||||||||||| t|||d | t|||d | t|||d | t|||d | t|||d | t|||d | t|||d | t|||d | d S )Nr`  r0   r   r   r   r-  r  )r   rV  r   r1  r.  r  r  r/  r  r  rK  rL  rM  rN  rI  rJ  r2  r3  r4  r5  r6  r7  r  r   r7   r7   r8   max_pool3d_backward_shape_check  s@   rT  c                 C   s   | j }t| ||||||||	|
|ddd|||||||d t|||d | t|||d | t|||d | t|||d | d S )Nr   Tr`  r0   r   rS  )r   rV  r1  r.  r  r  r/  r  r  rK  rL  rM  r2  r3  r4  r5  r6  r7  r  r   r7   r7   r8   r9    s:   r9  c                 C   sB  dd }|d|\}}t t|dv dd  t|dkr#||}	}
n|d|\}	}
|d	|\}}|d
|\}}| d}| d}| d}t| }|t jkr^t |  dkdd  n|t jkrpt |  dv dd  nt ddd  t	||||	||}t	||||
||}t
| |||	|
|||||||||| |||fS )Nc                    r  )Nr  c                      r  )Nzmax_pool2d: r  r7   r7   r  r7   r8   rX     r   zEmax_pool2d_checks_and_compute_shape.<locals>.unpack.<locals>.<lambda>r   r   r  r  r7   r  r8   r    r  z3max_pool2d_checks_and_compute_shape.<locals>.unpackr  r  c                   S   r]   )NzOmax_pool2d: stride must either be omitted, a single int, or a tuple of two intsr7   r7   r7   r7   r8   rX     r_   z5max_pool2d_checks_and_compute_shape.<locals>.<lambda>r   r   r.  r  r  r  r   r`  c                   S   r]   )NzMnon-empty 4D (batch mode) tensor expected for input with channels_last layoutr7   r7   r7   r7   r8   rX     r_   r  c                   S   r]   )Nz9non-empty 3D or 4D (batch mode) tensor expected for inputr7   r7   r7   r7   r8   rX   !  r_   Fc                   S   r]   )Nz?Unsupport memory format. Supports only ChannelsLast, Contiguousr7   r7   r7   r7   r8   rX   &  r_   )rL   r[   r   r   rG   r!   r  rt   r   r  r  )r   r  r   r.  r  r  r  r  r  r  r  r  r  rI  rJ  r  r  r  r   r  r  r7   r7   r8   r    sb   		









r  c                    s   t |||||\}tj jk fdd |jfdd}	|	  |	| t}
tjjjj	|
dS )Nc                      r  )NzExpected dtype z  for `gradOutput` but got dtype r}   r7   rE  r7   r8   rX   V  r   z7meta_max_pool2d_with_indices_backward.<locals>.<lambda>c                    s:   t | d   t | d  t | d  d S )Nr0   r   r   )r  )r  )r"  r   r  r  r7   r8   _check_dim_size\  s   z>meta_max_pool2d_with_indices_backward.<locals>._check_dim_sizerV  )
r  rL   r[   rS   r   rG   r!   ry   r   rq   )rV  r   r  r   r.  r  r  r   r  rU  r   r7   )rV  r"  r   r  r  r   r8   %meta_max_pool2d_with_indices_backwardA  s.   

rV  c                 C   s   t | |||||\}}}|  dkr| dnd}	t| }
|  dkr*|||g}n|	|||g}tj|| j| j|
dtj|tj	| j|
dfS r  )
r  rt   r   rG   r!   rL   ry   rS   rq   r   r  r7   r7   r8   meta_max_pool2d_with_indicesm  s2   
rW  c           
   	      s  t jdv fdd j}t|d |D ] t  dkd  d  d qt td	kd
d  t t|d	kdd  d}dd|dkr_d}nd}t jjkdd  t jdkfdd d}d}d	 t ||kd t ||kdd  t  d	k fdd t |d d  d kfdd t |d d  d kfdd  dkr|||d |d g}	n	||d |d g}	t j|	jj	dt j|	t j
j	dfS )Nr  c                      r|   )Nz:fractional_max_pool2d: Expected 3D or 4D tensor, but got: r  r7   r   r7   r8   rX     r   z,meta_fractional_max_pool2d.<locals>.<lambda>r0   r   z^fractional_max_pool2d: Expected input to have non-zero  size for non-batch dimenions, but got r>  z emptyr   c                   S   r]   )NzNfractional_max_pool2d: kernel_size musteither be a single int or tuple of Intsr7   r7   r7   r7   r8   rX     r_   c                   S   r]   )NzOfractional_max_pool2d: output_size must either be a single int or tuple of Intsr7   r7   r7   r7   r8   rX     r_   r  r  r   r`  r   c                   S   r]   )Nz6Expect _random_samples to have the same dtype as inputr7   r7   r7   r7   r8   rX     r_   c                      r|   )Nz1Expect _random samples to have 3 dimensions got, r  r7   )random_samplesr7   r8   rX     r   z=Expect _random_samples.size(0) no less then input batch size.c                   S   r]   )Nz<Expect _random_samples.size(1) equals to input channel size.r7   r7   r7   r7   r8   rX     r_   c                      r  )Nz/Expect _random_samples.size(2) equals to 2 got .r7   r7   )r   r7   r8   rX     r   c                         dd  d  S )Nz%fractional_max_pool2d: kernel height r   z' is too large relative to input height r7   r7   )input_heightr  r7   r8   rX     r   c                      rZ  )Nz$fractional_max_pool2d: kernel width r   z& is too large relative to input width r7   r7   )input_widthr  r7   r8   rX     r   r^  )rL   r[   r   r   r   r   rS   rt   ry   rq   r   )
r   r  r;  rX  r   Zinput_channelsZinput_batchr  cr   r7   )r   r[  r\  r  rX  r   r8   meta_fractional_max_pool2d  s   










r^  c                 C   s  t t|dv dd  |d }t|dkr|n|d }t|dkr$|n|d }t | p2t|dv dd  |s;|n|d }	|sC|nt|dkrK|	n|d }
|sS|nt|dkr[|	n|d }t t|dv dd  |d }t|dkrw|n|d }t|dkr|n|d }t t|dv d	d  |d }t|dkr|n|d }t|dkr|n|d }t | jd
v dd  | jdkr| dnd}| d}| d}| d}| d}t||||	||}t||||
||}t||||||}t| |||||	|
|||||||||||||d | jdkot| t j	k}| jdkr:| 
d}|  o2|jt j	d}||||f}n|||||f}| |}| j|t jd}|r_|jt j	d}|jt j	d}||fS )Nr&  c                   S   r]   NzMmax_pool3d: kernel_size must either be a single int, or a tuple of three intsr7   r7   r7   r7   r8   rX     r_   z.meta_max_pool3d_with_indices.<locals>.<lambda>r   r   r   c                   S   r]   NzQmax_pool3d: stride must either be omitted, a single int, or a tuple of three intsr7   r7   r7   r7   r8   rX     r_   c                   S   r]   NzImax_pool3d: padding must either be a single int, or a tuple of three intsr7   r7   r7   r7   r8   rX     r_   c                   S   r]   NzJmax_pool3d: dilation must be either a single int, or a tuple of three intsr7   r7   r7   r7   r8   rX     r_   r*  c                   S   r]   r+  r7   r7   r7   r7   r8   rX     r_   rq  r  r  r  r   zmax_pool3d_with_indices()r`  r   r}   )rL   r[   r   r   r   r  r-  rG   r!   r  r   r   r   r   r  )r   r  r   r.  r  r  r.  r  r  r/  r  r  rK  rL  rM  rN  rI  rJ  r?  r1  r2  r3  r4  r5  r6  r7  r  input_channels_last_checkr   r   r   r7   r7   r8   meta_max_pool3d_with_indices  s   

  







re  c                 C   s^  t t|dv dd  |d }t|dkr|n|d }	t|dkr$|n|d }
t | p2t|dv dd  |s;|n|d }|sC|	nt|dkrK|n|d }|sS|
nt|dkr[|n|d }t t|dv dd  |d }t|dkrw|n|d }t|dkr|n|d }t t|dv d	d  |d }t|dkr|n|d }t|dkr|n|d }t |jd
v dd  |d}|d}|d}|d}| d}| d}| d}t|| ||||	|
|||||||||||||||d |jdkot|t jk}|jdkr|	d}|
  o|j
t jd}||j}|r-|jt jd}|S )Nr&  c                   S   r]   r_  r7   r7   r7   r7   r8   rX   `  r_   z7meta_max_pool3d_with_indices_backward.<locals>.<lambda>r   r   r   c                   S   r]   r`  r7   r7   r7   r7   r8   rX   h  r_   c                   S   r]   ra  r7   r7   r7   r7   r8   rX   p  r_   c                   S   r]   rb  r7   r7   r7   r7   r8   rX   x  r_   r*  c                   S   r]   r+  r7   r7   r7   r7   r8   rX     r_   r  r  r  r   z"max_pool3d_with_indices_backward()rq  r`  r   )rL   r[   r   r   r   rT  rG   r!   r  r   r   r   r   r  )rV  r   r  r   r.  r  r  r   r.  r  r  r/  r  r  rK  rL  rM  rN  rI  rJ  r1  r2  r3  r4  r5  r6  r7  r  rd  rY  r7   r7   r8   %meta_max_pool3d_with_indices_backwardR  s   
  









