o
    h                 -   @   sl&  d dl Z d dlmZ d dlmZmZmZmZ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 d dl
mZmZmZmZmZmZm Z  d d	l!m"Z"m#Z#m$Z$m%Z% d d
l&m'Z'm(Z( d dl)m*Z*m+Z+ d dl,m-Z- e	j.j/Z/e	j01dddZ2dd Z3dd Z4dd Z5e3e/j6j7e/j6j8ge% dd Z9e3e/j:j7e/j:j8ge% ddddZ:e3e/j;e% dd Z;e3e/j<j7e/j<j8e/j=j7e/j=j8ge%ddd d! Z>e3e/j?j7e/j?j8ge% d"d# Z?d$d% Z@e3e/jAj7e/jAj8ge% d&d' ZBe3e/jCj7e/jCj8ge% d(d) ZDe3e/jEjFdd*d+d,ZGe3e/jEj7e	jHdddd-d.d/ZIe3e/jJj7e	jHdddd-d0d1ZKe3e/jJjLe	jHdddd-d2d3ZMe3e/jNj7ddddd-d4d5ZOe3e/jPj7e/jPj8ge% d6d7 ZQe3e/jRj7dd9d:ZSd;d< ZTe3e/jUj7d=d> ZVe3e/jWj7d?d@dAedBeXdCedDe	jdEeYdFeZdGefdHdIZ[e3e/j\j7d?d@dAedBeXdCedDe	jdEeYdFeZdGefdJdKZ]e3e/j^j7dLdM Z_e3e/j^j8dNdO Z`e3e/jaj7e/jajbge% dPdQ Zce3e/jajdddRdSZee3e/jfj7e/jfjbge% dTdU Zge3e/jfjdddVdWZhe3e/jij7dXdY Zje3e/jij8dZd[ Zke3e/jlj7d\d] Zme3e/jljnd^d_ Zoe3e/jpj7dddddd`dadbZqe3e/jrj7ddcddZre3e/jsj7ddedfZte3e/juj7ddgdhZue3e/jvj7didj Zwe3e/jxjndkdl ZydAedmeYfdndoZzdAedpedqeYfdrdsZ{	?ddtedmeYdueZfdvdwZ|ddpedmeYdxeYfdydzZ}dped{ed|eZdmeYfd}d~Z~	ddeYdededeYfddZdeYfddZe3e/jj7e/jjge%dd		?ddpedeYdeZfddZdedGefddZe3e/je% dAedpedeZdGefddZe3e/je% ddAedpedeZdGefddZe3e/je% ddAedeZdGefddZe3e/je% ddAedeZdGefddZe3e/jj7ddpedeZdeZfddZe3e/jj7e/jj8ge% dededGefddZe3e/jj7ddpedeZfddZe3e/jj7e/jj8ge%dddd8d8ddAedeZdeZdGeeeef fddZe3e/jj7e/jj8ge% d8ddeded{edeZdGef
ddZe3e/jj7e/jj8ge%dddd?ddpedeZdGeeeef fddZe3e/jj7e/jj8ge%dddd?d8ddpedeZdeZdGeeeef fddZe3e/jj7e/jj8ge% d?d8ddeded{ed|eZdeZdGefddZe3e/je%ddd	?	?ddededeZdeZdGeeeef f
ddZdeYdGeeZeZf fddZe3e/jj7e/jj8ge%dd	ÐddpedeYdGeeef fddńZe3e/jj7e/jjge%dddddpedGeeeeef fddɄZe3e/jj7	8	?	ddpedeZdeZdeeY fdd΄ZdededGeeeX eeX f fdd҄ZdededqeeY dGeeef fddԄZdededGeZfddׄZe3e/jd?d8ddddd؜dped{ed|eZdeZdee dee dee dee dGeeeeef fddڄZe3e/jj7e/jj8gd?d8ddۜdped{edeZd|eZdeZdee dGefdd߄Ze3e/je%dd	?	8	8ddAedpedeZdeZdeZdGeeef fddZe3e/jj7dd Ze3e/je% 	?	8ddededed|eZdeZdGefddZdd Zdd Ze3e/je% dd Ze3e/je% dd Zdd Ze3e/je% dd Ze3e/je% dd Zdd Ze3e/je% dd Ze3e/je% dd Ze3e/jj7e/jje/jj7e/jjge% dd Zdd  Ze3e/je% dd Ze3e/je% dd Ze3e/jj7e/jje/jj7e/jjge% dd Ze3e/jǃe% ddAededGefd	d
Ze3e/jʃe% dedAedeȐdedGef
ddZe3e/jj7e/jj8ge% dddddZe3e/jj7e/jj8ge% dd*ddZe3e/jjȃdddZe3e/jj҃dddZe3e/jj7	8	8dddZՐdd Ze3e/jj7dd Ze3e/jj7gd d! Zڐd"d# Zېdd%d&Z	dd'e	jd(e	jd)eeeX eXf d*eeeX eXf d+eeeX eXf d,eZd-eXd.eeeeX eXf  fd/d0Zݐd1d2 Ze3e/jj7d'e	jd(e	jd3e	jd)eeX d*eeX d+eeX d,eZd.eeX d-eXfd4d5Ze	jjre	j01d6ddZe3e	j.jjj7d7d8 Ze3e	j.jjj7d9d: Ze	jjre	j01d;ddZe3e	j.jjd<d= Ze	j01d>ddZe3e	j.jjj7d?d@ Ze3e	j.jjj7dAdB ZdCdD Ze3e/jj7	E	F	8	?	ddGdHZdIdJ Ze3e/jj7dKdL Ze3e/je% 	E	F	8	?	ddMdNZe3e/je% dOdP Ze3e/jj7dQdR Ze3e/j j7dSdT Ze3e/jj7dUdV Ze3e/je% dWdX ZdYedxeYfdZd[Ze3e/je%ddd\d] Ze3e/j	e% d^d_ Z
e3e/je%ddd`da Ze3e/je% dbdc Ze3e/jjddddeZe3e/jj7e/jj8ge% dfdg Ze3e/jjdhdi Ze3e/jj7djdk Ze3e/jj7e/jj8gddldmeXdneXfdodpZe3e/jje/jjgdqdr Ze3e/jj7gdsdt Ze3e/jj7e/jj8ge% ddddudvZe3e/jj7e/j j7e/j!j7e/j"j7gdwdx Z#e3e/j$j7e/j%j7e/j&j7e/j'j7gdydz Z(d{d| Z)e3e/j*je/j+je/j,je/j-je/j.je/j/jgdd}d~Z0e3e/j1je/j2je/j3je/j4je/j5je/j6jgdddZ7e3e/j1j8e/j3j8e/j2j8e/j4j8gdddZ9e3e/j*j8e/j-j8e/j,j8e/j+j8gdddZ:e3e/j;j8e/j<j8gdddZ=e3e/j>j8gdddZ?e3e/j@j8e/jAj8gdddZBe3e/jCjDgdd ZEe3e/j;je/j<jgdd ZFe3e/jGgdddZHe3e/jIj7gdddddZJe3e/jKj7gdddddZLe3e/jMge% dd ZNe3e/jOj7dd ZPe3e/jQe% dd ZRe3e/jSj7	8	 	8		8	dddZTe3e/jUj7dd ZVdddZWe3e/jXj7e/jXj8ge% dddddZYe3e/jZj7e/j[j7gdd Z\e3e/jZjde/jZj]e/j[jde/j[j]e/j^j7e/j^j_ge%dddddZ`e3e/jaj7dd Zbe3e/jcj7dd Zde3e/jej7dd Zfe3e/jgj8e/jhj8e/jgje/jhje/jij7e/jjj7e/jkj7gdd Zle3e/jmj8e/jnj8e/jmje/jnjgdddZoe3e/jpj7e/jpjqgdd Zrdd Zse3e/jtje/jtj8gdd Zue3e/jvje/jvj8gdd Zwe3e/jxj7dd Zye3e/jzje/jzj8gdd Z{e3e/j|je/j|j8gdd Z}e3e/j~j7dd Ze3e/jj7e/jj7gddÐdĄZe3e/jj8dŐdƄ Ze3e/jj7ddǐdȄZe3e/jj7dɐdʄ Zddːd̄Ze3e/jj7d͐d΄ ZdϐdЄ Zdѐd҄ ZdӐdԄ ZdՐdք Z	8ddedeXdeXdeXdeXdeXdeXdeXdeXdeXdeXdeXdeXdeXdeXdeXdeXdeXdeXdeXdeYdeZf,ddZdd ZdedYedeXdeXdeXdeXdeXdeXdeXdeXdeXdeXdeXdeXdeXdeXdeXdeXdeYf&ddZdd Ze3e/jj7dd Ze3e/jj7	8dddZe3e/je% dd Zdd Ze3e/je% dd Ze3e/je%dd	E	F		8dddZe3e/je% d d ZdedefddZG dd deZdededeXfdd	Ze3e/jj7d
d Ze3e/je% dd Ze3e/je%dddd Ze3e/jj7gdd Ze3e/jj7					dddZe3e/jjXdd Ze3e/jj7dd Ze3e/jj7dddZddBeXdeXdeZfddZd d! Zd"d# Ze3e/jj7dd$d%Zdd&d'Zdd(d)Zd*d+ Zdd,d-Zdd.d/Ze3e/jj7d0d1 Ze3e/jd2d3 Ze3e/jje/jje/jje/jjge% dd4d5Ze3e/jĐje/jĐje/jĐje/jĐjgdd6d7Ze3e/jg	8	8	8	dd9ed:ed;ed<eȐd=eZd>eZd?ee fd@dAZe3e/jg	ddBed9ed:ed;ededCedDedEedFeXdGeXd<eȐd=eZdHedIed?ee fdJdKZe3e/jg	8	8	dd9ed:ed;edLee dMeZd=eZd?ee fdNdOZe3e/jg	8	ddBed9ed:ed;edLee dedCedHedIed<eȐdPeeZ d=eZd?ee fdQdRZe3e/jΐje/jΐjge% ddSdTZe3e/jҐjσddUdVZe3e/jj7e/jj8ge% ddd*dWdXZՐdYdZ Z֐d[d\ Ze3e/jj7dd]d^Ze3e/jj7dd_d`Ze3e/jj7		ddYedaeeeXe	jېjf  dbeeeXe	jېjf  dcee ddee f
dedfZe3e/jj7ddgdhZe3e/jj7e/jސje/jސj_e/jސjgddidjZe3e/jj߃dd8dkdldmZdndo Ze3e/jj7	ddpdqZe3e/jj7drds Ze3e/jj7dtdu Zdvdw Zdxdy Ze3e/jj7e/jj7gddzd{Ze3e/jj7dd|d}Ze3e/jj7dd~dZe	jZdd Ze3e/jj7dd Ze3e/jj7dd Ze3e/jj7dd Ze3e/jje/jjge% d8d8dddZe3e/jj7	dddZe3e/jj7dd Ze3e/jj7e/jj8ge% dddZe3e	j.j/jdd Ze3e	j.j/j dd Z e3e/je% d8d8dddddZe3e/je% deXdAedGefddZdd Zdd Zee/j ee/j ee/j	 ee/j
 ee/j ee/j ee/j ee/j ee/j ee/jj7e/jj8e/jjg ee/jj7e/jj8e/jjg ee/jj7e/jj8e/jjg ee/jj7e/jj8e/jjg ee/jj7e/jj8e/jjg d dl&Z	d dlZ	d dlZ	dd Ze  dS (      N)Enum)ListOptionalSequenceTupleUnion)SymBoolSymFloatTensor)_add_op_to_registry_convert_out_paramsglobal_decomposition_table
meta_table)
OpOverload)_elementwise_meta$ELEMENTWISE_PRIM_TYPE_PROMOTION_KIND)corresponding_complex_dtypecorresponding_real_dtypeelementwise_dtypesELEMENTWISE_TYPE_PROMOTION_KINDIntLikemake_contiguous_strides_for
TensorLike)_maybe_resize_out_resize_output_check_safe_copy_outout_wrapper)_broadcast_shapes_maybe_broadcast)_constrain_range_for_sizeconstrain_range)tree_mapatenIMPLMetac                    s    fdd}|S )Nc                    s"   t    fdd}t|  S )Nc                    s   t t|   d S N)r   r   opfn O/var/www/html/ai/venv/lib/python3.10/site-packages/torch/_meta_registrations.pyregister1   s   z0register_meta.<locals>.wrapper.<locals>.register)r   r!   )r)   r,   r&   r(   r+   wrapper.   s   
zregister_meta.<locals>.wrapperr*   )r'   r-   r*   r&   r+   register_meta-   s   	r.   c                 C   s(   t jt jt jt jt jt ji}|| | S r%   )torch	complex32halfcfloatfloatcdoubledoubleget)dtypefrom_complexr*   r*   r+   toRealValueType:   s
   r9   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 r*   r*   broadcasted_shape
self_shaper*   r+   <lambda>G       z)check_inplace_broadcast.<locals>.<lambda>)tupler   r/   _check)r=   
args_shaper*   r;   r+   check_inplace_broadcastC   s
   rC   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 r7   r*   indexr*   r+   r>   Q       zmeta_take.<locals>.<lambda>r   c                   S      dS )Nz*take(): tried to take from an empty tensorr*   r*   r*   r*   r+   r>   V       )r/   rA   r7   long_check_indexnumel	new_emptyshape)selfrG   r*   rF   r+   	meta_takeK   s   

rQ   dimc                   sh   j }j }t||kdd  t dko dk fdd tjj}|S )Nc                   S   rI   )Nz=linalg.cross: inputs must have the same number of dimensions.r*   r*   r*   r*   r+   r>   b   rJ   zlinalg_cross.<locals>.<lambda>   c                      s"   d  d   d   S )Nzlinalg.cross: inputs dimension z must have length 3. Got  and sizer*   rT   otherrP   r*   r+   r>   f   s
   )ndimr/   rA   rX   r   rO   rN   )rP   rZ   rT   x_dy_d	out_shaper*   rY   r+   linalg_cross[   s   
r_   c                 C   s   t | d t| d t| S )Nzlinalg.matrix_exp
matrix_exp)squareCheckInputscheckFloatingOrComplexr/   
empty_likerP   r*   r*   r+   linalg_matrix_expo   s   