rf  gridc                    s   t j jk fdd t jt jko jt jk fdd t jd  jd k fdd t  jd jd k fdd tdjD ]t j dkfd	d qPd S )
Nc                      r  )NzNgrid_sampler(): expected input and grid to be on same device, but input is on z and grid is on r{  r7   rg  r   r7   r8   rX     r  z+check_grid_sampler_common.<locals>.<lambda>c                      r  )NzTgrid_sampler(): expected input and grid to have torch.strided layout, but input has z and grid has )rp   r7   rh  r7   r8   rX     r  r   c                      r  )NzZgrid_sampler(): expected grid and input to have same batch size, but got input with sizes  and grid with sizes r   r7   rh  r7   r8   rX     r  r   r   c                      s   dj d  d j S )Nz+grid_sampler(): expected grid to have size r   z, in last dimension, but got grid with sizes )r   r   r7   rh  r7   r8   rX     s   c                      rH  )NzYgrid_sampler(): expected input to have non-empty spatial dimensions, but input has sizes r>  r?  r   r7   rI  r7   r8   rX     rJ  )rL   r[   rq   rp   r!  r   r   r   )r   rg  r7   )rg  r   r   r8   check_grid_sampler_common  s,   
rj  c                   @   s   e Zd ZdZdZdZdS )GridSamplerInterpolationr   r   r   N)rj   
__module____qualname__ZBILINEARZNEARESTBICUBICr7   r7   r7   r8   rk    s    rk  interpolation_modec                    sP   t jdkoj jk fdd t jdko |tjjk dd  d S )Nrq  c                      r  )Nzdgrid_sampler(): expected 5D input and grid with same number of dimensions, but got input with sizes ri  r   r7   rh  r7   r8   rX     s
   z'check_grid_sampler_3d.<locals>.<lambda>c                   S   r]   )Nz<grid_sampler(): bicubic interpolation only supports 4D inputr7   r7   r7   r7   r8   rX     r_   )rL   r[   r   rk  rn  r/  )r   rg  ro  r7   rh  r8   check_grid_sampler_3d  s   

rp  c           
      C   s:   |d }|rt j|t jd}nd }t j|t jd}	||	fS Nr   r   )rL   r  r   r   
rV  r   rg  ro  padding_modealign_cornersrt  Zinput_requires_gradrY  	grad_gridr7   r7   r8   grid_sampler_2d_backward_meta  s   
rv  c           
      C   s\   t | | t| || | jd }| jd }|jd }|jd }|jd }	| |||||	fS )Nr   r   r   r0   )rj  rp  r   r   )
r   rg  ro  rs  rt  r  CZout_DZout_HZout_Wr7   r7   r8   grid_sampler_3d  s   
	




rx  ru  c           
      C   sP   t || t||| |d }|rtj|tjd}nd }tj|tjd}	||	fS rq  )rj  rp  rL   r  r  r   rr  r7   r7   r8   grid_sampler_3d_backward  s   
ry  c                 O   s:   | dd }|st|}||d< tj| g|R i |S )NrS   )rR   rG   Z	get_dtyperL   ry   )r   r\  rI   r{  rS   r7   r7   r8   full7  s
   
rz  c                 C   s   |t jkrJt |d u dd  t jd|d u r| jn|||d u r"| jn||d}| jr8||  | 	 | 
  n||  |  d |d |S tjj| |||||d}|d |S )Nc                   S   r]   )Nz9memory format option is only supported by strided tensorsr7   r7   r7   r7   r8   rX   M  r_   zzeros_like.<locals>.<lambda>r   r  Trz  )rL   Z
sparse_coor[   ry   rS   rq   	is_sparseZsparse_resize_and_clear_r   
sparse_dim	dense_dimrt   Z_coalesced_r.   r   r  fill_)r   rS   rp   rq   rr   r   r  r7   r7   r8   r  A  s:   
	

	r  ro   c                C   B   |d u rt  }|d u rt  }|d u rt j}t j| ||||dS r	  rL   rv   Zget_default_devicer!  ry   r   rS   rp   rq   rr   rs   r7   r7   r8   	meta_onesn     
r  c                C   r  r	  r  r  r7   r7   r8   
meta_zeros  r  r  c           	         s   ddl m}  }t|dkdd   dkr n |   }t| |kp1||k  fdd dkrAn| t }t } |    }| = | = 	|||S )Nr   rg  c                   S   r]   )Nz-select() cannot be applied to a 0-dim tensor.r7   r7   r7   r7   r8   rX     r_   zmeta_select.<locals>.<lambda>c                      s   d d   d  S )Nzselect(): index z! out of range for tensor of size z at dimension r   r7   rt   r   r   r7   r8   rX     s
    )
r   rh  rt   rL   r   r   r   r   r   r   )	r   rt   r   rh  r   r   new_sizer   Znew_storage_offsetr7   r  r8   meta_select  s(   
r  c                 C   r  r2   rG   Zclone_preserve_strides)r   r  rt   r   r7   r7   r8   meta_select_scatter  r  r  c                 C   r  r2   r  )r   r  rt   rl   rk   stepr7   r7   r8   meta_slice_scatter  r  r  dim_post_exprwrap_scalarc                 C   sb   |dkr
|sJ d}| }|d }| |k s| |kr'J d|  d| d| d| dk r/| |7 } | S )Nr   r   zdim z out of bounds (rg   rh   r7   )rt   r  r  r   r   r7   r7   r8   r     s   ,r   c                 C   s   |   dkrdS | j| S r  r  )r  rt   r7   r7   r8   ensure_nonempty_size  s   r  c                    st   t  d}t  d}t||kdd  t|D ] kr7tttk fdd qd S )Nr   c                   S   r]   )NzDIndex tensor must have the same number of dimensions as input tensorr7   r7   r7   r7   r8   rX     r_   z$gather_shape_check.<locals>.<lambda>c                      s$   d dj  dj  d   S )Nz!Size does not match at dimension z expected index  to be no larger than self  apart from dimension r   r7   rt   r   r   r   r7   r8   rX     s    )r   rt   rL   r[   r   r  )r   rt   r   	self_dimsZ
index_dimsr7   r  r8   gather_shape_check  s   r  c                    sn   ddl m} t||  }|  dk}|s1t jtjkp$ jtj	k fdd t
| |  |  jS )Nr   rg  c                      r|   )Nz8gather(): Expected dtype int32/int64 for index, but got r}   r7   r~   r7   r8   rX     r   zmeta_gather.<locals>.<lambda>)r   rh  r   rt   r   rL   r[   rS   r   r   r  r   r   )r   rt   r   Zsparse_gradrh  wrapped_dimZis_index_emptyr7   r~   r8   meta_gather  s   
r  c                 C   s   |r*| dkrdS | dkrdS | dkrdS | dkrdS | d	kr d
S t ddd  d S | dkr0dS | dkr6dS t ddd  d S )Nr  Z
REDUCE_ADDr  ZREDUCE_MULTIPLYmeanZREDUCE_MEANZamaxZREDUCE_MAXIMUMZaminZREDUCE_MINIMUMFc                   S   r]   )Nz=reduce argument must be either sum, prod, mean, amax or amin.r7   r7   r7   r7   r8   rX     r_   z#get_operator_enum.<locals>.<lambda>addmultiplyc                   S   r]   )Nz/reduce argument must be either add or multiply.r7   r7   r7   r7   r8   rX     r_   r  )reduce_use_new_optionsr7   r7   r8   get_operator_enum  s,   r  c                    sp   ddl m} || dkr"t|jtjkp|jtjk fdd |d ur6t|j|jk fdd d S d S )Nr   rg  c                      
     dS )Nz((): Expected dtype int32/int64 for indexr7   r7   method_namer7   r8   rX     r  z,scatter_gather_dtype_check.<locals>.<lambda>c                      r  )Nz0(): Expected self.dtype to be equal to src.dtyper7   r7   r  r7   r8   rX     r  )r   rh  r   rL   r[   rS   r   r   )r  r   r   src_optrh  r7   r  r8   scatter_gather_dtype_check  s   


r  c                 C   s
   t | dS r  )r   r   r7   r7   r8   ensure_nonempty_dim"  s   
r  c           	         s0  ddl m} | dkrd S tt t kdd  d}t }t|D ]}t|}| kr:q.|t|krEd} qFq.|scd urct|D ]}t|}|t|krbd} qcqPd urtt t kdd  t|  fdd d S t|  fd	d d S )
Nr   rg  c                   S   r]   NzCIndex tensor must have the same number of dimensions as self tensorr7   r7   r7   r7   r8   rX   .  r_   z%scatter_shape_check.<locals>.<lambda>FTc                   S   r]   r  r7   r7   r7   r7   r8   rX   H  r_   c                      s&   dj  dj  d  dj   S )NExpected index r  r  z and to be no larger than src r   r7   rt   r   r   r  r7   r8   rX   L  s    c                      s   dj  dj  d   S )Nr  r  r  r   r7   r  r7   r8   rX   R  s    )	r   rh  r   rL   r[   r  rt   r   r  )	r   rt   r   r  rh  Zis_wrong_shaper  r   Zindex_d_sizer7   r  r8   scatter_shape_check'  sJ   

r  c                 C   sD   t ||  }td| || t| ||| |d ur t|| d S d S )Nscatter)r   rt   r  r  r  )r   rt   r   r  r  r  r  r7   r7   r8   scatter_meta_implX  s   r  c                 C   s   t | |||d | | jS Nr  r  r   r   r   rt   r   r  r7   r7   r8   meta_scatter_adda  s   r  c                 C   s   t | |||d | S r  r  r  r7   r7   r8   meta_scatter_add_g  r  r  c                 C   s0   t |tjr|nd }t| |||| | | jS r2   )r`   rL   r   r  r   r   r   rt   r   Zsrc_or_valuer   r  r7   r7   r8   meta_scatterm  s   
r  c                 C   s(   t |tjr|nd }t| |||| | S r2   )r`   rL   r   r  r  r7   r7   r8   meta_scatter_|  s   	r          queryr   r/  	dropout_p	is_causalreturn_debug_maskr  c              	   C   sJ  |  d}|  d}|  d}	|  d}
| d}| dd}t|dd}tj|||	ftj| jd}|rb|
dkr=dnd}t|	| }|dkrMd}n|dkrSd}tj|||	|f| j	| jd}n
tjd| j	| jd}tj
jrtj rtjd	tjd
d}tjd	tjd
d}ntjdtjd
d}tjd	tjd
d}||d d |	||||f	S )Nr   r   r   r0   r^  @         r7   rn   )r   r  rL   r   ry   rO   rq   r  ceilrS   versionhipr   r  r   r  )r  r   r/  r  r  r  r  r   	num_headsmax_seqlen_batch_qhead_dimmax_seqlen_batch_kquery_t	attention	logsumexpblocksize_cmax_seqlen_k
debug_maskseedoffsetr7   r7   r8   (meta__scaled_dot_product_flash_attention  sP   