re   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devicer7   r   )	r/   emptyrO   ri   r7   int64rM   r[   maybe_wrap_dim)rP   rT   rf   rg   r*   r*   r+   	cummaxminw   s
   rm   c                 C   s   t || j t|  S r%   )rl   r[   r/   rc   
contiguous)rP   rT   r*   r*   r+   logcumsumexp   s   ro   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|D ]}|||  ||d	 < q| 	|} 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 Fr*   .0_r*   r*   r+   
<listcomp>   r?   z_exec_fft.<locals>.<listcomp>Tc                        |  S r%   r*   xself_stridesr*   r+   r>          z_exec_fft.<locals>.<lambda>keyreverserR   r      c                 S   rp   r   r*   rr   r*   r*   r+   ru      r?   )r[   lenlistrangeappendstridesortpermuterO   reshaperX   
as_stridedstorage_offset)outrP   	out_sizesrT   forwardr[   signal_ndim
batch_dimsdim_permuteis_transformed_dimdleftright	batch_endtmpinputbatched_sizes
batch_sizebatched_out_sizesiout_stridesbatch_numelr*   ry   r+   	_exec_fft   sL   





r   c                    sb   | j jsJ | j}| |}|s|S |d d  }|   |j fdddd t|| |||}|S )Nc                    rv   r%   r*   rw   ry   r*   r+   r>      r{   zmeta_fft_c2c.<locals>.<lambda>Tr|   )r7   
is_complexrO   rN   r   r   r   )rP   rT   normalizationr   r   outputsorted_dimsr*   ry   r+   meta_fft_c2c   s   
r   c                 C   sR   | j jsJ t|  }|r|d }|| d d }|||< | j|t| j dS )NrR      r   rE   )r7   is_floating_pointr   rX   rN   utilsr   )rP   rT   r   onesidedoutput_sizeslast_dimlast_dim_halfsizer*   r*   r+   meta_fft_r2c   s   r   )	generatorc                C   s    |j dkr|d| ksJ |S Nr   r   )r[   rX   )nr   r   r*   r*   r+   meta_randperm   s   r   r7   layoutri   
pin_memoryc                C      t j| ||||dS Nr   r/   rj   )r   r7   r   ri   r   r*   r*   r+   meta_randperm_default      
r   c                C   s   t j|||||dS r   r   )highrX   r7   r   ri   r   r*   r*   r+   meta_randint   r   r   c                C   s   t j|||||dS r   r   )lowr   rX   r7   r   ri   r   r*   r*   r+   meta_randint_low   s   
r   c                C   r   r   r   )rX   r7   r   ri   r   r*   r*   r+   meta_rand_default     
r   c                 C   s8   | j jsJ t|  }|||d < | j|t| j dS )NrR   rE   )r7   r   r   rX   rN   r9   )rP   rT   r   lastdimr   r*   r*   r+   meta_fft_c2r  s   r   Fc                 C   s2   | | |}|  | krtj||   | S r%   )torX   r"   expand_copydefault)rP   srcnon_blockingintermediater*   r*   r+   
meta_copy_  s   r   c                 C   sX   t |  }t |  }||  krdn|| ||  }||d ||| ||fS Nr   )r   rX   r   rT   insert)tensorrT   result_sizesresult_strides
new_strider*   r*   r+   inferUnsqueezeGeometry*  s    r   c                 C   s0   t ||  d }t| |\}}| || | S r   )rl   rT   r   as_strided_)rP   rT   g_sizes	g_stridesr*   r*   r+   meta_unsqueeze_3  s   r   T)include_selfrP   rT   rG   sourcereducer   returnc                C   s   t j| t jdS )Nmemory_format)r/   rc   contiguous_formatrP   rT   rG   r   r   r   r*   r*   r+   meta_index_reduce;  s   
r   c                C      | S r%   r*   r   r*   r*   r+   meta_index_reduce_H  s   
r   c                 C   s.   t |  }|  dkr| ||< | |S )Nr   )r   rX   rT   rM   rN   )rP   rT   rG   result_sizer*   r*   r+   meta_index_selectV  s   
r   c                 C   s(   t ||  | j |t | ||S r%   )r/   _resize_output_rX   ri   copy_index_select)rP   rT   rG   r   r*   r*   r+   meta_index_select_out^  s   r   c                 C   
   |  dS Nr*   rN   rd   r*   r*   r+   meta_maxd     
r   c                 C   6   t | j|f}t| ||}| || j|tjdfS NrE   r   reduction_dimsrO   _compute_reduction_shaperN   r/   rK   rP   rT   keepdimoutput_shaper*   r*   r+   meta_max_dimj  
   r   c                 C   r   r   r   rd   r*   r*   r+   meta_mint  r   r   c                 C   r   r   r   r   r*   r*   r+   meta_min_dimz  r   r   c                 C   s4   |   r
t| j}n	t| tjd\}}tj| |dS )Ntype_promotion_kindrE   )r   r   r7   r   r   INT_TO_FLOATr/   rc   )rP   result_dtypert   r*   r*   r+   
meta_angle  s   
r   c                 C   s$   t ||  | j |t | S r%   )r/   r   rX   ri   r   angle)rP   r   r*   r*   r+   meta_angle_out  s   r   c                 C      d S r%   r*   )valr*   r*   r+   assert_async     r   c                 C   r   r%   r*   )r   
assert_msgr*   r*   r+   assert_async_meta  r   r   r7   r   ri   r   r   c                 C   s   t jg ddS )Nmetari   r   r  r*   r*   r+   make_dep_token  s   	r  c                 C   (   t | ttfrtdt| ||d d S Nz'Constraining SymFloat or Symbool is nyiminmax)
isinstancer	   r   
ValueErrorr    rX   r  r	  r*   r*   r+   sym_constrain_range     r  c                 C      t j| ||d |S Nr  )r"   r  rX   r  r	  	dep_tokenr*   r*   r+   functional_sym_constrain_range     r  c                 C   r  r  )r
  r	   r   r  r   r  r*   r*   r+   sym_constrain_range_for_size  r  r  c                 C   r  r  )r"   r  r  r*   r*   r+   'functional_sym_constrain_range_for_size  r  r  c                 C   s   |S r%   r*   )r   r   r  r*   r*   r+   functional_assert_async_meta  r   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.rR   z5: A must be batches of square matrices, but they are  by 	 matrices)rT   rX   )rP   r  r*   r*   r+   ra     s    ra   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  r*   r  rP   r*   r+   r>     
   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  rE   r*   r   r*   r+   r>     r!  rR   r  c                      s   d  d d  d dS )Nz3A must be batches of square matrices, but they are r  r  rR   r  rW   r*   r  r*   r+   r>     s
   c                      s:   d d  d d  d d d d d 
S )NzIncompatible matrix sizes for z: each A matrix is rR   r  z but each b matrix is r  rW   r*   r  r  rP   r*   r+   r>     s   )r/   rA   ri   r7   rX   )rP   r  r  r*   r#  r+   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 r*   r*   r7   r  r*   r+   r>         z(checkFloatingOrComplex.<locals>.<lambda>c                      r'  )Nz*: Low precision dtypes not supported. Got r*   r*   r(  r*   r+   r>     r)  )	r7   r/   rA   r   r   r3   r5   r2   r4   )r%  r  r&  r*   r(  r+   rb      s   rb   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.r*   r*   r*  r  r*   r+   r>     r?   zcheckIsMatrix.<locals>.<lambda>)r/   rA   rT   )r  r  r*  r*   r,  r+   checkIsMatrix  s   
r-  Br   c                    sZ   t   t tr ddkn	 ddk fdd d S )Nr  rR   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  rx   rR   rV   )rW   r*   r  r.  r  r   r*   r+   r>   !  s   
z#checkInputsSolver.<locals>.<lambda>)ra   r-  r/   rA   rX   )r  r.  r   r  r*   r0  r+   checkInputsSolver  s   

*r1  resultfn_namer   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  r*   r3  r   r2  r4  r*   r+   r>   .  s   z!checkSameDevice.<locals>.<lambda>)r/   rA   ri   )r3  r2  r   r4  r*   r5  r+   checkSameDevice)  s   
r6  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 r*   r*   r7  r*   r+   r>   9     
 zcheckUplo.<locals>.<lambda>)upperr/   rA   r   )r7  UPLO_uppercaser*   r;  r+   	checkUplo5  s
   
r?  eigenvalueseigenvectorsr9  	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	row_majorr   rE   )
ra   r?  r   rO   rN   r   r   popr9   r7   )r  r7  rB  rO   vecsvalsr*   r*   r+   meta__linalg_eigh=  s   


rH  r   c                 C   s   | j jtjdddS )Nr   r  rR   )mTcloner/   r   	transpose)r   r*   r*   r+   cloneBatchedColumnMajorT  s   rL  r=  c                 C   s   t | S r%   )rL  )rP   r  r=  r*   r*   r+   _cholesky_solve_helperX  s   rM  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[   r*   rd   r*   r+   r>   c  r)  z cholesky_solve.<locals>.<lambda>c                      rN  )Nz-u should have at least 2 dimensions, but has rO  rP  r*   r"  r*   r+   r>   g  r)  cholesky_solve)r/   rA   r[   !_linalg_broadcast_batch_dims_namerM  )rP   r  r=  self_broadcastedA_broadcastedr*   r   r+   rQ  ^  s   

rQ  c                 C   s.   |   dkrtj| tjdS t| d t| S )Nr   r   cholesky)rM   r/   rc   legacy_contiguous_formatra   rL  rP   r=  r*   r*   r+   rU  o  s   
rU  c                 C   s   t | d t| S )Ncholesky_inverse)ra   rL  rW  r*   r*   r+   rX  x  s   
rX  check_errorsc                 C   sf   t | d t| d | j}t|}t|d}| |}||| | j|d|d  tjd}||fS )Nzlinalg.choleskyFr   r   rE   )	ra   rb   rO   r   r   rN   r   r/   int32)r  r=  rY  A_shaper[   	L_stridesr9  infosr*   r*   r+   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   rI   )NzHtorch.linalg.householder_product: input must have at least 2 dimensions.r*   r*   r*   r*   r+   r>     rJ   z,linalg_householder_product.<locals>.<lambda>r  rR   c                   S   rI   )Nzbtorch.linalg.householder_product: input.shape[-2] must be greater than or equal to input.shape[-1]r*   r*   r*   r*   r+   r>     rJ   c                   S   rI   )Nz`torch.linalg.householder_product: input.shape[-1] must be greater than or equal to tau.shape[-1]r*   r*   r*   r*   r+   r>     rJ   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 rP  r*   r   r_  r*   r+   r>     
   c                      r:  )Nzltorch.linalg.householder_product: Expected batch dimensions of tau to be equal to input.shape[:-2], but got r*   r*   actual_batch_tau_shaper*   r+   r>        c                      r`  )Nz,torch.linalg.householder_product: tau dtype z does not match input dtype rE   r*   rb  r*   r+   r>        
z torch.linalg.householder_productr_  FrC  rX   r   r7   ri   )
r/   rA   r[   rX   rO   r7   r6  empty_stridedr   ri   )r   r_  expected_batch_tau_shaper*   )re  r   r_  r+   linalg_householder_product  sD   


rk  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&  rC  r  rE   ra   rb   rN   rO   r   r   r/   rZ  )r  rY  r9  r]  r*   r*   r+   linalg_inv_ex_meta  s   
rm  LDpivotsinfo)	hermitianrY  rq  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_exFrC  rh  rR   rE   r  )
ra   rb   r/   ri  rO   r   r7   ri   rN   int)rP   rq  rY  rn  ro  rp  r*   r*   r+   linalg_ldl_factor_ex_meta  s   


rs  )rq  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                      rN  )NzMtorch.linalg.ldl_solve: Expected B to have at least 2 dimensions, but it has rO  rP  r*   )r.  r*   r+   r>        z'linalg_ldl_solve_meta.<locals>.<lambda>rR   c                      rN  )Nzjtorch.linalg.ldl_solve: Expected LD.shape[:-1] and pivots.shape to be the same, but got pivots with shape  insteadrO   r*   ro  r*   r+   r>     rt  c                      rD   )Nz<torch.linalg.ldl_solve: Expected pivots to be integers. Got rE   r*   rw  r*   r+   r>     rH   c                      r`  )Nz!torch.linalg.ldl_solve: LD dtype z does not match b dtype rE   r*   )r.  rn  r*   r+   r>         FrC  rh  )ra   rb   r$  r/   rA   r[   rO   r   is_integer_dtyper7   _linalg_broadcast_batch_dimsri  r   ri   )rn  ro  r.  rq  expected_pivots_shapeB_broadcast_sizert   r*   )r.  rn  ro  r+   linalg_ldl_solve_meta  s6   







r}  Pr8  )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                      rN  )Nz@linalg.lu: Expected tensor with 2 or more dimensions. Got size: ru  rv  r*   r"  r*   r+   r>     r)  z linalg_lu_meta.<locals>.<lambda>r  rR   r   )r/   rA   r[   r   rO   r  rN   )	r  r  sizesmr   kr~  r9  r8  r*   r"  r+   linalg_lu_meta  s$   





r  LU)r  rY  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                      rN  )NzFtorch.lu_factor: Expected tensor with 2 or more dimensions. Got size: ru  rv  r*   r"  r*   r+   r>   4  r)  z*linalg_lu_factor_ex_meta.<locals>.<lambda>r  rR   FrC  rh  rE   )r/   rA   r[   r   rO   ri  r   r7   ri   rE  r  rN   rr  )	r  r  rY  r  r  r   r  ro  rp  r*   r"  r+   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 ru  rE   r*   )r.  r  r*   r+   r>   \  r!  z&linalg_lu_solve_meta.<locals>.<lambda>c                   S   rI   )NzElinalg.lu_solve: pivots should be a Tensor of scalar type torch.int32r*   r*   r*   r*   r+   r>   c  rJ   zlinalg.lu_solverR   c                   S   rI   )NzYlinalg.lu_solve: Number of pivots per batch should be same as the dimension of the matrixr*   r*   r*   r*   r+   r>   k  rJ   c                      rN  )Nzclinalg.lu_solve: Expected LU.shape[:-1] and pivots.shape to be the same, but got pivots with shape ru  rv  r*   rw  r*   r+   r>   q  rt  rC  rh  r   )rb   r/   rA   r7   rr  ra   r1  rX   rO   rz  ri  r   ri   rM   r   conj)r  ro  r.  r   r  r|  rt   r2  r*   )r.  r  ro  r+   linalg_lu_solve_metaN  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                      rN  )NzFtorch.lu_unpack: Expected tensor with 2 or more dimensions. Got size: ru  rv  r*   r  r*   r+   r>     r)  z lu_unpack_meta.<locals>.<lambda>c                   S   rI   )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_factorr*   r*   r*   r*   r+   r>        r  rR   r   )	r/   rA   r[   r7   rZ  r   rO   r  rN   )r  ro  r  r  r  r  r   r  r~  r9  r8  r*   r  r+   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reducedTcompleteFrc                         d  dS )Nzqr received unrecognized mode 'z=' but expected one of 'reduced' (default), 'r', or 'complete'r*   r*   r  r*   r+   r>     s   z _parse_qr_mode.<locals>.<lambda>r/   rA   )r  	compute_qr  r*   r  r+   _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  rR   FrC  r   )	r-  rb   r  rO   r  r   rN   r   r   )r  r  r  reduced_moder  r   r  Q_shaper  R_shaper  r*   r*   r+   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  rE   rh  rR   )ra   rb   rO   rN   r9   r7   r/   ri  r   ri   rZ  )r  rO   r  r  r  ro  r*   r*   r+   _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  rR   FrC  cudar   rE   )r-  rb   r   rO   r  rN   r   r   device_hintr9   r7   )r  r  r  r  r   r  r   r  U_shaper8  V_shapeVis_cudaSr*   r*   r+   _linalg_svd_meta  s$   







r  arg1arg2c                 C   sn   | j d d }|j d d }t||}t|}|| d| dg7 }t|}||d|dg7 }||fS )Nr  rR   )rO   r   r   rX   )r  r  arg1_batch_sizesarg2_batch_sizesexpand_batch_portionarg1_expand_sizearg2_expand_sizer*   r*   r+   rz    s   
rz  c                 C   sV   |rt | || t| |\}}|| jkr| n| |}||jkr"|n||}||fS r%   )r$  rz  rO   expand)r  r  r  r  r  arg1_broadcastedarg2_broadcastedr*   r*   r+   rR  /  s   rR  rZ   c                 C   s6   | j d d }|jdkp| jd |jko|j |k}|S )NrR   r   )rO   r[   )r   rZ   expected_batched_rhs_shapevector_caser*   r*   r+   linalg_solve_is_vector_rhsA  s
   
r  )r   rY  r2  r  ro  rp  c                   sn  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} 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                      s   d j  dj  dS )NzKlinalg.solve: Expected A and B to have the same dtype, but found A of type r  ru  rE   r*   r  r.  r*   r+   r>   X  r!  z"_linalg_solve_ex.<locals>.<lambda>rR   c                   S   rI   )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)r*   r*   r*   r*   r+   r>   c  r  rh  FrE   r  c                 s   s    | ]}|d uV  qd S r%   r*   rs   rx   r*   r*   r+   	<genexpr>{      z#_linalg_solve_ex.<locals>.<genexpr>)	copy_fromcopy_toexact_dtype)rb   r/   rA   r7   r  	unsqueezer1  rz  ri  r   ri   rO   r[   rN   rZ  allzipr   r   r   r   )r  r.  r   rY  r2  r  ro  rp  r  B_B_broad_shapert   result_shaperesult_rO   r[   LU_pivots_info_r   resr  or*   r  r+   _linalg_solve_exI  sL   



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  rR   )rN   r
  r   r1  rR  rK  is_contiguousis_conjr   rO   r   resize_
transpose_)	r  r.  r=  r   r  r   r  A_avoid_copy_Ar*   r*   r+   linalg_solve_triangular_meta  s   
r  solutioncloned_coefficientrK  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                      rN  )NzMtorch.triangular_solve: Expected b to have at least 2 dimensions, but it has rO  rP  r*   rd   r*   r+   r>     rt  z'triangular_solve_meta.<locals>.<lambda>c                      rN  )NzMtorch.triangular_solve: Expected A to have at least 2 dimensions, but it has rO  rP  r*   r"  r*   r+   r>     rt  triangular_solveFrC  rh  r   c                   S   rI   )Nz+triangular_solve: Got an unexpected layout.r*   r*   r*   r*   r+   r>     rJ   )r/   rA   r[   r$  r   stridedrz  ri  r   r7   ri   
sparse_csr
sparse_bsrrc   rN   )	rP   r  r=  rK  r  self_broadcast_sizeA_broadcast_sizer  r  r*   r   r+   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  FrC  rR   rE   rl  )r  detr  ro  r*   r*   r+   _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   rI   )Nz3torch.ormqr: input must have at least 2 dimensions.r*   r*   r*   r*   r+   r>     rJ   zormqr.<locals>.<lambda>c                   S   rI   )Nz3torch.ormqr: other must have at least 2 dimensions.r*   r*   r*   r*   r+   r>     rJ   r  rR   c                      r  )Ntorch.ormqr: other.shape[z0] must be greater than or equal to tau.shape[-1]r*   r*   left_size_conditionr*   r+   r>     rH   c                      r  )Nr  z"] must be equal to input.shape[-2]r*   r*   r  r*   r+   r>     rH   c                   S   rI   )NzHtorch.ormqr: tau.shape[-1] must be less than or equal to input.shape[-1]r*   r*   r*   r*   r+   r>     rJ   r   c                      r`  )Nz[torch.ormqr: Expected tau to have one dimension less than input, but got tau.ndim equal to ra  rP  r*   rb  r*   r+   r>     rc  c                      r`  )Nzhtorch.ormqr: Expected other to have the same number of dimensions as input, but got other.ndim equal to ra  rP  r*   r   rZ   r*   r+   r>     rc  c                      r:  )NzWtorch.ormqr: Expected batch dimensions of tau to be equal to input.shape[:-2], but got r*   r*   rd  r*   r+   r>     rf  c                      r:  )NzYtorch.ormqr: Expected batch dimensions of other to be equal to input.shape[:-2], but got r*   r*   )actual_batch_other_shaper*   r+   r>     rf  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 rE   r*   rb  r*   r+   r>   "  rc  c                      r  )NzRtorch.ormqr: Expected input and other to have the same dtype, but input has dtype z and other has dtype rE   r*   r  r*   r+   r>   )  rc  ztorch.ormqrr_  rZ   FrC  rh  )	r/   rA   r[   rO   r7   r6  ri  r   ri   )r   r_  rZ   r   rK  expected_batch_shaper*   )r  re  r   r  rZ   r_  r+   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   r*   )rT   paddingr*   r+   r>   =      z,_padding_check_valid_input.<locals>.<lambda>r   r   c                      s    d d  d d  dj  S )Nz	Expected r   zD or r   zcD (batch mode) tensor with possibly 0 batch size and other non-zero dimensions for input, but got: rv  r*   )rT   r   r*   r+   r>   R  s   )r/   rA   r   r[   r   rX   )r   r  rT   	input_dimis_batch_modevalid_batch_modevalid_non_batch_moder   r*   )rT   r   r  r+   _padding_check_valid_input:  s$   r  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   rU   rS   c                         d d d  dj  S NzcArgument #4: Padding size should be less than the corresponding input dimension, but got: padding (, ) at dimension 
 of input rv  r*   dim_wr   pad_lpad_rr*   r+   r>   n     z_pad1d_common.<locals>.<lambda>c                         d  d S )Nz
input (W: z%) is too small. Calculated output W: r*   r*   )input_woutput_wr*   r+   r>   v  r?   r   )r[   rX   r  r/   rA   rN   )r   r  is_reflection	dim_planenbatchnplaner*   )r  r   r  r  r  r  r+   _pad1d_commonY  s0   




r  c                 C      t | |ddS NTr  r  r   r  r*   r*   r+   meta_reflection_pad1d     r  c                 C   r  NFr  r  r  r*   r*   r+   meta_replication_pad1d  r  r
  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   rI   )Nz padding size is expected to be 2r*   r*   r*   r*   r+   r>     rJ   z(_pad1d_backward_common.<locals>.<lambda>rU   c                      r  r  rv  r*   r  r*   r+   r>     r  c                         d d   S Nz(grad_output width unexpected. Expected: , Got: rW   r*   r  grad_outputr  r*   r+   r>         r/   rA   r   r[   rX   rN   rO   )r  r   r  r  r  r*   )r  r  r   r  r  r  r+   _pad1d_backward_common  s$   

r  c                 C      t | ||ddS r  r  r  r   r  r*   r*   r+   meta_reflection_pad1d_backward     r  c                 C   r  r	  r  r  r*   r*   r+   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   rS      c                      r  r  rv  r*   r  r*   r+   r>     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 (r  r  r  rv  r*   dim_hr   pad_bpad_tr*   r+   r>     r  c                      s   d  d d d S )Nz
input (H:  W: z%) is too small. Calculated output H: r*   r*   )input_hr  output_hr  r*   r+   r>     s
   rU   r  r[   rX   r/   rA   rN   )r   r  r  
dim_slicesr  r[   r   r*   )r  r  r   r!  r  r"  r  r  r  r  r  r+   _pad2d_common  sB   




r%  c                 C   r  r  r%  r  r*   r*   r+   meta_reflection_pad2d  r  r'  c                 C   r  r	  r&  r  r*   r*   r+   meta_replication_pad2d  r  r(  c                    s   dd d}d}|j }| dkr!|d }d7  d7  |d7 }|\}}}}	|| }
|  }| }|| |	 || | tkfdd t k fdd ||j S )Nr   r   r   r  c                      r  r  rW   r*   r  r*   r+   r>     r  z%meta_pad2d_backward.<locals>.<lambda>c                      r  Nz)grad_output height unexpected. Expected: r  rW   r*   r  r  r"  r*   r+   r>     r  )rO   rT   r/   rA   rX   rN   )r  rP   r  r  r  r=   r  r  r  r  r   r!  r  r*   )r  r  r  r"  r  r+   meta_pad2d_backward  s2   
r+  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 )NrU   r   r   r   rS      c                      r  r  rv  r*   r  r*   r+   r>   <  r  z_pad3d_common.<locals>.<lambda>c                      r  r  rv  r*   r  r*   r+   r>   C  r  c                      r  )NzcArgument #8: Padding size should be less than the corresponding input dimension, but got: padding (r  r  r  rv  r*   )dim_dr   pad_bkpad_fr*   r+   r>   J  r  c                      s(   d  d d d d d S )Nz
input (D:  H: r   z%) is too small. Calculated output D: r*   r*   )input_dr!  r  output_dr"  r  r*   r+   r>   R  s   r#  )r   r  r  r  
batch_moder  r   r*   )r-  r  r  r   r1  r!  r  r2  r"  r  r  r.  r/  r  r  r  r+   _pad3d_common  sP   





r4  c                 C   r  r  r4  r  r*   r*   r+   meta_reflection_pad3d^  r  r6  c                 C   r  r	  r5  r  r*   r*   r+   meta_replication_pad3dd  r  r7  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   rI   )Nz padding size is expected to be 6r*   r*   r*   r*   r+   r>   t  rJ   z%meta_pad3d_backward.<locals>.<lambda>rU   r   r   r,  c                      r  r  rW   r*   r  r*   r+   r>     r  c                      r  r)  rW   r*   r*  r*   r+   r>     r  c                      r  )Nz(grad_output depth unexpected. Expected: r  rW   r*   )r-  r  r2  r*   r+   r>     r  r  )r  r   r  r  r  r  r  r/  r.  r1  r!  r  r*   )r-  r  r  r  r2  r"  r  r+   meta_pad3d_backwardj  s<   




r9  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   rI   )Nz(_pdist_forward requires contiguous inputr*   r*   r*   r*   r+   r>     rJ   z%meta__pdist_forward.<locals>.<lambda>r   r   r   r   )r/   rA   r  rX   rN   r   rV  )rP   r:  r   r*   r*   r+   meta__pdist_forward  s   
r;  gradpdistc                 C   s8   t | dd  t | dd  t j|t jdS )Nc                   S   rI   )Nz._pdist_backward requires self to be contiguousr*   r*   r*   r*   r+   r>     rJ   z&meta__pdist_backward.<locals>.<lambda>c                   S   rI   )Nz/_pdist_backward requires pdist to be contiguousr*   r*   r*   r*   r+   r>     rJ   r   )r/   rA   r  rc   rV  )r<  rP   r:  r=  r*   r*   r+   meta__pdist_backward  s   r>  r   )betaalphac          	         s     d}  d} d}|||ft  dkdd  t dkdd  tj j  ko=jkn   fdd  j}j|d |d td kocd kfd	d   S )
Nr   r   r   rU   c                   S   rI   Nzbatch1 must be a 3D tensorr*   r*   r*   r*   r+   r>     rJ   zmeta_baddbmm.<locals>.<lambda>c                   S   rI   Nzbatch2 must be a 3D tensorr*   r*   r*   r*   r+   r>     rJ   c                      s   dj  d j  dj  S )Nz+Input dtypes must be the same, got: input: z
, batch1: z
, batch2: rE   r*   )batch1batch2rP   r*   r+   r>         c                	      &   d d d d  d d  d	S Nz@Expected size for first two dimensions of batch2 tensor to be: [r  z] but got: [r   r   ].r*   r*   batch2_sizesbscontraction_sizer*   r+   r>     s   )rX   r  r/   rA   rT   r7   rO   rN   )	rP   rC  rD  r?  r@  dim1dim2dim3batch1_sizesr*   )rC  rD  rJ  rK  rL  rP   r+   meta_baddbmm  s&   


rQ  c                C      t |  S r%   r/   rc   rn   )rP   r   r*   r*   r+   meta_bernoulli  s   rT        ?c                 C   r   r%   r*   rP   r:  r   r*   r*   r+   meta_bernoulli_  r   rW  c                 C   rR  r%   rS  rV  r*   r*   r+   meta_bernoulli_p  r  rX  c                 C   s6   t |
|  k dd  t j| t jd}t | |fS )Nc                   S   rI   )NzJError in fused_moving_avg_obs_fake_quant_cpu: ch_axis must be < self.dim()r*   r*   r*   r*   r+   r>     rJ   z6meta__fused_moving_avg_obs_fq_helper.<locals>.<lambda>rE   )r/   rA   rT   rc   bool)rP   observer_onfake_quant_onrunning_minrunning_maxscale
zero_pointaveraging_const	quant_min	quant_maxch_axisper_row_fake_quantsymmetric_quantmaskr*   r*   r+   $meta__fused_moving_avg_obs_fq_helper  s   
rg  c                    sT   t  dko  dk fdd  fdd}t    k| d S )Nr   c                         d   d    dS )Nz1D tensors expected, but got zD and z	D tensorsrS   r*   rZ   rP   r*   r+   r>     s    zdot_check.<locals>.<lambda>c                	      s.   d   d    d   d    d	S )Nz+inconsistent tensor size, expected tensor [z] and src [z.] to have thesame number of elements, but got rV   z elements respectivelyrM   r*   ri  r*   r+   numel_error   s   zdot_check.<locals>.numel_error)r/   rA   rT   rM   )rP   rZ   rk  r*   ri  r+   	dot_check  s   rl  c                 C   s   t | | | dS r   )rl  rN   )rP   r   r*   r*   r+   meta_dot	  s   

rm  c                    sn   t |  dkdd  t | dkdd  | j\ |j\t  k fdd | S )Nr   c                   S   rI   )Nza must be 2Dr*   r*   r*   r*   r+   r>     rJ   zmeta_mm.<locals>.<lambda>c                   S   rI   )Nzb must be 2Dr*   r*   r*   r*   r+   r>     rJ   c                	      s   d d  d d d	S )Nz/a and b must have same reduction dim, but got [r  z] X [rH  r*   r*   M1M2Nr~  r*   r+   r>     s    )r/   rA   rT   rO   rN   abr*   rn  r+   meta_mm  s   

ru  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   Nrv  rs   r   dimsrP   r*   r+   r    s   $ z+_compute_reduction_shape.<locals>.<genexpr>)r@   r   r[   r   compute_reduction_output_shaperO   )rP   rx  r   r*   rw  r+   r     s   r   strc                 C   s   t | tjjr| jjS dS )Nr  )r
  r/   _subclasses
FakeTensorfake_devicetype)r   r*   r*   r+   r  '  s   r  input_tensorweightr   r  dilationis_transposedgroupsoutput_paddingc                 C   s  dt dt dt dt dt dt fdd}dt dt dt dt dt d	t dt fd
d}	|jdd  }
| jdd  }|r<||jd  }n|jd }|jd | | jd krQtd| jd |g}t|tre|gt| }nt|dkrt|d gt| }t|tr|gt| }nt|dkr|d gt| }t|tr|gt| }nt|dkr|d gt| }d }|rt|tr|gt| }nt|dkr|d gt| }n|}tt|D ]2}|r||	|| || || |
| || ||  q|||| || || |
| ||  q|S )Nlnr:  r   r  sr   c                 S   s$   | 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   r*   )r  r:  r   r  r  r*   r*   r+   _formula8  s   $z+calc_conv_nd_return_shape.<locals>._formular'   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   r*   )r  r:  r   r  r  r'   r*   r*   r+   _formula_transposedI  s   (z6calc_conv_nd_return_shape.<locals>._formula_transposedr   r   r   zInvalid channel dimensions)rr  rO   RuntimeErrorr
  r   r   r   r   )r  r  r   r  r  r  r  r  r  r  kernel_sizerx  out_channels	ret_shapeoutput_padding_listr   r*   r*   r+   calc_conv_nd_return_shape.  sZ   "
&