r  	res_shape.c                    s   t jkrdd}t|dd}|S tg dfdddd fdd	 D } fd
d	tt D }tj|j	j
d|}|S )Nr   r   )r   r   r   r0   c                    s      |  S r2   r  )idx)r  r7   r8   rX     r   z,alloc_with_matching_layout.<locals>.<lambda>Tr   c                    s   g | ]} | qS r7   r7   )rA   r  )r  r7   r8   rE     r   z.alloc_with_matching_layout.<locals>.<listcomp>c                    s   g | ]}  |qS r7   r~   r  )	dim_orderr7   r8   rE     rF   r^  )rZ   r   r  rL   r   sortedr   r   ry   rS   rq   r   )r  r  r  r  Zpermuted_shapeZfinal_permuter7   )r  r  r  r8   alloc_with_matching_layout  s   
r  	attn_biascompute_log_sumexpc	              	   C   s   |  d}	|  d}
|  d}| d}| d}|	|
||f}t| |}tj|	|
|ftj| jd}tjdtjdd}tjdtjdd}||d d ||||d f	S Nr   r   r   r   r^  r7   rn   r   r  rL   ry   rO   rq   r   )r  r   r/  r  r  r  r  r  r  r  r	  S_QS_KVD_Vr  r  
logsum_expr  r  r7   r7   r8   (meta__scaled_dot_product_cudnn_attention  s0   





r  c              	   C   s   |  d}|  d}	|  d}
| d}| d}||	|
|f}t| |}tj||	|
ftj| jd}tjdtjdd}tjdtjdd}||d d |
|||d f	S r  r  )r  r   r/  r  r  r  r  r  r  ZH_Qr  r  r  r  r  r  r  r  r7   r7   r8   5meta__scaled_dot_product_fused_attention_overrideable  s0   





r  r@  r  	cum_seq_q	cum_seq_kmax_qmax_kphilox_seedphilox_offsetc                 C   sX   t |dddd}t |dddd}t |dddd}|||fS r  )rL   r   r  )r@  r  r   r/  r   r  r  r  r  r  r  r  r  r  r  grad_qgrad_kgrad_vr7   r7   r8   'meta__scaled_dot_product_flash_backward6  s   
r  	attn_maskc                 C   sR   |  d}|  d}|  d}	t| }
tj||	|ftj| jddd}|
|fS )Nr   r   r   r^  )r   rL   r   ry   rO   rq   r  )r  r   r/  r  r  r  r  r   r  r  r  r  r7   r7   r8   0meta__scaled_dot_product_flash_attention_for_cpuR  s"   




r  c
                 C   s   | d}
| d}| d}| d}| d}tj|
|||fd|j|jd}tj|
|||fd|j|jd}tj|
|||fd|j|jd}|||fS )Nr   r   r0   r   r   r   r   r0   r^  )r   rL   empty_permutedrS   rq   )r@  r  r   r/  r   r  r  r  r  r  r   r  r  len_qZlen_kr  r  r  r7   r7   r8   9meta__scaled_dot_product_flash_attention_for_cpu_backwardt  s0   








r  c                 C   s   |  dd} | dd}| dd}| d}| d}	| d}
|d}tj||	|
|| j| jd}tjjrDtj	 rD	 |rA|	nd}n|rOt
|	d d nd}tj||
|ftj| jd}| dd}tjdtjd	d}tjdtjd	d}||||fS )
Nr   r   r   r  r   r^  r  r7   rn   )r  r   rL   ry   rS   rq   r  r  r   r  r  r  rO   r   )r  r   r/  r  r  r  r  r  r  r  r  Kvr  logsumexp_dimr  r  r  r7   r7   r8   ,meta__scaled_dot_product_efficient_attention  s*   



r  grad_input_maskc                 C   s  | d}| d}| d}| d}| d}| d}tj||||fd|j|jd}tj||||fd|j|jd}tj||||fd|j|jd}d }|d ur|
d r| d}|d dkrb|n|d |d  }t|  }||d< tj||j|jd}|d	d |f }||||fS )
Nr   r   r   r0   r  r^  r   r<  .)r   rL   r  rS   rq   r   ry   )r@  r  r   r/  r  r   r  r  r  r  r  r  r  r   r  r  r  Z
head_dim_vr  r  r  r  	grad_biaslastDimlastDimAligned	new_sizesr7   r7   r8   +meta__scaled_dot_product_efficient_backward  sF   









 
r  c                 C   s(   t |}t |}t |}|||fS r2   r  )r@  r  r   r/  r   r  r  r  r  r  r  r  r  r  r  r  r  r  r  r7   r7   r8   'meta__scaled_dot_product_cudnn_backward  s   



r  window_size_leftwindow_size_right	seqused_kalibi_slopesc                 C   s  |d u r	|  dn| d }|d u r|  dn|}|d u r#| dn|}|  d}|  d}t| }|d u rFtj|||ftj| jd}n|  d}tj||ftj| jd}|	r|dkr_dnd}t|| }|dkrod}n|dkrud}tj||||f| j	| jd}n
tjd| j	| jd}d	\}}tj
jrtj rtjd
tjdd}tjd
tjdd}ntjdtjdd}tjd
tjdd}|||||fS )Nr   r   r  r   r^  r  r  r  NNr7   rn   r   )r   r   rL   r   ry   rO   rq   r  r  rS   r  r  r   r  r   r  )r  r   r/  r  r  r  r  r  r  r  r  r  r  r  r  r   r  r  r  r  r  r  Ztotal_qr  r  r  r  r  r7   r7   r8   meta__flash_attention_forward)  sR   




r  c                 C   s(   t |}t |}t |}|||fS r2   r  )r@  r  r   r/  r   r  r  r  r  r  r  r  r  r  r  r  r  
grad_querygrad_key
grad_valuer7   r7   r8   meta__flash_attention_backwardy  s   



r  cu_seqlens_qcu_seqlens_kmax_seqlen_qr  custom_mask_typecausal_diagonalseqlen_kwindow_sizec                 C   s   |  d}|  d}| d}|  d}| d}tj||||| j| jd}|d ur1| dd n|}|}|d urA|d us?J |}|d urG|n|}|
rTt|d d nd}tj|||ftj| jd}tjdtjdd}tjdtjdd}||||||fS )	Nr   r   r  r   r^  r  r7   rn   )	r   rL   ry   rS   rq   r  r  rO   r   )r  r   r/  r$  r  r  r  r  r  r  r  r  r   r  r  r  r  r  r  r  r  Zlogsumexp_batch_dimZactual_max_seqlen_qZactual_max_seqlen_kr  r  r  r  r7   r7   r8   !meta__efficient_attention_forward  s,   




r  bias_requires_gradnum_splits_keyshared_storage_dqdkdvc                 C   sL  |rSt |jd |jd kdd  t |jd |jd kdd  t jg |jdd d|jd |jd R |j|jd	}|d
d}|d
d}|d
d}nt |}t |}t |}|d ur|d}|d dkrs|n|d |d  }t	| }||d< t j||j|jd	}|dd |f }nt jd|jd}||||fS )Nr   c                   S   r]   )Nz,seqlen must match for `shared_storage_dqdkdvr7   r7   r7   r7   r8   rX     r_   z4meta__efficient_attention_backward.<locals>.<lambda>r0   c                   S   r]   )Nz3embedding dim must match for `shared_storage_dqdkdvr7   r7   r7   r7   r8   rX     r_   r   r  r   r^  r  r   r<  .r7   r{  )
rL   r[   r   ry   rS   rq   rj  r   r   r   )r@  r  r   r/  r$  r  r  r  r  r  r  r  r  r  r  r  r  r  chunkr  r  r  r  r  r  r  r7   r7   r8   "meta__efficient_attention_backward  s:   *



 r  scale_ascale_bscale_resultuse_fast_accumc                    sn  dd }t  dko dkfdd t |jo$|jfdd tdkrdd	 }	d
d }
dd }t |	 pJ|fdd t |
 p\|fdd t dd dkfdd t dd dkodd dkfdd j\}djt jkojt jkpjt j	kojt j	k}
 dkrψ
 dkrt jt jkoɈjt jkdd  n|r7jt j	krd}|d }nd}d}dd }|||}||dd }||| |  ||| | 
  kr)
 kr)t  dd  t  dd  qt d fdd nft jt jkoEjt jkdd  t  dkoX dkfd d dkrddkrddkrdkrt  o d!d  nt dfd"d |d ur|nj}t jdd|jd#S )$Nc                 S   s   | t jt jt jt jt jfv S r2   )rL   r@  Zfloat8_e5m2Zfloat8_e4m3fnuzZfloat8_e5m2fnuzZfloat4_e2m1fn_x2r}   r7   r7   r8   is_fp8_or_fp4_type  s   z*meta_scaled_mm.<locals>.is_fp8_or_fp4_typer   c                      s   d   d    S )Nz%Inputs must be 2D but got self.dim()=z and mat2.dim()=r   r7   r.  r   r7   r8   rX     r/  z meta_scaled_mm.<locals>.<lambda>c                      r  )Nz?Expected both inputs to be fp8 or fp4 types but got self.dtype=z and mat2.dtype=r}   r7   r  r7   r8   rX   "  r   r   c                 S   s   | d | d ko| d dkS r  r7   r  r7   r7   r8   is_row_major'     z$meta_scaled_mm.<locals>.is_row_majorc                 S   s   | d dko| d dkS r  r7   r  r7   r7   r8   is_col_major*  r  z$meta_scaled_mm.<locals>.is_col_majorc                 S   s   |  ddkp|  ddkS r  r   )Z	tensor_2dr7   r7   r8   has_zero_dim-  r  z$meta_scaled_mm.<locals>.has_zero_dimc                      rG  )Nz#self must be row_major, got stride r  r7   r   r7   r8   rX   2  r  c                      rG  )Nz#mat2 must be col_major, got stride r  r7   r.  r7   r8   rX   6  r  r   r<  r   c                      s   d  d S )NzBExpected self.size(1) to be divisible by 16, but got self.size(1)=r   r   r7   r   r7   r8   rX   :  rY   c                      r|   )Nz>Expected both dimensions of mat2 to be divisble by 16 but got r   r7   r  r7   r8   rX   >  r   c                   S   r]   )NzNFor tensorwise scaling, both scale_a and scale_b must be float (fp32) tensors.r7   r7   r7   r7   r8   rX   R  r_   r  r  c                 S   s   | | d | S r  r7   r  r7   r7   r8   ceil_divb  r:   z meta_scaled_mm.<locals>.ceil_divr`  c                   S   r]   )Nzscale_a must be contiguousr7   r7   r7   r7   r8   rX   u  r_   c                   S   r]   )Nzscale_b must be contiguousr7   r7   r7   r7   r8   rX   y  r_   Fc                	      s&   d  d   d d   d	S )NzTInvalid blockwise scaling configuration. For blockwise scaling, scale_a should have  elements, got z, scale_b should have rY  r  r7   )expected_a_sizeexpected_b_sizer	  r
  r7   r8   rX   ~  s   c                   S   r]   )NzKFor rowwise scaling, both scale_a and scale_b must be float (fp32) tensors.r7   r7   r7   r7   r8   rX     r_   c                      s   d   d  S )NzLFor non-tensorwise scaling, scale tensors must be 2D, but got scale_a.dim()=z and scale_b.dim()=r   r7   )r	  r
  r7   r8   rX     r/  c                   S   r]   )Nz@Both scale_a and scale_b must be contiguous for rowwise scaling.r7   r7   r7   r7   r8   rX     r_   c                      sB   d  d d d d d d d d d dS )	Nz}Invalid scaling configuration. For tensorwise scaling, both scales should be scalar. For rowwise scaling, scale_a should be (z, 1), scale_b should be (1, z). Got scale_a.size()=(r   rg   r   z) and scale_b.size()=(rh   r   r7   )rB  r  r	  r
  r7   r8   rX     s   r^  )rL   r[   rt   rS   r   r   r   r   Zfloat8_e8m0fnur@  r   r=  r   ry   rq   )r   r.  r	  r
  r$  r  r&  r  r  r  r  r  Z_kZis_blockwise_scalingZblock_size_kZblock_size_mnr  Znum_k_blocksZpadded_num_k_blocksZ
_out_dtyper7   )r  r  rB  r.  r  r	  r
  r   r8   meta_scaled_mm  s   	


"








	 r  c                 C   s    t | ||||dd | | jS NT)r  r  r   rt   r   r  r   rE  r7   r7   r8   meta_scatter_reduce_two  s   r  c                 C   s   t | ||||dd | S r  r  r  r7   r7   r8   meta_scatter_reduce__two  s   r  c                   sh   t d    k odkn   fdd   dkr&t j|t j jdS t j d|t j jdS )Nr   r   c                      rG  )Nz@The probabilty distributions dimensions must be 1 or 2, but got r   r7   rM  r7   r8   rX     r  z"meta_multinomial.<locals>.<lambda>r   r^  )rL   r[   rt   ry   r   rq   r   )r   num_samplesreplacementr  r7   rM  r8   meta_multinomial  s   
r  c                 C   s   d}| D ]}||9 }q|S r  r7   )vsr  vr7   r7   r8   multiply_integers  s   
r"  c                    s   t tkfdd d  t t k fdd t tdd dd  D o9tdd D fdd d d \}}||gR S )Nc                         d  dt  S )Nz%It is expected output_size equals to , but got size r-  r7   )num_spatial_dimsr;  r7   r8   rX     r   z'upsample_common_check.<locals>.<lambda>r   c                      r#  )Nz$It is expected input_size equals to r$  r-  r7   )expected_input_dimsr$  r7   r8   rX     r   c                 s   r  r  r7   r   r7   r7   r8   rc     r  z(upsample_common_check.<locals>.<genexpr>c                      re   )NzDInput and output sizes should be greater than 0, but got input size z and output size r7   r7   )r$  r;  r7   r8   rX     s
    )rL   r[   r   r  )r$  r;  r%  r?  Zchannelsr7   )r&  r$  r%  r;  r8   upsample_common_check  s   

*r'  c                    sZ   t   dkpt  dd   fdd t  |dd} |jt	 dS )Nr   r   c                      rG  )Nz>Non-empty 3D data tensor expected but got a tensor with sizes r   r7   rM  r7   r8   rX     r  z$upsample_nearest1d.<locals>.<lambda>r%  r   
rL   r[   r   r"  r   r'  r   r  rG   r!   )r   r;  scalesfull_output_sizer7   rM  r8   upsample_nearest1d     


r,  c           	         s   t   dkpt  dd   fdd t  |dd} |}t } j	\}}}} j
jdkr?|dk r?t j}|j|d	}|S )
Nr   r   c                      rG  Nz>Non-empty 4D data tensor expected but got a tensor with sizes r   r7   rM  r7   r8   rX     r  z$upsample_nearest2d.<locals>.<lambda>r   r(  r   r`  r   )rL   r[   r   r"  r   r'  r   rG   r!   r   rq   ri   r   
contiguous)	r   r;  scales_hscales_wr+  r  r   rJ   Z
n_channelsr7   rM  r8   upsample_nearest2d  s   



r2  r;  r$  r0  r1  c                    st   t ||dd tjdkfdd tdD ]t  k fdd q|jt	dS )Nr   r(  r`  c                      r|   NzFExpected grad_output to be a tensor of dimension 4 but got: dimension r  r7   rP  r7   r8   rX     r   z-upsample_nearest2d_backward.<locals>.<lambda>c                
      &   d d   d d  S )NzCExpected grad_output to have the same shape as output; output.size() = z but got grad_output.size(r   r7   r+  rV  r   r7   r8   rX   #  s   r   )
r'  rL   r[   r   r   r   r   r  rG   r!   )rV  r;  r$  r0  r1  r7   r6  r8   upsample_nearest2d_backward  s   

	r7  c                    sZ   t   dkpt  dd   fdd t  |dd} |jt	 dS )Nr   r   c                      rG  )Nz>Non-empty 5D data tensor expected but got a tensor with sizes r   r7   rM  r7   r8   rX   5  r  z$upsample_nearest3d.<locals>.<lambda>r0   r(  r   r)  )r   r;  Zscales_dr0  r1  r+  r7   rM  r8   upsample_nearest3d/  r-  r8  c           
      C   s   t | t j| t jd}}|d urQ|d urQt|tsJ t|ts$J |j}| }	t||}t||}|||	 |||	 t	||d t	||d ||fS ||fS )Nr}   )r  r  )
rL   r   r   r`   r"   r   r   r$   r   r&   )
r   stablert   Z
descendingr   r   r!  r   r   
out_strider7   r7   r8   	meta_sort?  s   	

r;  c                    s  t jdkfdd t jjkfdd dd urPt jdkfdd t  kfdd t jjkfdd t jdkfd	d d
   t   k fdd t tfddfD dd  d S )Nr   c                          j  dS Nz != 2r  r7   input_gatesr7   r8   rX   b  r   z%rnn_cell_checkSizes.<locals>.<lambda>c                         j  d j  S N != r   r7   )hidden_gatesr?  r7   r8   rX   e  rN  r   c                      r<  )Nz != 1r  r7   )
input_biasr7   r8   rX   i  r   c                      s      d  S rA  r  r7   )
gates_sizerD  r7   r8   rX   l  rN  c                      r@  rA  r   r7   )hidden_biasrD  r7   r8   rX   p  rN  c                      r<  r=  r  r7   )prev_hiddenr7   r8   rX   r  r   r   c                
      s,      dd d d d  d
S )NrB  r   z * z // z (aka rh   )r   r   r7   )expected_prev_hidden_numelfactorrE  r?  rG  r7   r8   rX   v  s   , c                 3   s    | ]	}|j  j kV  qd S r2   r{  r@   r>  r7   r8   rc   y  s
    

z&rnn_cell_checkSizes.<locals>.<genexpr>c                   S   r]   )Nz%expected all inputs to be same devicer7   r7   r7   r7   r8   rX   }  r_   )rL   r[   r   r   r   r   r  )r?  rC  rD  rF  rI  rG  r7   )rH  rI  rE  rF  rC  rD  r?  rG  r8   rnn_cell_checkSizesZ  s8   





rJ  c                 C   sL   t | |||d| tj| tjd}tj|tjd}tj|tjd}|||fS )Nr`  r   )rJ  rL   r   r   )r?  rC  cxrD  rF  	workspacehycyr7   r7   r8   _thnn_fused_lstm_cell_meta  s
   