"r  c                 C      t j| t jkS r%   r/   _prims_commonsuggest_memory_formatchannels_lasttenr*   r*   r+   is_channels_last     r  biasc	              	      sH    fdd}	t  ||||||r|n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  r/   r  r  r   preserve_formatr*   r  r  r*   r+   pick_memory_format  s   z%meta_conv.<locals>.pick_memory_formatr   )r  rN   r   )r  r  r  r   r  r  r  r  r  r  	shape_outr   r*   r  r+   	meta_conv  s   

r  mkldnnc
              	   C   s6   t | ||||d|g }
| |
}tj}|j|d}|S )NFr   )r  rN   r/   r  r   )r  r  r  r  r   r  r  attrscalars	algorithmr  r   out_memory_formatr*   r*   r+   meta_mkldnn_convolution_default  s   
r  c                 C   s$   |  g | jd d |jd R S NrR   r   rN   rO   )r  r  r  r  r  r  r*   r*   r+   meta_linear_pointwise_default  s   $r  mklc                 C   s$   |  g | jd d |jd R S r  r  )r  packed_weightorig_weightr  r   r*   r*   r+   meta_mkl_linear  s   r  onednnc              	   C   s@   t | ||||	d|
d }| j||rtjnd d}|jtjd}|S )NFrE   r   )r  rN   r/   float32r   r  )rx   x_scalex_zpww_scalew_zpr  r   r  r  r  output_scaleoutput_zero_pointfp32_outputr  r  r  r  r   r*   r*   r+   meta_qconv2d_pointwise  s   
r  c                 C   s4   t | j}|jd |d< | j||	rtjnd d}|S )Nr   rR   rE   )r   rO   rN   r/   r  )rx   r  r  r  r  r  r  r  r  r  post_op_namepost_op_argspost_op_algorithmr   r   r*   r*   r+   meta_qlinear_pointwise  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] == r  zbut got : dimension z] = rT   rO   r*   rT   dim_sizerX   r   r*   r+   r>   0  s    z check_dim_size.<locals>.<lambda>)r/   rA   rT   rO   )r   rT   r  rX   r*   r  r+   check_dim_size-  s   r  r*   r   c                 C   sb  dd }|d|\}}	t t|dv dd  t|dkr#||	}
}nt|dkr3|d |d }
}n|d	|\}
}|d
|\}}t |d u pJ|dkdd  |  dkrZ| 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 intsr*   r*   r  r*   r+   r>   B  rH   z1meta_avg_pool2d.<locals>.unpack.<locals>.<lambda>r   r   r/   rA   r   r  r   HWr*   r  r+   unpack?     

zmeta_avg_pool2d.<locals>.unpackr  r   r   r   c                   S   rI   NzOavg_pool2d: stride must either be omitted, a single int, or a tuple of two intsr*   r*   r*   r*   r+   r>   K  rJ   z!meta_avg_pool2d.<locals>.<lambda>r   r   r   r  c                   S   rI   Nzdivisor must be not zeror*   r*   r*   r*   r+   r>   X  rJ   r  r  rR   rU   r7   ri   r   )r/   rA   r   rT   rX   pooling_output_shaper   r  pool2d_shape_checkrj   r7   ri   )r   r  r   r  	ceil_modecount_include_paddivisor_overrider  kHkWdHdWpadHpadWr  nInputPlaneinputHeight
inputWidthoutputHeightoutputWidthr   rX   r*   r*   r+   meta_avg_pool2d5  sb   
	




r  c                 C   sj   t | ||||||dd|	|
|||| |  }|	}t|||d | t|||d | t|||d | d S )Nr   rU   r   )r  rT   r  )r   
gradOutputr  r  r  r  r  r  r  r  r  r  r  r  
mem_formatr[   nOutputPlaner*   r*   r+   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   rI   )NzKavg_pool2d: kernel_size must either be a single int, or a tuple of two intsr*   r*   r*   r*   r+   r>     rJ   z*meta_avg_pool2d_backward.<locals>.<lambda>r   c                   S   rI   r  r*   r*   r*   r*   r+   r>     rJ   c                   S   rI   )NzGavg_pool2d: padding must either be a single int, or a tuple of two intsr*   r*   r*   r*   r+   r>     rJ   c                   S   rI   r  r*   r*   r*   r*   r+   r>     rJ   r  r  r  r  rR   r  )r/   rA   r   rO   rT   r  r   r  r  rj   r7   ri   )gradOutput_r   r  r   r  r  r  r  r  r  r  r  r  r  
input_sizer  r  r  r  r  r  r  r*   r*   r+   meta_avg_pool2d_backward  sj   "(
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 | 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   rU   c                   S   rI   NzFavg_pool3d: kernel_size must be a single int, or a tuple of three intsr*   r*   r*   r*   r+   r>   	  rJ   z!meta_avg_pool3d.<locals>.<lambda>r   r   r   c                   S   rI   NzJavg_pool3d: stride must be omitted, a single int, or a tuple of three intsr*   r*   r*   r*   r+   r>   	  rJ   c                   S   rI   NzBavg_pool3d: padding must be a single int, or a tuple of three intsr*   r*   r*   r*   r+   r>   	  rJ   r  r,  c                   S   rI   Nz9non-empty 4D or 5D (batch mode) tensor expected for inputr*   r*   r*   r*   r+   r>   	  rJ   c                   S   rI   r  r*   r*   r*   r*   r+   r>   "	  rJ   r  r  r  rR   zavg_pool3d()T)check_input_sizer  )r/   rA   r   r[   rX   r  pool3d_shape_checkrN   )r   r  r   r  r  r  r  kTr  r  dTr  r  padTr  r  r  nslicesitimeiheightiwidthotimeoheightowidthr*   r*   r+   meta_avg_pool3d  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 |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   rI   r  r*   r*   r*   r*   r+   r>   \	  rJ   z*meta_avg_pool3d_backward.<locals>.<lambda>r   r   r   c                   S   rI   r  r*   r*   r*   r*   r+   r>   d	  rJ   c                   S   rI   r  r*   r*   r*   r*   r+   r>   l	  rJ   r  c                   S   rI   r  r*   r*   r*   r*   r+   r>   t	  rJ   c                   S   rI   r  r*   r*   r*   r*   r+   r>   y	  rJ   r  r  r  rR   zavg_pool3d_backward())	r/   rA   r   r[   rX   r  avg_pool3d_backward_shape_checkrN   rO   )r  r   r  r   r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  otime_for_shape_checkoheight_for_shape_checkowidth_for_shape_checkr*   r*   r+   meta_avg_pool3d_backwardN	  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 )NrU   r  c                      rD   )Nz"Expected 3D or 4D tensor, but got rv  r*   rd   r*   r+   r>   	  rH   z*meta_adaptive_avg_pool2d.<locals>.<lambda>r  r  )
r/   rA   r[   rO   r@   r   r  rj   r7   ri   )rP   output_sizer   r   r*   rd   r+   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  r,  c                      rD   )Nz"Expected 4D or 5D tensor, but got rv  r*   rd   r*   r+   r>   	  rH   z*meta_adaptive_avg_pool3d.<locals>.<lambda>r  )r/   rA   r[   rN   rO   r@   )rP   r  r*   rd   r+   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                      s   d j  d dS )Nz{adaptive_avg_pool2d_backward(): Expected grad_output to have non-zero                       size for non-batch dimensions,  with dimension  being emptyrv  r*   )grad_outr   r*   r+   r>   	  s
    z4meta__adaptive_avg_pool2d_backward.<locals>.<lambda>rU   r  c                      rD   )NzBadaptive_avg_pool2d_backward(): Expected 3D or 4D tensor, but got rv  r*   rd   r*   r+   r>   	  rH   c                      r`  Nzexpected dtype z! for `grad_output` but got dtype rE   r*   )r  rP   r*   r+   r>   	  rx  r   )r[   r   r/   rA   rX   r7   r   r  r  rN   rO   r   )r  rP   r[   r   r*   )r  r   rP   r+   "meta__adaptive_avg_pool2d_backward	  s$   

r  c                 C   s   t | d tj|tjdS )Nadaptive_avg_pool3d_backwardr   )!_adaptive_pool_empty_output_checkr/   rc   rV  r  rP   r*   r*   r+   "meta__adaptive_avg_pool3d_backward	  s   
r
  r  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  rv  r*   r*  r  r   r*   r+   r>   	  s
   z3_adaptive_pool_empty_output_check.<locals>.<lambda>)r[   r   r/   rA   rX   )r  r*  r[   r*   r  r+   r  	  s   r  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 )NrU   r  c                      rD   )Nz:adaptive_max_pool2d(): Expected 3D or 4D tensor, but got: rv  r*   r   r*   r+   r>   	  rH   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  rv  r*   r   r   r*   r+   r>   	  
   r   c                   S   rI   )NzCadaptive_max_pool2d(): internal error: output_size.size() must be 2r*   r*   r*   r*   r+   r>   	  rJ   r  rU   rE   r   )r[   r/   rA   r   rX   r   rN   rk   r   r  r   )r   r  r[   dimHsizeBsizeDosizeHosizeWr^   r   rg   r   r*   r  r+   meta_adaptive_max_pool2d	  sD   







r  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                      rD   )NzKadaptive_max_pooling2d_backward(): Expected 3D or 4D grad_output, but got: rv  r*   r  r*   r+   r>   
  rH   z3meta_adaptive_max_pool2d_backward.<locals>.<lambda>adaptive_max_pool2d_backwardc                      r`  r  rE   r*   r  r   r*   r+   r>    
  rx  r   )
r[   r/   rA   r  r7   r   r  rN   rO   r   )r  r   rg   r[   r   r*   r  r+   !meta_adaptive_max_pool2d_backward
  s   



r  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                      rD   )Nz:adaptive_max_pool3d(): Expected 4D or 5D tensor, but got: rv  r*   r  r*   r+   r>   -
  rH   z*meta_adaptive_max_pool3d.<locals>.<lambda>r   r   c                      r  )Nzjadaptive_max_pool3d(): Expected input to have non-zero size for non-batch dimensions, but input has sizes r  r  rv  r*   r  r*   r+   r>   2
  r  rU   c                   S   rI   )NzCadaptive_max_pool3d(): internal error: output_size.size() must be 3r*   r*   r*   r*   r+   r>   :
  rJ   r,  r  rE   )r[   r/   rA   r   rX   r   rN   rk   )r   r  r[   dimDr  r  osizeTr  r  r^   r   rg   r*   r  r+   meta_adaptive_max_pool3d'
  s8   





r  c                 C   s   t | d ||jS )Nadaptive_max_pool3d_backward)r  rN   rO   )r  r   rg   r*   r*   r+   !meta_adaptive_max_pool3d_backwardS
  s   
r  c                 C   s   |d u rt d| |S )Nz:cannot repeat_interleave a meta tensor without output_size)r  rN   )repeatsr  r*   r*   r+   meta_repeat_interleave_TensorZ
  s   
r!  c                 C   s:   | j jsJ |j jsJ t| j|j}| j|t| j dS r   )r7   r   r   rO   rN   r   )realimagr^   r*   r*   r+   meta_complexa
  s   r$  c                 C   s   t | |S r%   )r   clone_preserve_stridesr   )rP   r7   r*   r*   r+   
view_dtypej
  s   r&  c                 C   sv   | j s	t| |S |  r#| rt| |  S t|  |S | r1t| |  S t| | | dS r   )r   r/   dotr  vdotr  rl  rN   rP   rZ   r*   r*   r+   r(  o
  s   

r(  )
fill_valuerX   r*  c                C   s   | j ||  ftjdS r   )rN   rT   r/   rK   )rP   rX   r*  r*   r*   r+   nonzero_static
  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 nqd}|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"|	| |
 S )Nc                   S   rI   )Nz#at least one index must be providedr*   r*   r*   r*   r+   r>   
  rJ   z#meta_index_Tensor.<locals>.<lambda>c                   S   rI   )Nz?tensors used as indices must be long, int, byte or bool tensorsr*   r*   r*   r*   r+   r>   
  rJ   c                      rD   )N)too many indices for tensor of dimension rP  r*   rd   r*   r+   r>   
  rH   c                	      s$   dj  d  dj  d  S )NzThe shape of the mask z
 at index z0 does not match the shape of the indexed tensor rv  r*   )r   rG   jr  rP   r*   r+   r>   
  s
    r   c                      s   dj  dt  dS )Nr,  z (got r/  )r[   r   r*   )rg   rP   r*   r+   r>   
  r  r   Fr   T)r/   rA   rY  	enumerater7   rK   rr  int8nonzeror   rL   r[   r   rO   r   selecttorch._refs_refsr   r   r   rN   )rP   rg   r2  r0  refsstatehas_contiguous_subspacerx  transposed_indicesbefore_shapeafter_shapereplacement_shaperT   r*   )r   rG   rg   r-  r  rP   r+   meta_index_Tensor
  s   








r;  c                 C   sT   d }d }d }|
d r|  | }|
d r|  | }|
d r%|  |}|||fS )Nr   r   r   )rN   rX   )grad_output_input_weight_bias_sizes_optr   r  r  
transposedr  r  output_maskbackend_grad_inputbackend_grad_weightbackend_grad_biasr*   r*   r+   meta_convolution_backward
  s   

rE  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   rU   c                   S   rI   rA  r*   r*   r*   r*   r+   r>     rJ   zmeta_addbmm.<locals>.<lambda>c                   S   rI   rB  r*   r*   r*   r*   r+   r>     rJ   r   c                         d  d d d S )Nz8batch1 and batch2 must have same number of batches, got r   rV   rW   r*   rC  rD  r*   r+   r>     rE  c                
      6   d  d d  d d d d d d	S )Nz#Incompatible matrix sizes for bmm (r   rx   r   rV   r/  rW   r*   rG  r*   r+   r>     
   c                   S   rI   )Nz.self tensor does not match matmul output shaper*   r*   r*   r*   r+   r>     rJ   )rX   r  r/   rA   rT   rN   )rP   rC  rD  r?  r@  rM  rN  r*   rG  r+   meta_addbmm  s$   

rJ  c                       t t t fdd d S )Nc                         dt   S NzExpect List[Tensor] but got r~  r*   rd   r*   r+   r>   ,  r)  z&meta__foreach_unaop_.<locals>.<lambda>r/   rA   r
  r   rd   r*   rd   r+   meta__foreach_unaop_!     	
rP  c                    (   t t t fdd dd  D S )Nc                      rL  rM  rN  r*   rd   r*   r+   r>   ;  r)  z%meta__foreach_unaop.<locals>.<lambda>c                 S      g | ]}t |qS r*   r/   rc   rs   r  r*   r*   r+   ru   =  r  z'meta__foreach_unaop.<locals>.<listcomp>rO  rd   r*   rd   r+   meta__foreach_unaop0  
   	
rV  c                    sX   t ttot t fdd t tdko"tt k fdd d S )Nc                         dt  dt   dS )Nz9The first two arguments of must be List[Tensor], but got rV   .rN  r*   ri  r*   r+   r>   C  
   z3_check_foreach_binop_tensor_lists.<locals>.<lambda>r   c                      rX  )Nz>self and other must be non-empty and match in length, but got rV   rY  r  r*   ri  r*   r+   r>   J  rZ  )r/   rA   r
  r   r   r)  r*   ri  r+   !_check_foreach_binop_tensor_lists@  s   r[  c                 C   s   t | | dd | D S )Nc                 S   rS  r*   rT  rU  r*   r*   r+   ru   ]  r  z,meta__foreach_binop_list.<locals>.<listcomp>r[  rP   rZ   r@  r*   r*   r+   meta__foreach_binop_listQ  s   
r^  c                 C      t | | d S r%   r\  r]  r*   r*   r+   meta__foreach_binop__list`  s   r`  c                    rK  )Nc                         dt   dS Nz4The first argument of must be List[Tensor], but got rY  rN  r*   rd   r*   r+   r>   y  r?   z-meta__foreach_binop__scalar.<locals>.<lambda>rO  rP   scalarr*   rd   r+   meta__foreach_binop__scalarn  rQ  re  c                    rR  )Nc                      ra  rb  rN  r*   rd   r*   r+   r>     r?   z,meta__foreach_binop_scalar.<locals>.<lambda>c                 S   rS  r*   rT  rU  r*   r*   r+   ru     r  z.meta__foreach_binop_scalar.<locals>.<listcomp>rO  rc  r*   rd   r+   meta__foreach_binop_scalar}  rW  rf  c                    st   t tdd  fD  fdd t t dkdd  t t tko3t tkdd  d S )Nc                 s       | ]}t |tV  qd S r%   r
  r   rs   lr*   r*   r+   r        z/meta__foreach_addcop__scalar.<locals>.<genexpr>c                      "   dt   dt  dt  S )Nz?All arguments of _foreach_addc*_ must be List[Tensor], but got r  , and rN  r*   rP   tensor1tensor2r*   r+   r>        z.meta__foreach_addcop__scalar.<locals>.<lambda>r   c                   S   rI   Nz$input tensor list must not be empty.r*   r*   r*   r*   r+   r>     rJ   c                   S   rI   Nz0All input tensor lists must have the same lengthr*   r*   r*   r*   r+   r>     rJ   r/   rA   r  r   rP   ro  rp  rd  r*   rn  r+   meta__foreach_addcop__scalar  s   rv  c                 C   r_  r%   r\  )rP   rZ   rd  r*   r*   r+   meta__foreach_lerp__scalar  s   rw  c                    s~   t tdd  fD  fdd t t dkdd  t t tko3t tkdd  dd	  D S )
Nc                 s   rg  r%   rh  ri  r*   r*   r+   r    rk  z.meta__foreach_addcop_scalar.<locals>.<genexpr>c                      rl  )Nz,All arguments must be List[Tensor], but got r  rm  rN  r*   rn  r*   r+   r>     rq  z-meta__foreach_addcop_scalar.<locals>.<lambda>r   c                   S   rI   rr  r*   r*   r*   r*   r+   r>     rJ   c                   S   rI   rs  r*   r*   r*   r*   r+   r>     rJ   c                 S   rS  r*   rT  rU  r*   r*   r+   ru     r  z/meta__foreach_addcop_scalar.<locals>.<listcomp>rt  ru  r*   rn  r+   meta__foreach_addcop_scalar  s   rx  c                    rR  )Nc                      rL  Nz'exponent must be a tensor list but got rN  r*   exponentr*   r+   r>     r)  z5meta__foreach_pow_scalar_and_tensor.<locals>.<lambda>c                 S   rS  r*   rT  )rs   er*   r*   r+   ru     r  z7meta__foreach_pow_scalar_and_tensor.<locals>.<listcomp>rO  )rP   r{  r*   rz  r+   #meta__foreach_pow_scalar_and_tensor  s
   
r}  c                    s   t tdd fD ot t j fdd t tdkdd  t ttko:ttkdd  d S )Nc                 s   rg  r%   rh  ri  r*   r*   r+   r    rk  z.meta__foreach_addcop_tensor.<locals>.<genexpr>c                	      s,   dt  dt  dt  dt   S )Nzi_foreach_addc*_ op expects arguments of type: List[Tensor], List[Tensor], List[Tensor], tensor, but got: r  rm  rN  r*   r  rP   ro  rp  r*   r+   r>     s   z-meta__foreach_addcop_tensor.<locals>.<lambda>r   c                   S   rI   rr  r*   r*   r*   r*   r+   r>     rJ   c                   S   rI   rs  r*   r*   r*   r*   r+   r>     rJ   )r/   rA   r  r
  r
   r   )rP   ro  rp  r  r*   r~  r+   meta__foreach_addcop_tensor  s   
r  c                 C   r_  r%   r\  )rP   r   r   r*   r*   r+   meta__foreach_copy_inplace  s   r  )
grad_scale	found_infc       	            s4   | |||||fD ] t t t fdd qd S )Nc                      rL  ry  rN  r*   rj  r*   r+   r>     r)  z#meta__fused_adam_.<locals>.<lambda>rO  )rP   gradsexp_avgsexp_avg_sqsmax_exp_avg_sqsstate_stepslrbeta1beta2weight_decayepsamsgradmaximizer  r  r*   r  r+   meta__fused_adam_  s   
r  c       	            sZ   | |||||fD ] t t t fdd qdd }|| ||||||||fS )Nc                      rL  ry  rN  r*   r  r*   r+   r>     r)  z"meta__fused_adam.<locals>.<lambda>c                 S   s   dd | D S )Nc                 S   rS  r*   rT  )rs   r%  r*   r*   r+   ru     r  z=meta__fused_adam.<locals>.empty_like_list.<locals>.<listcomp>r*   )tensor_listr*   r*   r+   empty_like_list  s   z)meta__fused_adam.<locals>.empty_like_listrO  )rP   r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r*   r  r+   meta__fused_adam  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   rI   )Nza must be a 2D tensorr*   r*   r*   r*   r+   r>   #  rJ   zmeta__int_mm.<locals>.<lambda>c                   S   rI   )Nzb must be a 2D tensorr*   r*   r*   r*   r+   r>   $  rJ   c                      rD   )Nzexpected self to be int8, got rE   r*   )rs  r*   r+   r>   '  rH   c                      rD   )Nzexpected mat2 to be int8, got rE   r*   )rt  r*   r+   r>   +  rH   r   r   c                
      rH  )Nz'Incompatible matrix sizes for _int_mm (r   rx   r   rV   r/  rW   r*   rr  r*   r+   r>   /  rI  rE   )r/   rA   rT   r7   r/  rX   rN   rZ  rr  r*   rr  r+   meta__int_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: DrS   r*   )x1r*   r+   r>   ;  r?   z$meta_cdist_forward.<locals>.<lambda>c                      r  )Nz1cdist only supports at least 2D tensors, X2 got: r  rS   r*   )x2r*   r+   r>   ?  r?   rR   c                      rF  )Nz4X1 and X2 must have the same number of columns. X1: rR   z X2: rW   r*   )r  r  r*   r+   r>   C  rE  c                   S   rI   )Nz=cdist only supports floating-point dtypes, X1 got: {x1.dtype}r*   r*   r*   r*   r+   r>   G  rJ   c                   S   rI   )Nz=cdist only supports floating-point dtypes, X2 got: {x2.dtype}r*   r*   r*   r*   r+   r>   K  rJ   r   c                   S   rI   )Nz)cdist only supports non-negative p valuesr*   r*   r*   r*   r+   r>   M  rJ   )Nr   r   c                      r:  )Nz%possible modes: None, 1, 2, but was: r*   r*   )compute_moder*   r+   r>   P  r<  r  )r/   rA   rT   rX   r   is_float_dtyper7   rO   r   broadcast_shapesextendrN   )	r  r  r:  r  r1r2batch_tensor1batch_tensor2r   r*   )r  r  r  r+   meta_cdist_forward7  s@   









r  c                 C   s   |j d }|j d }|j d }|j d d }|j d d }	tt||	}
|
d d  }|||g t|
}|dksG|dksG|dksG|dkrLt|S |t|j krX||}tj	|tj
dS )NrR   r  r   r   )rO   r   r/   r  r  mathprod
zeros_liker  rc   r   )r<  r  r  r:  cdistc1r  r  r  r  r  tensor1_expand_sizebatch_productr*   r*   r+   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}
t	d\}}}d urt ||kd	d  t jjkf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rʈ  d}  }||krĈ |	d}nT d}nN||
|}|||fv s|s d}nd}|	}jd }||kr|rt |dkdd  |d8 }|jd }n| }|
|||fS )Nc                      rD   )Nz(expected indices to be long or int, got rE   r*   rg   r*   r+   r>     rH   z$meta_embedding_bag.<locals>.<lambda>c                      rD   )Nz(expected offsets to be long or int, got rE   r*   )offsetsr*   r+   r>     rH   c                      rD   )Nz/expected weight to be floating point type, got rE   r*   )r  r*   r+   r>     rH   r   r   c                   S   rI   Nz1include_last_offset: numBags should be at least 1r*   r*   r*   r*   r+   r>     rJ   rU   c                   S   rI   )Nz@embedding_bag: per_sample_weights only supported with mode='sum'r*   r*   r*   r*   r+   r>     rJ   c                      r  )Nzexpected weight (z) and per_sample_weights (z) to have same dtyperE   r*   )per_sample_weightsr  r*   r+   r>     r  c                      rN  )Nz1expected per_sample_weights to be 1D tensor, got r  rP  r*   )r  r*   r+   r>     r)  c                      rh  )Nz%expected per_sample_weights.numel() (z$ to be the same as indices.numel() (r/  rj  r*   )rg   r  r*   r+   r>     s   c                    s    | ||o| ddkS Nr   r   r   r   r^  r   padding_idx)is_fast_path_index_selectr*   r+   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 r   )r7   r/   r3   r1   r   )r   r   r  r*   r*   r+   r    s   z5meta_embedding_bag.<locals>.is_fast_path_index_selectc                    s"   |d ur| |||S  | ||S r%   r*   r  )r  r  r*   r+   is_fast_path  s   z(meta_embedding_bag.<locals>.is_fast_pathcpuc                   S   rI   r  r*   r*   r*   r*   r+   r>     rJ   )r/   rA   r7   rK   rr  r   r  rX   rN   r   r[   rM   r  rO   )r  rg   r  scale_grad_by_freqr  sparser  include_last_offsetr  num_bagsr   MODE_SUM	MODE_MEANMODE_MAXr  
offset2bagbag_sizemax_indicesfast_path_sumnumBagsr*   )rg   r  r  r  r  r  r+   meta_embedding_bagr  s~   










r  c                 G   sB   t | ||g|R  \}}}}t|dkr|| }||||fS )Nr  )r  r  rN   rX   )r  rg   r  argsr   r  r  r  r*   r*   r+   meta_embedding_bag_forward_only  s   r  c                 C   s.   |r|S | j js| j jr| j S |rtjS | j S r%   )r7   r   r   r/   rK   )r   r7   promote_int_to_longr*   r*   r+   _get_reduction_dtype  s   r  rE   c                C   s6   t | |dd}t| j|}t| ||}| j||dS )NT)r  rE   )r  r   r   rO   r   rN   )r   rx  r   r7   output_dtyper   r*   r*   r+   meta_nansum  s   r  c                 C   s$   t | jtt|  }| |S r%   )r   ry  rO   r@   r   rT   rN   )r   r   r*   r*   r+   meta_median  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 outputrE   )	r  r   alert_not_deterministicr   rO   r   rN   r/   rK   )r   rT   r   r   r*   r*   r+   meta_median_mode_dim  s   
r  c                 C   r   r%   r*   rd   r*   r*   r+   meta_logical_not_  r   r  c                    sd   t t|  kdd  t|   }d| t| j   fddttD }| |S )Nc                   S   rI   )NzZNumber of dimensions of repeat dims can not be smaller than number of dimensions of tensorr*   r*   r*   r*   r+   r>   "  rJ   zmeta_repeat.<locals>.<lambda>r   c                    s   g | ]
} | |  qS r*   r*   rv  padded_sizer   r*   r+   ru   )  rE  zmeta_repeat.<locals>.<listcomp>)r/   rA   r   rT   r@   rO   r   rN   )rP   r   num_new_dimensionstarget_sizer*   r  r+   meta_repeat  s   
r  c                 C   r   r%   r*   rd   r*   r*   r+   
meta_zero_-  r   r  c                 C      t |tjrt| j|j | S r%   r
  r/   r
   rC   rO   r)  r*   r*   r+   meta_binop_inplace2  s   r  c                 C   r  r%   r  r]  r*   r*   r+   meta_binop_inplace_alphaC  s   	r  c                 K      t | tjdS Ntype_promotion)r   r   DEFAULT)rP   kwargsr*   r*   r+   
meta_roundQ  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 rE   r*   )r3  rP   r*   r+   r>   [  r?   z#shift_dtype_check.<locals>.<lambda>c                      r  )Nz6: Expected shift value to have an integral dtype. Got rE   r*   r3  r   r*   r+   r>   `  r?   c                      s     d S )Nz): Expected shift value to be an int. Got r*   r*   r  r*   r+   r>   e  r)  )r/   rA   r   ry  r7   r
  r
   r   r3  rP   r   r*   r  r+   shift_dtype_checkX  s   