rO  c                 C   s(  t |dk}|rt |}|d }| jd }n|
r| jd n| jd }|
r)| jd n| jd }d}|r4dnd}|dkr<|n|}|rG||| g}n|
rP|||| gn|||| g}| |}|	| ||g}|d u rptjd| jd}n||}||	| ||g}|rdnd}| j|tjd}|||||fS )Nr   r   r   r   r{  r}   )r   r   r   rL   ry   rq   r  )r   r"  Zweight_stride0Z
weight_bufhxrK  r  hidden_sizeZ	proj_size
num_layersbatch_firstZdropouttrainbidirectionalbatch_sizesZdropout_stateZis_input_packed
seq_length
mini_batchZbatch_sizes_sumZnum_directionsout_sizer   r  Z
cell_shaperN  rM  Zreserve_shapeZreserver7   r7   r8   
_cudnn_rnn  s2   

rZ  c                 C   s   |r| j d n| j d }|r| j d n| j d }|
}|r!|||gn|||g}| |}|d u r8tjd| jd}n||j }|d u rKtjd| jd}n||j }tjd| jtjd}||||fS )Nr   r   r{  r   )r   r   rL   ry   rq   r  )r   Zw0Zw1Zw2Zw3hx_Zcx_r   rV  r  rQ  rR  
has_biasesrU  rS  rT  rW  rX  Zoutput_chanelsr   r  rM  rN  rL  r7   r7   r8   mkldnn_rnn_layer  s    
r]  c                    sT   | j dkrt dkp dk fdd d S t|  dk fdd d S )Nr   r   c                      r  )Nz4: Expected reduction dim -1 or 0 for scalar but got r7   r7   rt   r  r7   r8   rX     r  z'zero_numel_check_dims.<locals>.<lambda>c                      r  )Nz: Expected reduction dim z to have non-zero size.r7   r7   r^  r7   r8   rX     rY   )r   rL   r   r   )r   rt   r  r7   r^  r8   zero_numel_check_dims  s   
r_  c                    sF   |d urt || }t||  d S t| dk fdd d S )Nr   c                      r  )Nz@: Expected reduction dim to be specified for input.numel() == 0.r7   r7   r  r7   r8   rX     r  z%check_argmax_argmin.<locals>.<lambda>)r   rt   r_  rL   r[   r   )r  r   rt   r7   r  r8   check_argmax_argmin  s   

r`  c                 C   sD   t d| | t| j|d ur|fnd }t| ||}| j|tjdS )Nargmaxr}   )r`  rG   rd  r   re  r   rL   r   )r   rt   rg  r  r   r7   r7   r8   argmax_argmin_meta	  s   rb  c                 C   s$   |t jkrt j}t jd||||dS )Nr7   r  )rL   Zjaggedr!  ry   )r   rS   rp   rq   rr   r7   r7   r8   scalar_tensor  s
   

rc  c                 C   s   t ||  dd}|  dkrdn| |}t| t||kdd  t| j}t|dkr4|||< | 	|| j	|tj
dfS )NTr  r   r   c                   S   r]   )Nzk not in range for dimensionr7   r7   r7   r7   r8   rX   #  r_   ztopk_meta.<locals>.<lambda>r}   )r   rt   r   rL   r  r[   r   r   r   r   r   )r   rA  rt   Zlargestr  Z	sliceSizeZtopKSizer7   r7   r8   	topk_meta  s   

re  c           
      C   s@   |d us|d usJ d|  }|   }	tj||	j|	j|	jdS )Nz;segment_reduce(): Either lengths or offsets must be defined)rS   rq   rp   )r/  rL   r   rS   rq   rp   )
r  r  rU  r   rP  rQ  rR  rT  Zdata_contigZgrad_contigr7   r7   r8   meta__segment_reduce_backward+  s   rf  c                    s   ddl m} t |  dd |  dkr|  nd}t||dk||k fdd t| jd   | j d d   }|rM|  dkrM|	 d | 
|| j
|tjdfS )	Nr   )sym_andTrd  r   c                      r  )Nz9kthvalue(): selected number k out of range for dimension r7   r7   r   r7   r8   rX   F  r  zkthvalue_meta.<locals>.<lambda>r}   )r   rg  r   rt   r   rL   r[   r   r   r  r   r   )r   rA  rt   rg  rg  ZdimSizer   r7   r   r8   kthvalue_meta=  s   
$rh  c                 C   s   | d ur| n|}t | dkdd  | }| d ur(t |  |kdd  |d ur8t | |kdd  t | |kdd  t | |kdd  t | dkdd  t | |d	 |d
  d kdd  d S )Nr   c                   S   r]   N r7   r7   r7   r7   r8   rX   U  r_   z(checkLSTMBackwardSizes.<locals>.<lambda>c                   S   r]   ri  r7   r7   r7   r7   r8   rX   X  r_   c                   S   r]   ri  r7   r7   r7   r7   r8   rX   Z  r_   c                   S   r]   ri  r7   r7   r7   r7   r8   rX   [  r_   c                   S   r]   ri  r7   r7   r7   r7   r8   rX   \  r_   c                   S   r]   ri  r7   r7   r7   r7   r8   rX   ]  r_   r   r   r`  c                   S   r]   ri  r7   r7   r7   r7   r8   rX   ^  r_   )rL   r[   rt   r   r   )grad_hygrad_cyrK  rN  rL  Zdefined_gradZexp_sizer7   r7   r8   checkLSTMBackwardSizesS  s   ,rm  c           	      C   s`   | d u r
|d u r
dS t | |||| tj|td}tj|td}|r)|jdddnd }|||fS )NNNNr   r   F)rg  )rm  rL   r   legacy_contiguous_memory_formatr  )	rk  rl  rK  rN  rL  Zhas_biasZ
grad_gatesZgrad_cxr  r7   r7   r8   #_thnn_fused_lstm_cell_backward_implb  s   
rp  c                 C   sf   d }d }d }|d r| |  }|d s|d r.| |d| df}| |d}|||fS )Nr   r   r   r   rp  )rr  rq  rs  rt  rY  grad_weightr  r7   r7   r8   linear_backwardp  s   
rr  c                    s   t jdkrjd ||  dksJ dj d| dd   fdd	}jd ||  }jd
 | }jd | }g jd d |||R }|}|j| d}|S )Nr   r  r   z'Invalid input shape for pixel_shuffle: z with upscale_factor = c                 S   r  r2   r  r  r7   r7   r8   r    r  z,meta_pixel_shuffle.<locals>.is_channels_lastc                      sL    rt dkrtjS tjS jtjdrtjS jtjdr$tjS d S r  )r   rL   r   r  r   r  r7   r  r   r7   r8   r    s   z.meta_pixel_shuffle.<locals>.pick_memory_formatr  r   r   )r   r   r   r  )r   Zupscale_factorr  rw  ZHrZWrr   r   r7   rs  r8   meta_pixel_shuffle}  s    
rt  c                 C   sZ   |  | j}| |j}| |j}| |j}| |j}| |j}|||||||fS r2   r  )r   Zweight0Zweight1Zweight2Zweight3r[  Zcx_tmpr  Zhy_Zcy_Zgrad_output_r_optZgrad_hy_r_optZgrad_cy_r_optr   r  rQ  rR  r\  rT  rU  rV  rS  rL  Zdiff_xZdiff_hxZdiff_cxZdiff_w1Zdiff_w2Zdiff_br7   r7   r8   mkldnn_rnn_layer_backward  s   ru  )	out_int32r   c                C   s   t j| |rt jnt jt jdS )NrS   r   )rL   r   r)  r   r   )r   Z
boundariesrv  r   r7   r7   r8   meta_bucketize  s
   rx  d   c                    s   dt dkrt fdd t dkr# r#td tt t fdd t dk fd	d tttfd
d tttfdd tkdd  tj	 j
jdS )Nzhistc()r  c                      r  )Nz%"histogram_cpu" not implemented for 'rJ  r}   r7   rM  r7   r8   rX     r  zmeta_histc.<locals>.<lambda>r   z%_histc_cuda with floating point inputc                      s    dt   S )Nz#: argument 'bins' must be int, not r  r7   binsr  r7   r8   rX     rN  r   c                      r  )Nz: bins must be > 0, but got r7   r7   rz  r7   r8   rX     r  c                           dt  S )Nz%: argument 'min' must be Number, not r  r7   )r  r   r7   r8   rX     rN  c                      r|  )Nz%: argument 'max' must be Number, not r  r7   )r  r   r7   r8   rX     rN  c                   S   r]   )Nz&{fn_name}: max must be larger than minr7   r7   r7   r7   r8   rX     r_   r   )r   rL   r[   r   rG   r  r`   r   r    ry   rq   rS   )r   r{  r   r   r7   )r{  r  r   r   r   r8   
meta_histc  s.   

r}  c                    sd   t   |dd}t  dkptdd   dd  D  fdd  |jt	 d	S )
Nr   r(  r   c                 s   r  r  r7   )rA   r   r7   r7   r8   rc     r  z,meta_upsample_bimode2d_aa.<locals>.<genexpr>r   c                      rG  r.  r   r7   rM  r7   r8   rX     r  z+meta_upsample_bimode2d_aa.<locals>.<lambda>r   )
r'  r   rL   r[   r   r  r   r  rG   r!   )r   r;  rt  r0  r1  r+  r7   rM  r8   meta_upsample_bimode2d_aa  s   