r  c                 C   L   t d| | t| tjd}|  dkr$t|tjr$tj|j	|j
|jdS |S )Nrshiftr  r   rh   r  r   r   r  rT   r
  r/   r
   rj   rO   ri   r7   rP   rZ   element_wiser*   r*   r+   meta_rshiftsi     r  c                 C   r  )Nlshiftr  r   rh   r  r  r*   r*   r+   meta_lshiftsw  r  r  c                 C      |  | jS r%   r  rd   r*   r*   r+   	meta_zero     r  c                 C   r   r%   r*   rP   r   r*   r*   r+   
meta_fill_  r   r  c                 C   
   t | S r%   rT  r  r*   r*   r+   	meta_fill     
r  c                 C   r   r%   r*   rd   r*   r*   r+   
meta_relu_  r   r  c                 C   r  r%   rT  rP   rg   rf   
accumulater*   r*   r+   meta_index_put  r  r  c                 C   s   t | j|j | S r%   )rC   rO   )rP   rf  valuer*   r*   r+   meta_masked_fill_  s   r  c                 C   r   r%   r*   r  r*   r*   r+   meta_index_put_  r   r  c                 C   r  r%   )viewrO   rd   r*   r*   r+   
meta_alias  r  r  c                    s   t |  dkdd  t | dkdd  |  }|  |d |d |d } d }||ft  d koB d k fdd |}|sqd urqt  dkd	d  t  kfd
d |S )NrU   c                   S   rI   rA  r*   r*   r*   r*   r+   r>     rJ   z)common_meta_baddbmm_bmm.<locals>.<lambda>c                   S   rI   rB  r*   r*   r*   r*   r+   r>     rJ   r   r   r   c                	      rF  rG  r*   r*   rI  r*   r+   r>     s    c                   S   rI   )Nzself must be a 3D tensorr*   r*   r*   r*   r+   r>     rJ   c                      s   d  d   S )Nz*Expected an input tensor shape with shape z but got shape: rW   r*   )r  self_baddbmmr*   r+   r>     rx  )r/   rA   rT   rX   rN   )rC  rD  is_bmmr  rP  res_rowsres_colsr   r*   )rJ  rK  rL  r  r  r+   common_meta_baddbmm_bmm  s*   


r  c                 C   s   t | |dS )NT)r  )rP   mat2r*   r*   r+   meta_bmm  r  r  c                 C   s<   | | }| | }|dkrt |dk t |dk kr|d8 }|S r  )rY  )rx   yqr  r*   r*   r+   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
kernelSizer  r  r   r  r  
outputSizer*   r*   r+   pooling_output_shape_pad_lr  s*   
	r
  c                    s^   t |dkdd  t dkfdd t  d k fdd t|  |||S )Nr   c                   S   rI   )Nzstride should not be zeror*   r*   r*   r*   r+   r>     rJ   z&pooling_output_shape.<locals>.<lambda>c                      r:  )Nz'pad must be non-negative, but got pad: r*   r*   )padr*   r+   r>     r<  r   c                      r:   )Nz7pad should be at most half of kernel size, but got pad=z and kernel_size=r*   r*   r  r  r*   r+   r>     r?   )r/   rA   r
  )r  r  r  r   r  r  r*   r  r+   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   rI   )NzCkernel size should be greater than zero, but got kH: {kH}, kW: {kW}r*   r*   r*   r*   r+   r>     rJ   z$pool2d_shape_check.<locals>.<lambda>c                   S   rI   )Nz>stride should be greater than zero, but got dH: {dH}, dW: {dW}r*   r*   r*   r*   r+   r>     rJ   c                   S   rI   )Nz\dilation should be greater than zero, but got dilationH: {dilationH}, dilationW: {dilationW}r*   r*   r*   r*   r+   r>     rJ   r   r   r  rU   c                   S   rI   )NzExpected 4D (batch mode) tensor expected for input with channels_last layout with optional 0 dim batch size for input, but got: {input.size()}r*   r*   r*   r*   r+   r>   %  rJ   c                         d    S )NzYExpected 3D or 4D (batch mode) tensor with optional 0 dim batch size for input, but got: rW   r*   r  r*   r+   r>   ,  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 = r*   r*   )r  r  r  r  r*   r+   r>   1  s    c                      s*   d d  d d d d dS NzGiven input size: (rx   z). Calculated output size: (z). Output size is too smallr*   r*   )r  r  r  r  r  r  r*   r+   r>   7  s    )rT   r/   rA   rX   r  )r   r  r  r  r  r  r  	dilationH	dilationWr  r  r  r  r  r   r[   
valid_dimsr*   )r   r  r  r  r  r  r  r  r  r  r  r+   r    sB   

r  r  r  r  r  r  r  r  pTpHpW	dilationTr  r  r  r  r  r  r  r  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: r*   r*   )r  r  r  r*   r+   r>   Y     z$pool3d_shape_check.<locals>.<lambda>c                      r  )Nz0stride should be greater than zero, but got dT: z, dH: z, dW: r*   r*   )r  r  r  r*   r+   r>   `  r  c                      r  )Nz9dilation should be greater than zero, but got dilationT: z, dilationH: z, dilationW: r*   r*   )r  r  r  r*   r+   r>   g  r  r  c                      r  )Nz/: Expected 4D or 5D tensor for input, but got: rv  r*   )r3  r   r*   r+   r>   o  r?   r,  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!)rO   rX   r*   r3  r   r   r*   r+   r>   x  s
   c                      s*   d d  d d d d dS )Nzinput image (T: r0  r   z ) smaller than kernel size (kT:  kH:  kW: r/  r*   r*   )r  r  r  r  r  r  r*   r+   r>     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: r  r  z padT: z padW: z padH: r*   r*   )r  r  r  r  r  r  r*   r+   r>     s   r   c                      s6   d d d  d d d d d dS r  r*   r*   )r  r  r  r  r  r  r  r*   r+   r>     s   )r[   r/   rA   r   rX   )r   r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r3  r  r[   r*   )r  r  r  r  r  r  r3  r   r  r   r  r  r  r  r  r  r  r  r  r  r  r  r+   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  rU   r   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  r3  r[   r*   r*   r+   max_pool3d_backward_shape_check  s@   r  c                 C   s   | j }t| ||||||||	|
|ddd|||||||d t|||d | t|||d | t|||d | t|||d | d S )Nr   Tr  rU   r   r  )r   r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r3  r[   r*   r*   r+   r    s:   r  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  r*   r*   r  r*   r+   r>     rH   zEmax_pool2d_checks_and_compute_shape.<locals>.unpack.<locals>.<lambda>r   r   r  r  r*   r  r+   r    r  z3max_pool2d_checks_and_compute_shape.<locals>.unpackr  r  c                   S   rI   )NzOmax_pool2d: stride must either be omitted, a single int, or a tuple of two intsr*   r*   r*   r*   r+   r>     rJ   z5max_pool2d_checks_and_compute_shape.<locals>.<lambda>r   r   r  r  r  r  rR   r  c                   S   rI   )NzMnon-empty 4D (batch mode) tensor expected for input with channels_last layoutr*   r*   r*   r*   r+   r>   0  rJ   r  c                   S   rI   )Nz9non-empty 3D or 4D (batch mode) tensor expected for inputr*   r*   r*   r*   r+   r>   5  rJ   Fc                   S   rI   )Nz?Unsupport memory format. Supports only ChannelsLast, Contiguousr*   r*   r*   r*   r+   r>   :  rJ   )r/   rA   r   rX   r   r  r  rT   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+   #max_pool2d_checks_and_compute_shape  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 rE   r*   r	  r*   r+   r>   j  rx  z7meta_max_pool2d_with_indices_backward.<locals>.<lambda>c                    s:   t | d   t | d  t | d  d S )NrU   r   r   )r  )r%  )r  r[   r  r  r*   r+   _check_dim_sizep  s   z>meta_max_pool2d_with_indices_backward.<locals>._check_dim_sizer  )
r  r/   rA   r7   r[   r   r  rj   rO   ri   )r  rP   r  r   r  r  r  rg   r  r  r   r*   )r  r  r[   r  r  rP   r+   %meta_max_pool2d_with_indices_backwardU  s.   

r  r  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 )Nr  r  r   rU   r  )
r  rT   rX   r   r  r/   rj   r7   ri   rk   )r   r  r   r  r  r  r  r  r  r  r   rX   r*   r*   r+   meta_max_pool2d_with_indices  s2   
r   c           	         s  t d tjtjkfdd ttdkfdd \}}tjdv fdd tjjkfdd t	d	jD ] t
 d
k fdd qG }jdkrr|
d
}||||f}|S |
d
}|
d	}|||||f}|S )Nmax_unpooling2d_forward_outc                      rD   )Nz2elements in indices should be type int64 but got: rE   r*   r  r*   r+   r>     rH   z#meta_max_unpool2d.<locals>.<lambda>r   c                      ra  )NzMThere should be exactly two elements (height, width) in output_size, but got 
 elements.r  r*   r  r*   r+   r>        r  c                      rN  )NzLInput to max_unpooling2d should be a 3d or 4d Tensor, but got a tensor with  dimensions.rP  r*   )self_r*   r+   r>     rt  c                      r`  NzBExpected shape of indices to be same as that of the input tensor (z%) but got indices tensor with shape: rv  r*   )rg   r&  r*   r+   r>     rg  r   r   c                      r  )NzZmax_unpooling2d(): Expected input to have non-zero size for non-batch dimensions, but got r   being empty.rv  r*   )r   r&  r*   r+   r>     s
   rU   )r   r  r/   rA   r7   rk   r   r[   rO   r   rX   rn   rN   )	r&  rg   r  r  r  rP   	nchannelsr2  r  r*   )r   rg   r  r&  r+   meta_max_unpool2d  s@   






	



r*  c                    s  t jt jkdd  t jdv fdd t tdkfdd t tdkfdd t tdkfdd t jjkfd	d td
jD ]t dk fdd qXt d dkod
 dkod dkfdd d S )Nc                   S   rI   )Nz(elements in indices should be type int64r*   r*   r*   r*   r+   r>     rJ   z._max_unpooling3d_shape_check.<locals>.<lambda>r  c                      rN  )NzLInput to max_unpooling3d should be a 4d or 5d Tensor, but got a tensor with r%  rP  r*   r  r*   r+   r>     r)  rU   c                      ra  )NzVThere should be exactly three elements (depth, height, width) in output_size, but got r"  r  r*   r#  r*   r+   r>     r$  c                      ra  )NzRThere should be exactly three elements (depth, height, width) in stride, but got: r"  r  r*   r  r*   r+   r>     r?   c                      ra  )NzSThere should be exactly three elements (depth, height, width) in padding, but got: r"  r  r*   )r  r*   r+   r>     r?   c                      r`  r'  rv  r*   )rg   r   r*   r+   r>     rg  r   r   c                      s     dj  d dS )NzI: Expected input to have non-zero size for non-batch dimensions, but got r  r(  rv  r*   r  r*   r+   r>     s
   r   c                      r:  )Nz5strides should be greater than zero, but got stride: r*   r*   r  r*   r+   r>     r<  )	r/   rA   r7   rk   r[   r   rO   r   rX   )r   rg   r  r   r  r3  r*   )r3  r   rg   r   r  r  r   r+   _max_unpooling3d_shape_check  s@   







	"
r+  c                 C   s   t d t| ||||d |  }|\}}}| jdkr,|d}	||	|||f}
|
S |d}|d}	|||	|||f}
|
S )Nmax_unpooling3d_forward_outzmax_unpooling3d()r  r   r   )r   r  r+  rn   r[   rX   rN   )r&  rg   r  r   r  rP   odepthr  r  r)  r2  r  r*   r*   r+   meta_max_unpool3d  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   rI   NzMmax_pool3d: kernel_size must either be a single int, or a tuple of three intsr*   r*   r*   r*   r+   r>   /  rJ   z.meta_max_pool3d_with_indices.<locals>.<lambda>r   r   r   c                   S   rI   NzQmax_pool3d: stride must either be omitted, a single int, or a tuple of three intsr*   r*   r*   r*   r+   r>   7  rJ   c                   S   rI   NzImax_pool3d: padding must either be a single int, or a tuple of three intsr*   r*   r*   r*   r+   r>   ?  rJ   c                   S   rI   NzJmax_pool3d: dilation must be either a single int, or a tuple of three intsr*   r*   r*   r*   r+   r>   G  rJ   r  c                   S   rI   r  r*   r*   r*   r*   r+   r>   O  rJ   r,  r  r  r  rR   zmax_pool3d_with_indices()r  r   rE   )r/   rA   r   r[   rX   r  r  r   r  channels_last_3dr  r  rN   rk   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  input_channels_last_checkr^   r   rg   r*   r*   r+   meta_max_pool3d_with_indices#  s   