(

r~  c                    st   t ||dd tjdkfdd tdD ]tj   k fdd q|jt	dS )Nr   r(  r`  c                      r|   r3  r  r7   rP  r7   r8   rX   
  r   z4meta_upsample_bimode2d_aa_backward.<locals>.<lambda>c                
      r4  )NzD
Expected grad_output to have the same shape as output; output.size(r5  z
but got grad_output_size(r   r7   r6  r7   r8   rX     s    r   )
r'  rL   r[   r   r   r   r   r  rG   r!   )rV  r;  r$  rt  r0  r1  r7   r6  r8   "meta_upsample_bimode2d_aa_backward  s   	

r  c                 C   s\   t | dkdd  t | dkdd  t |jjdd  t |jjdd  d S )Nr   c                   S   r]   )Nz%found_inf must be a 1-element tensor.r7   r7   r7   r7   r8   rX     r_   z<_amp_foreach_non_finite_check_and_unscale_.<locals>.<lambda>c                   S   r]   )Nz%inv_scale must be a 1-element tensor.r7   r7   r7   r7   r8   rX     r_   c                   S   r]   )Nz!found_inf must be a float tensor.r7   r7   r7   r7   r8   rX   #  r_   c                   S   r]   )Nz!inv_scale must be a float tensor.r7   r7   r7   r7   r8   rX   '  r_   )rL   r[   r   rS   r   )r   r~  Z	inv_scaler7   r7   r8   *_amp_foreach_non_finite_check_and_unscale_  s   r  c                 C   s   t |  }| |S r2   )r   r   r   )r   nanZposinfZneginfrN  r7   r7   r8   
nan_to_num,  s   
r  c                 C   s   | j tjtjtjtjhvsJ d| j  d| j}t||}t||}||kr)| S t| 	 }t| 
 }|| || ||< ||< || || ||< ||< | || | S )Nz>torch.transpose_: in-place transposition is not supported for z layout)rp   rL   r"  Z
sparse_cscr#  Z
sparse_bscr   r   r   r   r   r   )r   Zdim0r  ndimsr   r   r7   r7   r8   r  3  s&   	

r  c                 C   sz   | j }| jr"|  }|  }|dkr|dks!J d| d| dn|  dks0J d| dt| d|dk r:dS dS )	Nr   r   zEt_ expects a tensor with <= 2 sparse and 0 dense dimensions, but got z sparse and z dense dimensionsz6t_ expects a tensor with <= 2 dimensions, but self is r  r   )r   r{  r|  r}  rt   r  )r   r  r|  r}  r7   r7   r8   t_P  s   
r  )rv  r   sidesorterc                   s   t tjdkpjd d  jd d k fdd t d u p)jjkfdd t |dkp9| d |rAt jnt j}t t jrSt j |t j	dS t j
d	|jd
S )Nr   r   c                      s   dt j dt  j S )Nztorch.searchsorted(): boundaries tensor should be 1 dimension or the first N-1 dimensions of boundaries tensor and input value tensor must match, but we got boundaries tensor z and input value tensor r   r   r7   )r   sorted_sequencer7   r8   rX   s  s
   z#meta_searchsorted.<locals>.<lambda>c                      s,   dt  j dd urt j S g  S )Nz[torch.searchsorted(): boundary and sorter must have the same size, but got boundary tensor z and got sorter tensor r  r7   )r  r  r7   r8   rX   ~  s   r   zetorch.searchsorted(): side and right can't be set to opposites, got side of left while right was Truerw  r7   r^  )rL   r[   r   r   r)  r   r`   r   r   r   ry   rq   )r  r   rv  r   r  r  rS   r7   )r   r  r  r8   meta_searchsortedc  s&   
r  c                    s(   t  t jt jt jfv fdd d S )Nc                      r  )Nz/Unsupported input type encountered for isin(): r7   r7   r}   r7   r8   rX     r  z3_check_for_unsupported_isin_dtype.<locals>.<lambda>)rL   r[   rO  Z
complex128Z	complex64r}   r7   r}   r8   !_check_for_unsupported_isin_dtype  s   
r  c                 C   s   |  || df}|S )Nr   rp  )rV  r   num_weightsr  r  rq  r7   r7   r8   meta_embedding_dense_backward  s   r  c                 C   s:   |	rt | ||||||||
|
S t| ||||||||
|
S r2   )r.   Z_embedding_bag_sparse_backward!meta_embedding_bag_dense_backward)r  r   rQ  r  r  maximum_indicesr  r  r  r  r  r  r7   r7   r8   meta_embedding_bag_backward  s2   r  c
                    sX   t  jt jt jt jt jfv  fdd |tkr t |d u  | 	df}
|
S )Nc                      r|   )Nz$Unsupported input type encountered: r}   r7   r  r7   r8   rX     r   z3meta_embedding_bag_dense_backward.<locals>.<lambda>r   )
rL   r[   rS   r>  r?  r=  Zfloat64r  r   r   )r  r   r  r  r  r  r  r  r  r  Zindex_grad_weightr7   r  r8   r    s   
r  c           
      C   s~   |  d}t|tkd t|  dk t| dk | d}t| dk t| d|k | |f}	|	S )Nr   zHembedding_bag_backward: per_sample_weights only supported for mode='sum'r   r   )r   rL   r[   r  rt   r   )
r  r"  r   rQ  r  r  r  Zembedding_featuresr  r  r7   r7   r8   .meta_embedding_bag_per_sample_weights_backward  s   


r  )assume_uniqueinvertc                C   sx   t t| tpt|tdd  t| tst j| |jd} t|ts*t j|| jd}t| j t|j t j| t j	dS )Nc                   S   r]   )Nz<At least one of elements and test_elements must be a Tensor.r7   r7   r7   r7   r8   rX     r_   zmeta_isin.<locals>.<lambda>r{  r}   )
rL   r[   r`   r   r  rq   r  rS   r   rO  )elementsZtest_elementsr  r  r7   r7   r8   	meta_isin  s   



r  r  c                 C   s4   t | dkdd  t|tjd\}}t j||dS )Nr   c                   S   r]   )Nz,polygamma(n, x) does not support negative n.r7   r7   r7   r7   r8   rX     r_   z meta_polygamma.<locals>.<lambda>rm  r}   )rL   r[   r   r   rn  r   )r  r   rJ   rD   r7   r7   r8   meta_polygamma  s   
r  c                 C   s   t d)Nz.Tensor.item() cannot be called on meta tensors)r  r   r7   r7   r8   meta_local_scalar_dense  s   r  c                 C   r  r2   r  r   r7   r7   r8   silu$  r`  r  c                 C   s    t | tjd\}}tj| |dS rl  )r   r   rn  rL   r   )r   rJ   rD   r7   r7   r8   sigmoid*  s
   
r  c                 C   sF  |   dk}|  dk}|r8|r|d| d|dg}qot|d|dkd | d|dg}n7|rSt|d| dkd | d|dg}nt| d|dkd | d| d|dg}|ps| j}d|j }|d | d | | }||kr|d | |dg}	n|dg}	tj||	|| jd}
|
S )	Nr   r   r   z matrix batch sizes have to matchr   zbatched dimension has to matchr<  r^  )rt   r   rL   r[   rS   itemsizer  rq   )r,  r.  offsr&  Z
mat1_is_2dZ
mat2_is_2drY  	alignmentZsize_paddedr:  r   r7   r7   r8    _create_grouped_mm_output_tensor4  s0   

r  mat_amat_br  c	                    sv  t |d u |d u kdd  |d uo|d u}	|	r.t  jt jko%jt jk fdd nt  jt jko;jt jk fdd t   dv oP dv  fdd   dk}
 dk}|	rdd	 }d
d }t |  fdd t |fdd dd }|d  |d |d ur|d urt |jt jko|jt jkdd  d dd}d ur|
r|rjd nd}|d| d| |d|d| t |d u dd  |
s|rt d u fdd d urt  dkfdd t jt jkfdd n
t d u dd  t |d u dd  t |d u p/|t jkdd  t	 |S )!Nc                   S   r]   )Nz,Either both scale factors are given, or noner7   r7   r7   r7   r8   rX   d  r_   z)_meta_grouped_mm_common.<locals>.<lambda>c                      r  )Nz5Expected inputs of E4M3 FP8 type but got mat_a.dtype= and mat_b.dtype=rY  r}   r7   r  r  r7   r8   rX   o  rF   c                      r  )Nz1Expected inputs of BF16 type but got mat_a.dtype=r  rY  r}   r7   r  r7   r8   rX   t  rF   )r   r0   c                      s   d    d   S )Nz3Multiplicands must be 2D or 3D but got mat_a.dim()=z and mat_b.dim()=r   r7   r  r7   r8   rX   y  r/  r   c                 S   s    |   }|d dko|d dkS Nr  r   r   r  mat
mat_strider7   r7   r8   r       z-_meta_grouped_mm_common.<locals>.is_row_majorc                 S   s    |   }|d dko|d dkS r  r  r  r7   r7   r8   r    r  z-_meta_grouped_mm_common.<locals>.is_col_majorc                         d   dd   S )NzNExpected mat_a tensor to be row major in the last two dimensions, got strides r  r  r7   )r  r7   r8   rX     rF   c                      r  )NzQExpected mat_b tensor to be column major in the last two dimensions, got strides r  r  r7   )r  r7   r8   rX     rF   c                    s     d  d  }  d  dkr:  tdj d  kr:t  | dk fdd d S   dkrd d  tdj  krdt d  | dk fdd d S tdfdd d S )	Nr   r<  r   c                      s   d d  d   dS )Nr0   stride along % dim to be multiple of 16 bytes, got rY  r7   r7   end_dimmat_namer  r7   r8   rX     r  zF_meta_grouped_mm_common.<locals>.check_valid_strides.<locals>.<lambda>c                      s$   d d d  d d   dS )Nr0  r  r   r  rY  r7   r7   r  r7   r8   rX        $ Fc                      s   d d j  dS )NzInvalid strides/sizes, got z for strides and z for sizes.r   r7   r  r7   r8   rX     r   )rt   Zelement_sizer   r   r   rL   r[   )r  r  r  r7   )r  r  r  r  r8   check_valid_strides  s*   
z4_meta_grouped_mm_common.<locals>.check_valid_stridesr  r  c                   S   r]   )NzBoth scale_a and scale_b must be float (fp32) tensors, but got scale_a.dtype={scale_a.dtype} and scale_b.dtype={scale_b.dtype}.r7   r7   r7   r7   r8   rX     r_   r   c                    s     dkr;t  dkfdd t fdd tjd  j  k fdd d S t  dkfdd tddkfd	d tjd  jd k fd
d tjd  jd  k fdd d S )Nr   r   c                         d d    dS )Nr0  z to be 1D tensor, but got 	D tensor.r   r7   r  
scale_namer7   r8   rX     rF   z>_meta_grouped_mm_common.<locals>.check_scale.<locals>.<lambda>c                      r  )Nr0  z to be contiguous.r7   r7   r  r7   r8   rX     r   r   c                      s(   d d j    dj d  dS )Nr0  z	 to have r  r   z
 elements.r   r7   r  r  scale_multiplierr  
scaled_dimr7   r8   rX        ( c                      r  )Nr0  z to be 2D tensor, but got r  r   r7   r  r7   r8   rX     rF   c                      r  )Nr0  z( to be contiguous in the last dimension.r7   r7   r  r7   r8   rX     r   c                      s$   d d j d  dj d  dS )Nr0  z batch dimension to be r   , got rY  r   r7   )r  r  r  r7   r8   rX     r  c                      s(   d d j d   dj d  dS )Nr0  z non-batch dimension to be r   r  rY  r   r7   )r  r  r  r  r7   r8   rX     r  )rt   rL   r[   r   r   r   )r  r  r  r  r  r7   r  r8   check_scale  s:   