  







r6  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   rI   r/  r*   r*   r*   r*   r+   r>     rJ   z7meta_max_pool3d_with_indices_backward.<locals>.<lambda>r   r   r   c                   S   rI   r0  r*   r*   r*   r*   r+   r>     rJ   c                   S   rI   r1  r*   r*   r*   r*   r+   r>     rJ   c                   S   rI   r2  r*   r*   r*   r*   r+   r>     rJ   r  c                   S   rI   r  r*   r*   r*   r*   r+   r>     rJ   r  r  r  rR   z"max_pool3d_with_indices_backward()r,  r  r   )r/   rA   r   r[   rX   r  r   r  r4  r  r  rN   rO   r   )r  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  r  r5  
grad_inputr*   r*   r+   %meta_max_pool3d_with_indices_backward  s   
  









r8  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  r*   r9  r   r*   r+   r>     rc  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 )r   r*   r:  r*   r+   r>     rc  r   c                      r`  )NzZgrid_sampler(): expected grid and input to have same batch size, but got input with sizes  and grid with sizes rv  r*   r:  r*   r+   r>     rc  rR   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[   rO   r*   r:  r*   r+   r>     s   c                      r  )NzYgrid_sampler(): expected input to have non-empty spatial dimensions, but input has sizes r  r  rv  r*   r  r*   r+   r>     r  )r/   rA   ri   r   r  rO   r[   r   )r   r9  r*   )r9  r   r   r+   check_grid_sampler_common  s,   
r<  c                   @   s   e Zd ZdZdZdZdS )GridSamplerInterpolationr   r   r   N)__name__
__module____qualname__BILINEARNEARESTBICUBICr*   r*   r*   r+   r=    s    r=  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 )Nr,  c                      r`  )Nzdgrid_sampler(): expected 5D input and grid with same number of dimensions, but got input with sizes r;  rv  r*   r:  r*   r+   r>   $  s
   z'check_grid_sampler_3d.<locals>.<lambda>c                   S   rI   )Nz<grid_sampler(): bicubic interpolation only supports 4D inputr*   r*   r*   r*   r+   r>   /  rJ   )r/   rA   r[   r=  rC  r  )r   r9  rD  r*   r:  r+   check_grid_sampler_3d!  s   

rE  c           
      C   s:   |d }|rt j|t jd}nd }t j|t jd}	||	fS Nr   r   )r/   r  r   rc   
r  r   r9  rD  padding_modealign_cornersrA  input_requires_gradr7  	grad_gridr*   r*   r+   grid_sampler_2d_backward_meta3  s   
rL  c           
      C   s\   t | | t| || | jd }| jd }|jd }|jd }|jd }	| |||||	fS )Nr   r   r   rU   )r<  rE  rO   rN   )
r   r9  rD  rH  rI  rq  Cout_Dout_Hout_Wr*   r*   r+   grid_sampler_3dF  s   
	




rQ  r7  rK  c           
      C   sP   t || t||| |d }|rtj|tjd}nd }tj|tjd}	||	fS rF  )r<  rE  r/   r  rV  rc   rG  r*   r*   r+   grid_sampler_3d_backwardY  s   
rR  c                 O   s:   | dd }|st|}||d< tj| g|R i |S )Nr7   )r6   r   	get_dtyper/   rj   )rX   r*  r  r  r7   r*   r*   r+   fullq  s
   
rT  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   rI   )Nz9memory format option is only supported by strided tensorsr*   r*   r*   r*   r+   r>     rJ   zzeros_like.<locals>.<lambda>r   r   Tr  )r/   
sparse_coorA   rj   r7   ri   	is_sparsesparse_resize_and_clear_rX   
sparse_dim	dense_dimrT   _coalesced_r"   rc   r   fill_)rP   r7   r   ri   r   r   r  r*   r*   r+   r  {  s:   
	

	r  c                    s     }t|dkdd   dkr n |   }t |kp'|k  fdd dkr7n| t }t } |    }| = | = |||S )Nr   c                   S   rI   )Nz-select() cannot be applied to a 0-dim tensor.r*   r*   r*   r*   r+   r>     rJ   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 rW   r*   rT   rG   rP   r*   r+   r>     s
    )rT   r/   rL   rX   r   r   r   r   )rP   rT   rG   r[   rX   new_sizer   new_storage_offsetr*   r\  r+   meta_select  s$   
r_  c                 C   r  r%   r   r%  )rP   r   rT   rG   r*   r*   r+   meta_select_scatter  r  ra  c                 C   r  r%   r`  )rP   r   rT   startendstepr*   r*   r+   meta_slice_scatter  r  re  dim_post_exprwrap_scalarc                 C   sb   |dkr
|sJ d}| }|d }| |k s| |kr'J d|  d| d| d| dk r/| |7 } | S )Nr   r   zdim z out of bounds (r  r/  r*   )rT   rf  rg  r  r	  r*   r*   r+   rl     s   ,rl   c                 C   s   |   dkrdS | j| S r  r  )r%  rT   r*   r*   r+   ensure_nonempty_size  s   rh  c                    st   t  d}t  d}t||kdd  t|D ] kr7tttk fdd qd S )Nr   c                   S   rI   )NzDIndex tensor must have the same number of dimensions as input tensorr*   r*   r*   r*   r+   r>     rJ   z$gather_shape_check.<locals>.<lambda>c                      s$   d dj  dj  d   S )Nz!Size does not match at dimension z expected index  to be smaller than self  apart from dimension rv  r*   rT   r   rG   rP   r*   r+   r>     s    )r	  rT   r/   rA   r   rh  )rP   rT   rG   	self_dims
index_dimsr*   rk  r+   gather_shape_check  s   rn  c                    sR   t ||  }  dk}|s#t jtjk fdd t| |  |  j	S )Nr   c                      rD   )Nz2gather(): Expected dtype int64 for index, but got rE   r*   rF   r*   r+   r>     rH   zmeta_gather.<locals>.<lambda>)
rl   rT   rM   r/   rA   r7   rK   rn  rN   rO   )rP   rT   rG   sparse_gradwrapped_dimis_index_emptyr*   rF   r+   meta_gather  s   

rr  c                 C   s   |r*| dkrdS | dkrdS | dkrdS | dkrdS | d	kr d
S t ddd  d S | dkr0dS | dkr6dS t ddd  d S )Nsum
REDUCE_ADDr  REDUCE_MULTIPLYmeanREDUCE_MEANamaxREDUCE_MAXIMUMaminREDUCE_MINIMUMFc                   S   rI   )Nz=reduce argument must be either sum, prod, mean, amax or amin.r*   r*   r*   r*   r+   r>     rJ   z#get_operator_enum.<locals>.<lambda>addmultiplyc                   S   rI   )Nz/reduce argument must be either add or multiply.r*   r*   r*   r*   r+   r>     rJ   r  )reduce_use_new_optionsr*   r*   r+   get_operator_enum  s,   r  c                    sT   |  dkrt|jtjk fdd |d ur(t|j|jk fdd d S d S )Nr   c                      
     dS )Nz"(): Expected dtype int64 for indexr*   r*   method_namer*   r+   r>     r<  z,scatter_gather_dtype_check.<locals>.<lambda>c                      r  )Nz0(): Expected self.dtype to be equal to src.dtyper*   r*   r  r*   r+   r>   $  r<  )rM   r/   rA   r7   rK   )r  rP   rG   src_optr*   r  r+   scatter_gather_dtype_check  s   



r  c                 C   s
   t | dS r   )r	  rS   r*   r*   r+   ensure_nonempty_dim(  s   
r  c                    s     dkrd S tt t kdd  d}t }t|D ]}t|}| kr2q&|t|kr=d} nq&|s[d ur[t|D ]}t|}|t|krZd} nqHd urtt t kdd  t|  fdd d S t|  fdd d S )	Nr   c                   S   rI   NzCIndex tensor must have the same number of dimensions as self tensorr*   r*   r*   r*   r+   r>   2  rJ   z%scatter_shape_check.<locals>.<lambda>FTc                   S   rI   r  r*   r*   r*   r*   r+   r>   L  rJ   c                      s&   dj  dj  d  dj   S )NExpected index ri  rj  z and to be smaller than src rv  r*   rT   rG   rP   r  r*   r+   r>   P  s    c                      s   dj  dj  d   S )Nr  ri  rj  rv  r*   r\  r*   r+   r>   V  s    )rM   r/   rA   r  rT   r   rh  )rP   rT   rG   r  is_wrong_shaperl  r   index_d_sizer*   r  r+   scatter_shape_check-  sH   

r  c                 C   sD   t ||  }td| || t| ||| |d ur t|| d S d S )Nscatter)rl   rT   r  r  r  )rP   rT   rG   r   r~  r  rp  r*   r*   r+   scatter_meta_impl\  s   r  c                 C   s   t | |||d | | jS Nr|  r  rN   rO   rP   rT   rG   r   r*   r*   r+   meta_scatter_adde  s   r  c                 C   s   t | |||d | S r  r  r  r*   r*   r+   meta_scatter_add_k  r  r  c                 C   s0   t |tjr|nd }t| |||| | | jS r%   )r
  r/   r
   r  rN   rO   rP   rT   rG   src_or_valuer   r   r*   r*   r+   meta_scatterq  s   
r  c                 C   s(   t |tjr|nd }t| |||| | S r%   )r
  r/   r
   r  r  r*   r*   r+   meta_scatter_  s   	r          queryr}   r  	dropout_p	is_causalreturn_debug_maskr^  c                 C   s   |  d}|  d}|  d}	|  d}
| d}||	 }| dd}||||
}tj|| jd}|||	||
dd}t| dkrtj||	|ftj	| jddd}||tjdtj
d	dtjdtj
d	dddtjdtjd	dtjdtjd	dtjd| j| jdf	S t|	d
 d
 }tj|||ftj	| jd}tj|d tj
d	d}tj|d tj
d	d}|r|
dkrdnd}t|	| }|dkrd}n|dkrd}tj||||f| j| jd}n
tjd| j| jd}|||||	|tjdtjd	dtjdtjd	d|f	S )Nr   r   r   rU   r  r  r7   ri   r*   r     @         )rX   rK  r   r/   rc   ri   r  r  rj   r3   rZ  rK   r7   r  ceil)r  r}   r  r  r  r  r^  r   	num_headsmax_seqlen_batch_qhead_dimmax_seqlen_batch_kNnz_qquery_tquery_reshaped	attention	logsumexpmax_seqlen_qcumulative_sequence_length_qcumulative_sequence_length_kblocksize_cmax_seqlen_k
debug_maskr*   r*   r+   meta__scaled_dot_product_flash  s   






r  r  r  	cum_seq_q	cum_seq_kmax_qmax_kphilox_seedphilox_offsetc                 C   s   | d}| d}| d}t|dkr| dn|}t|dkr'| dn|	}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   rU   r  r   r   r   r   rU   r  )rX   r  r/   empty_permutedr7   ri   )r  r  r}   r  r   r  r  r  r  r  r  r  r  r  r^  r   r  r  len_qlen_kgrad_qgrad_kgrad_vr*   r*   r+   'meta__scaled_dot_product_flash_backward  s0   






r  	attn_biascompute_log_sumexpc                 C   s   |  dd} | dd}| dd}| d}| d}	|d}
| d}| d}|d}tj||	||| j| jd}|rHt|	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  rR   r      r*   r  )
rK  rX   r/   rj   r7   ri   r  r  r3   rK   )r  r}   r  r  r  r  r  r^  r.  Mrq  r  KKvr  logsumexp_dim
logsum_expseedoffsetr*   r*   r+   "meta__scaled_dot_product_efficient   s(   





r  grad_input_maskc                 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}d }|d ur_|
d r_tj|  | |j|jd}||||fS )Nr   r   r   rU   r  r  )rX   r/   r  r7   ri   ri  r   )r  r  r}   r  r  r   r  r  r  r  r  r  r^  r   r  r  r  r  r  r  r  	grad_biasr*   r*   r+   +meta__scaled_dot_product_efficient_backwardL  s@   







r  c                 C   s    t | ||||dd | | jS NT)r  r  rP   rT   rG   r   r   r   r*   r*   r+   meta_scatter_reduce_two  s   r  c                 C   s   t | ||||dd | S r  r  r  r*   r*   r+   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                      r  )Nz@The probabilty distributions dimensions must be 1 or 2, but got rS   r*   r  r*   r+   r>     r)  z"meta_multinomial.<locals>.<lambda>r   r  )r/   rA   rT   rj   rK   ri   rX   )r   num_samplesreplacementr   r*   r  r+   meta_multinomial  s   
r  c                 C   s   d}| D ]}||9 }q|S r   r*   )vsr  vr*   r*   r+   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  r*   )num_spatial_dimsr  r*   r+   r>     rx  z'upsample_common_check.<locals>.<lambda>r   c                      r  )Nz$It is expected input_size equals to r  r  r*   )expected_input_dimsr  r*   r+   r>     rx  c                 s       | ]}|d kV  qdS r   Nr*   rU  r*   r*   r+   r    r  z(upsample_common_check.<locals>.<genexpr>c                      r  )NzDInput and output sizes should be greater than 0, but got input size z and output size r*   r*   )r  r  r*   r+   r>     s
    )r/   rA   r   r  )r  r  r  r  channelsr*   )r  r  r  r  r+   upsample_common_check  s   

*r  c                    sZ   t   dkpt  dd   fdd t  |dd} |jt	 dS )Nr   r   c                      r  )Nz>Non-empty 3D data tensor expected but got a tensor with sizes rW   r*   r  r*   r+   r>     r)  z$upsample_nearest1d.<locals>.<lambda>r  r   
r/   rA   rM   r  rX   r  rN   r   r   r  )r   r  scalesfull_output_sizer*   r  r+   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                      r  Nz>Non-empty 4D data tensor expected but got a tensor with sizes rW   r*   r  r*   r+   r>     r)  z$upsample_nearest2d.<locals>.<lambda>r   r  r  r  r   )r/   rA   rM   r  rX   r  rN   r   r  rO   ri   r~  r   rn   )	r   r  scales_hscales_wr  r   r   rt   
n_channelsr*   r  r+   upsample_nearest2d  s   



r  r  r  r  r  c                    st   t ||dd tjdkfdd tdD ]t  k fdd q|jt	dS )Nr   r  r  c                      rD   )NzFExpected grad_output to be a tensor of dimension 4 but got: dimension rP  r*   r  r*   r+   r>     rH   z-upsample_nearest2d_backward.<locals>.<lambda>c                
      s&   d d   d d  S )NzCExpected grad_output to have the same shape as output; output.size(z) = z but got grad_output.size(rW   r*   r  r  r   r*   r+   r>     s   r   )
r  r/   rA   r[   r   rX   rN   r   r   r  )r  r  r  r  r  r*   r  r+   upsample_nearest2d_backward  s   

	r  c                    sZ   t   dkpt  dd   fdd t  |dd} |jt	 dS )Nr   r   c                      r  )Nz>Non-empty 5D data tensor expected but got a tensor with sizes rW   r*   r  r*   r+   r>     r)  z$upsample_nearest3d.<locals>.<lambda>rU   r  r   r  )r   r  scales_dr  r  r  r*   r  r+   upsample_nearest3d   r  r  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 )NrE   )r  r  )
r/   rc   rk   r
  r   rO   r   r   r   r   )
rP   stablerT   
descendingrf   rg   r  r   r^   
out_strider*   r*   r+   	meta_sort  s   	