z,_meta_grouped_mm_common.<locals>.check_scaler   r	  r
  c                   S   r]   )Nz:Scale result tensor provided, but it is not supported yet.r7   r7   r7   r7   r8   rX     r_   c                      s   d    d   dS )Nz/Offsets tensor not provided, but is needed for zD/zD multiplicand layouts.r   r7   r  r7   r8   rX     s    c                      r  )Nz.Offsets tensor must be 1D, but got offs.dim()=rY  r   r7   r  r7   r8   rX     rY   c                      r  )Nz7Offsets tensor must be integer (int32) tensor, but got rY  r}   r7   r  r7   r8   rX     r  c                   S   r]   )NzJOffsets tensor provided, but is not needed for 3D/3D multiplicand layouts.r7   r7   r7   r7   r8   rX     r_   c                   S   r]   )Nz2Bias tensor provided, but it is not supported yet.r7   r7   r7   r7   r8   rX     r_   c                   S   r]   )Nz4If output dtype provided, it must be torch.bfloat16.r7   r7   r7   r7   r8   rX     r_   r  )
rL   r[   rS   r@  r?  rt   r=  r   r)  r  )r  r  r	  r
  r  r$  r  r&  r  ZscaledZmat_a_is_2dZmat_b_is_2dr  r  r  r  r  r7   )r  r  r  r8   _meta_grouped_mm_commonW  s   




!





r  c              
   C   s   t | |d d ||d |dS )N)r	  r
  r  r$  r  r&  r  )r  r  r  r$  r&  r7   r7   r8   
grouped_mm  s   	r  c	           	      C   s   t | ||||||||d	S )N)r	  r
  r  r$  r  r&  r  r  )	r  r  r	  r
  r  r$  r  r&  r  r7   r7   r8   meta_scaled_grouped_mm  s   r  rB   half_to_floatc                 C   sL   |r
| j tjks
J tj| tjjd\}}|s|n|}tj| |tjd}|S )Nrm  rw  )	rS   rL   rM   rG   r   r   rH   r   r   )rB   rt   r  Zcomputation_dtyperD   r  r7   r7   r8   softmax-  s   
r  c              	      s   t td dkfdd | jttd }| t |kfdd td  }t|D ]1 t d d       d   }t |dk fdd || q:t j|| j| j	| j
t| dS )	Nr   r   c                      r  )Nz1Length of pad must be even but instead it equals r-  r7   rD  r7   r8   rX   A  r  z'_constant_pad_nd_meta.<locals>.<lambda>c                      s   dt  d  dS )Nz`Length of pad should be no more than twice the number of dimensions of the input. Pad length is z while the input has z dimensions.r-  r7   )l_inprE  r7   r8   rX   K  s
    r   c                	      s6   d    d  dd   d   d	S )NzThe input size z, plus negative padding r   r   zG resulted in a negative output size, which is invalid. Check dimension z of your input.r7   r7   )r   r   l_diffrE  pad_idxr7   r8   rX   V  s    
)rS   rq   rs   r   )rL   r[   r   r   r   r   r   ry   rS   rq   rs   r!   )r   rE  r/  Zl_padr   Znew_dimr7   )r   r   r  r  rE  r  r8   _constant_pad_nd_meta;  s8   
 r  r  r  r  c           	      C   sx   |   dks
J d| j}|j}|jdkr|d f}n|jdkr)|d |d f}n	g ||d R }| j}| j||dS )Nr   z'weight' must be 2-Dr   r   r}   )rt   r   r   rS   r   )	r"  r   r  r  r  Zweight_shapeZindices_shaper   r&  r7   r7   r8   	embeddinge  s   	

r  max_lengthspadding_valuec                 C   s\   t |dksJ t |dksJ |d jd d }|d }||g| jdd  R }| |S r  )r   r   r   )r   rQ  r  r  r  r  rC  r7   r7   r8   $meta__jagged_to_padded_dense_forward}  s   
r  c                 C      t | t dd }|S )Nc                 S   r  r	  rK   r   rn  r   r7   r7   r8   _f  s   z)_create_unary_float_meta_func.<locals>._fr=   r'   funcr  r7   r7   r8   _create_unary_float_meta_func     r  c                 C   r  )Nc                 S   r   r	  r  )rB   r>  r7   r7   r8   r    r"  z*_create_binary_float_meta_func.<locals>._fr  r  r7   r7   r8   _create_binary_float_meta_func  r  r  c                    s<   t   fdd} j d}||_ttt||}|S )Nc                    s(    | g|R i |}t | j|j | S r2   r.  )r   rI   r{  r   r5   r7   r8   _fn  s   z#_register_inplace_meta.<locals>._fnrJ   )r   rj   r=   getattrr.   )r6   r  Zinplace_namer7   r5   r8   _register_inplace_meta  s   r  c                    sr   t j jk fdd  g}ttr1jdkr,t jjkfdd | t|dtj	iS )Nc                      r  )NrB  z for `end`, but got dtype r}   r7   )rk   rl   r7   r8   rX     r   zlerp.<locals>.<lambda>r   c                      r+  )NrB  z for `weight`, but got dtype r}   r7   )rl   r"  r7   r8   rX     r   r>   )
rL   r[   rS   r`   r"   r   r   rK   r   rH   )rl   rk   r"  rI   r7   )rk   rl   r"  r8   lerp  s"   




r  )r/  c                C   s   t | ||tjdS r	  r  r   Ztensor1Ztensor2r/  r7   r7   r8   addcmul  s   
r  c                C   s8   t t|jot|j dd  t| ||tjdS )Nc                   S   r]   )N)zFInteger division with addcdiv is no longer supported, and in a future zErelease addcdiv will perform a true division of tensor1 and tensor2. z4The historic addcdiv behavior can be implemented as zA(input + value * torch.trunc(tensor1 / tensor2)).to(input.dtype) zfor integer inputs and as z6(input + value * tensor1 / tensor2) for float inputs. z?The future addcdiv behavior is just the latter implementation: z4(input + value * tensor1 / tensor2), for all dtypes.r7   r7   r7   r7   r8   rX     r_   zaddcdiv.<locals>.<lambda>r
  )rL   r[   rG   r  rS   rK   r   rH   r  r7   r7   r8   addcdiv  s   

r  c                  C   s4  i } dD ]}t | }|D ]}|| vr|| | |< qq|  D ]y\}}t|tjjr*qt|ts1J |tjj	j
| tj| drR|t d v rQt| dq|jrVq| dv r]qd| v rjt|| qd| v rwt|| qd| v rt|| qd	| v rt|| qt|| qd S )
N)rn   Zpost_autogradZpre_autogradZCompositeImplicitAutogradrn   z is a CompositeImplicitAutograd op, we shouldn't register meta function for it. Instead, we should let the decomposition run and write meta kernels for the base operators.>   zaten::empty_stridedzaten::_to_copyzaten::rot90zaten::copy_zaten::constant_pad_ndzaten::clonezaten::as_strided_scatterzmkldnn::zmkl::zonednn::zquantized::)r   itemsr`   rL   Z_opsZHigherOrderOperatorr   Zpy_impl_CZDispatchKeyr/   Z%_dispatch_has_kernel_for_dispatch_keyr  r  Zis_view2_meta_lib_dont_use_me_use_register_meta_for_mkldnnimpl/_meta_lib_dont_use_me_use_register_meta_for_mkl2_meta_lib_dont_use_me_use_register_meta_for_onednn5_meta_lib_dont_use_me_use_register_meta_for_quantized'_meta_lib_dont_use_me_use_register_meta)Zactivate_meta_tableri   registryZopoZop_overloadr6   r7   r7   r8   activate_meta  sP   r  r   )Trn  r2   )NNNFr   r   r   r  r  )r  )r  T)FF)TT)r  )FTN)TFF)TF)r   )r  N)r1   r  )r7   r   r  F)r7   r   FTN)Fr   FNFr   )NF)r   F)r#  r$  FN)NNNNN)r   NNr   )NNF)r  FFN)Nr  FFN)r  FNN)r  FN)FN)FNNNN)NNNF)Nr   FNN)NNNN)r   TT)NNr   N)ry  r   r   )r   )NNNNF)r   FF)r  (  r  r   collections.abcr   enumr   	functoolsr   r   typingr   r   r   r	   Ztyping_extensionsr
   rL   Ztorch._prims_commonr  rG   r   r   r   Ztorch._decompr   r   r   r   Z
torch._opsr   Ztorch._primsr   r   r   r   r   r   r   r   r   r   r   r   r   r    r!   r"   Ztorch._prims_common.wrappersr#   r$   r%   r&   r'   rk  r(   r)   Ztorch.fx.experimentalr*   r  Ztorch.utilsr+   r;   r,   r-   opsr.   ZlibraryLibraryr  r   r  r  r  r=   rK   rT   r\   ZlinspaceZlogspacer!  r{   Ztaker  r   r   r   r   r   r   r7  r   r   ZcummaxZcumminr   r   r   r   r   rO  r   Z_fft_c2cr   r   r   Z_fft_r2cr  ZrandpermZgenerator_outr  r   r  randintr  r  Zlow_outr  Zrandr  Z_fft_c2rr  rp  r  r   Z
unsqueeze_r!  Z_sparse_semi_structured_linearr  rS   r+  Z_sparse_semi_structured_mmr0  Z_sparse_semi_structured_addmmr3  Z_cslt_sparse_mmrD  Zindex_reducerJ  Zindex_reduce_rL  Zindex_selectrO  Zsegment_reducer[  r   Z	unary_outr_  rt   rh  r   rj  rk  rq  ro  rr  Z_assert_asyncru  r   rx  Z_printry  Z_make_dep_tokenr|  r  Z_functional_sym_constrain_ranger  r  Z(_functional_sym_constrain_range_for_sizer  Z_functional_assert_asyncr  r   r  r   r  r  r  r  Z_linalg_eighr  r  Z_linalg_eigvalsZlinalg_eigvalsr  Z
linalg_eigr  r  r  r  r  r  r  r  Zlinalg_inv_exr  Zlinalg_ldl_factor_exrZ   r  Zlinalg_ldl_solver  Z	linalg_lur  Zlinalg_lu_factor_exr  Zlinalg_lu_solver  Z	lu_unpackr  r  Z	linalg_qrr  r   r  Z_linalg_svdr  r  r  r  r  Zlinalg_solve_triangularr  r   r$  Z_linalg_detr%  r,  r1  rA  Zreflection_pad1drF  Zreplication_pad1drP  rX  Zreflection_pad1d_backwardr]  Zreplication_pad1d_backwardr_  rk  Zreflection_pad2drl  Zreplication_pad2drm  Zreflection_pad2d_backwardrY  Zreplication_pad2d_backwardrp  rx  Zreflection_pad3dry  Zreplication_pad3drz  Zreflection_pad3d_backwardZreplication_pad3d_backwardr|  Z_pdist_forwardrO   r~  Z_pdist_backwardr  Zbaddbmmr  Z	bernoullir  Z
bernoulli_r  r}  r  Zpoissonr  Z_fused_moving_avg_obs_fq_helperr  mmr  re  r   r  r  Zmiopen_batch_normr  Zconvolutionr  r  Z_has_mkldnnr  r  Z_convolution_pointwiser  Z_linear_pointwiser  Zhas_mklr  r  Z_mkl_linearr  r  r  Zqconv2d_pointwiseZqconv_pointwiser  binaryr  Zqlinear_pointwiser  r  Zbinary_tensorr  Zlinear_dynamic_fp16Zlinear_relu_dynamic_fp16r  r  r  Z
max_pool2dr  Zint4mm_packed_weight_cpur  r  Z
avg_pool2dr   r#  Zavg_pool2d_backwardr%  Z
avg_pool3dr8  Zavg_pool3d_backwardr:  Z_adaptive_avg_pool2dr<  Z_adaptive_avg_pool3dr=  Z_adaptive_avg_pool2d_backwardrC  Z_adaptive_avg_pool3d_backwardrF  rD  Zadaptive_max_pool2drO  rQ  rS  Zadaptive_max_pool3drT  rU  rV  Zrepeat_interleaverX  ra   r[  r]  r_  r   Z_unsafe_indexro  Zconvolution_backwardru  Zaddbmmrz  Zrandint_liker|  Z_fused_adam_Z_fused_adamw_r  Z_fused_adamr  Z_int_mmr  Z_convert_weight_to_int4packr  Z#_convert_weight_to_int4pack_for_cpur  Z_weight_int4pack_mmr  Z_weight_int4pack_mm_for_cpur  r  r  r  Z_dyn_quant_pack_4bit_weightr  Z_dyn_quant_matmul_4bitr  Z_weight_int8pack_mmr  Z_cdist_forwardr  Z_cdist_backwardr  Z_embedding_bagr  Z_embedding_bag_forward_onlyr  r  Znansumr  ZmedianZ	nanmedianr  Z
dim_valuesr  r   r  Zlogical_not_r  repeatr  Zzero_r  Zmul_ZScalarZdiv_Zlogical_and_Zlogical_or_Zlogical_xor_r   Zadd_Zsub_r  roundZdecimalsr  r  
__rshift__r  
__lshift__r  zeror  r~  r  fillr  Zrelu_r  Z	_add_relur!  Zrrelu_with_noiser(  Zrrelu_with_noise_functionalr)  Zrrelu_with_noise_r*  Z	index_putZ_unsafe_index_putr-  Zmasked_fill_r0  Z_masked_scaler1  Zmasked_scatter_r3  Zmasked_scatterr4  Zmasked_scatter_backwardr5  Z
index_put_r6  aliasr8  r:  Zbmmr<  r=  r@  rC  r  r  r-  rT  r9  r  Z max_pool2d_with_indices_backwardrV  Zmax_pool2d_with_indicesrW  Zfractional_max_pool2dr^  Zmax_pool3d_with_indicesre  Z max_pool3d_with_indices_backwardrf  rj  rk  rp  Zgrid_sampler_2d_backwardrv  rx  ry  rz  r  Zonesr  Zzerosr  rj  r  Zselect_scatterr  Zslice_scatterr  r   r  r  Zgatherr  r  r  r  r  r  Zscatter_addr  Zscatter_add_r  r  r  r/  Zvalue_reducer  Zscatter_r  Z#_scaled_dot_product_flash_attentionr  r  Z#_scaled_dot_product_cudnn_attentionr  Z0_scaled_dot_product_fused_attention_overrideabler  Z,_scaled_dot_product_flash_attention_backwardr  Z+_scaled_dot_product_flash_attention_for_cpur  Z4_scaled_dot_product_flash_attention_for_cpu_backwardr  Z'_scaled_dot_product_efficient_attentionr  Z0_scaled_dot_product_efficient_attention_backwardr  Z,_scaled_dot_product_cudnn_attention_backwardr  Z_flash_attention_forwardr  Z_flash_attention_backwardr  Z_efficient_attention_forwardr  Z_efficient_attention_backwardZSymIntr  Z
_scaled_mmr  Zscatter_reducetwoZtwo_outr  Zscatter_reduce_r  Zmultinomialr  r"  r'  r,  Z_upsample_nearest_exact1dr2  Z_upsample_nearest_exact2dr7  Z"_upsample_nearest_exact2d_backwardr8  Z_upsample_nearest_exact3dr   r9  Zvalues_stabler;  rJ  Z_thnn_fused_lstm_cellrO  rZ  r]  r_  r`  ra  Zargminrb  rc  Ztopkre  Z_segment_reduce_backwardrf  Zkthvaluerh  r   ro  rm  rp  rr  Zpixel_shufflert  ru  Z	bucketizeZ
Tensor_outrx  Zhistcr}  Z_upsample_bilinear2d_aaZ_upsample_bicubic2d_aar~  Z _upsample_bilinear2d_aa_backwardr  r  r  r  r  Zsearchsortedr  r  Zembedding_dense_backwardr  Z_embedding_bag_backwardr  Z_embedding_bag_dense_backwardr  Z*_embedding_bag_per_sample_weights_backwardr  isinr  Z	polygammar  Z_local_scalar_denser  r  r  r  r  Z_grouped_mmr  Z_scaled_grouped_mmr  Z_softmaxr  Zconstant_pad_ndr  r  Z_jagged_to_padded_dense_forwardr  r  r  Zspecial_airy_aiZspecial_bessel_y0Zspecial_bessel_y1Zspecial_modified_bessel_i0Zspecial_modified_bessel_i1Zspecial_modified_bessel_k0Zspecial_modified_bessel_k1Z!special_scaled_modified_bessel_k0Z!special_scaled_modified_bessel_k1Zspecial_chebyshev_polynomial_tZspecial_chebyshev_polynomial_uZspecial_chebyshev_polynomial_vZspecial_chebyshev_polynomial_wZ&special_shifted_chebyshev_polynomial_tZ&special_shifted_chebyshev_polynomial_uZ&special_shifted_chebyshev_polynomial_vZ&special_shifted_chebyshev_polynomial_wZspecial_hermite_polynomial_hZspecial_hermite_polynomial_heZspecial_laguerre_polynomial_lZspecial_legendre_polynomial_pr  r  r  r  Zlerp_Zaddcmul_Zaddcdiv_Ztorch._refs.nn.functionalZtorch._refs.specialr  r7   r7   r7   r8   <module>   s  <(
	8I

D
		6

;
"
'
	
!"	
0



#
	

	











	




'



"

2
*
*
"7
(*&
%
	
:

/Z&5?'&
.



l%	,$Q,
H
XN



.


*" 
$ m
#c	







-


)
T	
\>	
6L+&
T

ge( 

	, $1	






@	+*	
		
*,	
7	
	
K	
	
/	
7 #



'7'

"
0


*


"
	


(	 
)	
(

(




E