r  )rT   r  c                C   s   t | |||dd S )N)r  rT   r  r   )r  )rP   r  rT   r  r*   r*   r+   meta_argsort)  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 != 2rP  r*   input_gatesr*   r+   r>   1  rH   z%rnn_cell_checkSizes.<locals>.<lambda>c                         j  d j  S N != rv  r*   )hidden_gatesr  r*   r+   r>   4      r   c                      r  )Nz != 1rP  r*   )
input_biasr*   r+   r>   8  rH   c                      s      d  S r  rj  r*   )
gates_sizer  r*   r+   r>   ;  r  c                      r  r  rv  r*   )hidden_biasr  r*   r+   r>   ?  r  c                      r  r  rP  r*   )prev_hiddenr*   r+   r>   A  rH   r   c                
      s,      dd d d d  d
S )Nr  r   z * z // z (aka r/  )rM   rX   r*   )expected_prev_hidden_numelfactorr   r  r  r*   r+   r>   E  s   , c                 3   s    | ]	}|j  j kV  qd S r%   r  r  r  r*   r+   r  H  s
    

z&rnn_cell_checkSizes.<locals>.<genexpr>c                   S   rI   )Nz%expected all inputs to be same devicer*   r*   r*   r*   r+   r>   L  rJ   )r/   rA   r[   rO   rX   rM   r  )r  r  r  r  r  r  r*   )r  r  r   r  r  r  r  r  r+   rnn_cell_checkSizes.  s8   





r  c                 C   sL   t | |||d| tj| tjd}tj|tjd}tj|tjd}|||fS )Nr  r   )r  r/   rc   r   )r  r  cxr  r  	workspacehycyr*   r*   r+   _thnn_fused_lstm_cell_metaP  s
   
r
  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   rR   r   r  rE   )r   rO   rN   r/   rj   ri   uint8)r   r  weight_stride0
weight_bufhxr  r  hidden_size	proj_size
num_layersbatch_firstdropouttrainbidirectionalbatch_sizesdropout_stateis_input_packed
seq_length
mini_batchbatch_sizes_sumnum_directionsout_sizer^   r   
cell_shaper	  r  reserve_shapereserver*   r*   r+   
_cudnn_rnn[  s2   

r!  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  rh   )rO   rN   r/   rj   ri   r  )r   w0w1w2w3hx_cx_r~   r  r  r  r  
has_biasesr  r  r  r  r  output_chanelsr^   r   r  r	  r  r*   r*   r+   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   rR   c                      r'  )Nz4: Expected reduction dim -1 or 0 for scalar but got r*   r*   rT   r3  r*   r+   r>     r)  z'zero_numel_check_dims.<locals>.<lambda>c                      r+  )Nz: Expected reduction dim z to have non-zero size.r*   r*   r+  r*   r+   r>     r?   )r[   r/   rL   rX   )rP   rT   r3  r*   r+  r+   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.r*   r*   r  r*   r+   r>     r<  z%check_argmax_argmin.<locals>.<lambda>)rl   rT   r,  r/   rA   rM   )r  rP   rT   r*   r  r+   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argmaxrE   )r-  r   r   rO   r   rN   r/   rk   )rP   rT   r   rx  rO   r*   r*   r+   argmax_argmin_meta  s   r/  c                 C   s   t jd||||dS )Nr*   r   r   )r  r7   r   ri   r   r*   r*   r+   scalar_tensor  r   r0  c                 C   s   t ||  dd}t|dko||  dkr| |ndkdd  |  dkr*dn| |}t|dko8||kdd  t| j}t|dkrL|||< | || j|tj	dfS )	NT)rg  r   r   c                   S   rI   )Nzselected index k out of ranger*   r*   r*   r*   r+   r>     rJ   ztopk_meta.<locals>.<lambda>c                   S   rI   )Nzk not in range for dimensionr*   r*   r*   r*   r+   r>     rJ   rE   )
rl   rT   r/   rA   rX   r   rO   r   rN   rk   )rP   r  rT   largestsorted	sliceSizetopKSizer*   r*   r+   	topk_meta  s   $
r5  c                 C   s   | d ur| n|}t | dkdd  | }| d ur(t |  |kdd  |d ur8t | |kdd  t | |kdd  t | |kdd  t | dkdd  t | |d	 |d
  d kdd  d S )Nr   c                   S   rI   N r*   r*   r*   r*   r+   r>     rJ   z(checkLSTMBackwardSizes.<locals>.<lambda>c                   S   rI   r6  r*   r*   r*   r*   r+   r>     rJ   c                   S   rI   r6  r*   r*   r*   r*   r+   r>     rJ   c                   S   rI   r6  r*   r*   r*   r*   r+   r>      rJ   c                   S   rI   r6  r*   r*   r*   r*   r+   r>     rJ   c                   S   rI   r6  r*   r*   r*   r*   r+   r>     rJ   r   r   r  c                   S   rI   r6  r*   r*   r*   r*   r+   r>     rJ   )r/   rA   rT   rX   rM   )grad_hygrad_cyr  r	  r  defined_gradexp_sizer*   r*   r+   checkLSTMBackwardSizes  s   ,r<  c           	      C   s`   | d u r
|d u r
dS t | |||| tj|td}tj|td}|r)|jdddnd }|||fS )NNNNr   r   F)r   )r<  r/   rc   legacy_contiguous_memory_formatrs  )	r8  r9  r  r	  r  has_bias
grad_gatesgrad_cxr  r*   r*   r+   #_thnn_fused_lstm_cell_backward_impl  s   
rB  c                    s   t jdkrjd ||  dksJ dj d| dd   fdd	}jd ||  }jd
 | }jd | }g jd d |||R }|}|j| d}|S )Nr   r  r   z'Invalid input shape for pixel_shuffle: z with upscale_factor = c                 S   r  r%   r  r  r*   r*   r+   r    r  z,meta_pixel_shuffle.<locals>.is_channels_lastc                      sL    rt dkrtjS tjS jtjdrtjS jtjdr$tjS d S r  )r  r/   r   r  r  r  r*   r  rP   r*   r+   r    s   z.meta_pixel_shuffle.<locals>.pick_memory_formatr  rR   r   )r   rO   rN   r   )rP   upscale_factorr  rM  HrWrr^   r   r*   rC  r+   meta_pixel_shuffle  s   & 
rG  c                 C   sZ   |  | j}| |j}| |j}| |j}| |j}| |j}|||||||fS r%   r  )r   weight0weight1weight2weight3r&  cx_tmpr   hy_cy_grad_output_r_optgrad_hy_r_optgrad_cy_r_optr~   r  r  r  r(  r  r  r  r  r  diff_xdiff_hxdiff_cxdiff_w1diff_w2diff_br*   r*   r+   mkldnn_rnn_layer_backward2  s   rX  )	out_int32r   c                C   s   t j| |rt jnt jd S r   )r/   rc   rZ  rk   rn   )rP   
boundariesrY  r   r*   r*   r+   meta_bucketizeU  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  r*   )rs   rX   r*   r*   r+   r  e  r  z.meta_upsample_bilinear2d_aa.<locals>.<genexpr>r   c                      r  r  rW   r*   r  r*   r+   r>   f  r)  z-meta_upsample_bilinear2d_aa.<locals>.<lambda>r   )
r  rX   r/   rA   rM   r  rN   r   r   r  )r   r  rI  r  r  r  r*   r  r+   meta_upsample_bilinear2d_aa]  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   rI   )Nz%found_inf must be a 1-element tensor.r*   r*   r*   r*   r+   r>   q  rJ   z<_amp_foreach_non_finite_check_and_unscale_.<locals>.<lambda>c                   S   rI   )Nz%inv_scale must be a 1-element tensor.r*   r*   r*   r*   r+   r>   t  rJ   c                   S   rI   )Nz!found_inf must be a float tensor.r*   r*   r*   r*   r+   r>   x  rJ   c                   S   rI   )Nz!inv_scale must be a float tensor.r*   r*   r*   r*   r+   r>   |  rJ   )r/   rA   rM   r7   r   )rP   r  	inv_scaler*   r*   r+   *_amp_foreach_non_finite_check_and_unscale_n  s   r^  c                 C   s   t |  }| |S r%   )r   rX   rN   )rP   nanposinfneginfr   r*   r*   r+   
nan_to_num  s   
rb  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)r   r/   r  
sparse_cscr  
sparse_bscr[   rl   r   rX   r   r   )rP   dim0rM  ndimsrX   r   r*   r*   r+   r    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[   rV  rX  rY  rT   r  )rP   rf  rX  rY  r*   r*   r+   t_  s   
rg  )rY  r   sidesorterc                C   s@   |rt jnt j}t|t jrt j||d S t jd|| jdS )NrE   r*   r  )	r/   rZ  rk   r
  r
   rc   rn   rj   ri   )sorted_sequencerP   rY  r   rh  ri  r7   r*   r*   r+   meta_searchsorted  s   rk  r   c                 C   s4   t | dkdd  t|tjd\}}t j||dS )Nr   c                   S   rI   )Nz,polygamma(n, x) does not support negative n.r*   r*   r*   r*   r+   r>     rJ   z meta_polygamma.<locals>.<lambda>r   rE   )r/   rA   r   r   r   rc   )r   rP   rt   r   r*   r*   r+   meta_polygamma  s   
rl  c                 C      t | t dd }|S )Nc                 S   r  r  )r   r   r   rw   r*   r*   r+   _f  s   z)_create_unary_float_meta_func.<locals>._fr.   r   funcrn  r*   r*   r+   _create_unary_float_meta_func  s   rr  c                 C   rm  )Nc                 S   s   t | |\} }t| |tjdS r  )r   r   r   r   )rx   r  r*   r*   r+   rn    s   z*_create_binary_float_meta_func.<locals>._fro  rp  r*   r*   r+   _create_binary_float_meta_func  s   rs  c                  C   s
  i } dD ]}t | }|D ]}|| vr|| | |< qq|  D ]d\}}t|ts)J |tjjj| tj	|
 drJ|t d v rIt| dq|jrNq|
 dv rUqd|
 v rbt|| qd|
 v rot|| qd|
 v r|t|| qt|| qd S )	N)r  post_autogradpre_autogradCompositeImplicitAutogradr  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.>   aten::cloneaten::copy_aten::rot90aten::_to_copyaten::empty_stridedaten::constant_pad_ndaten::as_strided_scatterzmkldnn::zmkl::zonednn::)r   itemsr
  r   py_implr/   _CDispatchKeyr$   %_dispatch_has_kernel_for_dispatch_keyr  r  is_view2_meta_lib_dont_use_me_use_register_meta_for_mkldnnimpl/_meta_lib_dont_use_me_use_register_meta_for_mkl2_meta_lib_dont_use_me_use_register_meta_for_onednn'_meta_lib_dont_use_me_use_register_meta)activate_meta_tabler~  registryopoop_overloadr)   r*   r*   r+   activate_meta  s>   	r  rq   )NNr=  )Tr"  )r2  )r9  T)FF)TT)r  )FTN)TFF)TF)r   )rU  N)r   rz  r%   )r*   r   FTN)Fr   FNFrR   )NF)rR   F)r*   r   r  F)NNNNN)r   NNr   )NNF)r  FFN)r  FN)FN)NrR   FNN)NNNN)rR   TT(  r  enumr   typingr   r   r   r   r   r/   torch._prims_commonr  r   r   r	   r
   torch._decompr   r   r   r   
torch._opsr   torch._primsr   r   r   r   r   r   r   r   r   torch._prims_common.wrappersr   r   r   r   r2  r   r   %torch.fx.experimental.symbolic_shapesr   r    torch.utils._pytreer!   opsr"   libraryLibraryr  r.   r9   rC   taker   r   rQ   r_   re   cummaxcumminrm   ro   r   _fft_c2cr   _fft_r2cr   randpermgenerator_outr   rK   r   randintr   r   r   randr   _fft_c2rr   r   r   r   
unsqueeze_r   index_reducerr  rz  rY  r   index_reduce_r   r   r   r   r	  	unary_outr   rT   r   r  r   r   r   r   r   _assert_asyncr   msgr   _make_dep_tokenr  r  _functional_sym_constrain_ranger  r  (_functional_sym_constrain_range_for_sizer  _functional_assert_asyncr  ra   r$  rb   r-  r1  r6  r?  _linalg_eighr@  rH  rL  rM  rQ  rU  rX  r^  rk  linalg_inv_exrm  linalg_ldl_factor_exrs  linalg_ldl_solver}  	linalg_lur  linalg_lu_factor_exr  linalg_lu_solver  	lu_unpackr  r  	linalg_qrr  r  r  _linalg_svdr  rz  rR  r  r  linalg_solve_triangularr  r  r  _linalg_detr  r  r  r  reflection_pad1dr  replication_pad1dr
  r  reflection_pad1d_backwardr  replication_pad1d_backwardr  r%  reflection_pad2dr'  replication_pad2dr(  reflection_pad2d_backwardr7  replication_pad2d_backwardr+  r4  reflection_pad3dr6  replication_pad3dr7  reflection_pad3d_backwardreplication_pad3d_backwardr9  _pdist_forwardr3   r;  _pdist_backwardr>  baddbmmrQ  	bernoullirT  
bernoulli_rW  r:  rX  _fused_moving_avg_obs_fq_helperrg  rl  r'  rm  mmru  r   r  r  r  convolutionr  r  _has_mkldnnr  r  _convolution_pointwiser  _linear_pointwiser  has_mklr  r  _mkl_linearr  r  r  qconv2d_pointwiser  qlinear_pointwiser  r  
avg_pool2dr  r  avg_pool2d_backwardr  
avg_pool3dr  avg_pool3d_backwardr  _adaptive_avg_pool2dr   _adaptive_avg_pool3dr  _adaptive_avg_pool2d_backwardr  _adaptive_avg_pool3d_backwardr
  r  adaptive_max_pool2dr  r  r  adaptive_max_pool3dr  r  r  repeat_interleaver!  complexr$  r  r7   r&  r(  r+  rG   _unsafe_indexr;  convolution_backwardrE  addbmmrJ  _foreach_neg__foreach_reciprocal__foreach_sqrt__foreach_sign_rP  _foreach_neg_foreach_reciprocal_foreach_sqrt_foreach_signrV  r[  _foreach_add_foreach_sub_foreach_mul_foreach_div_foreach_maximum_foreach_minimumr^  _foreach_add__foreach_sub__foreach_mul__foreach_div__foreach_maximum__foreach_minimum_r`  Scalarre  rf  _foreach_addcdiv__foreach_addcmul_rv  _foreach_lerp_rw  _foreach_addcdiv_foreach_addcmulrx  _foreach_powScalarAndTensorr}  r  _foreach_copy_r  _fused_adam_r  _fused_adamr  _int_mmr  _cdist_forwardr  _cdist_backwardr  _embedding_bagr  _embedding_bag_forward_onlyr  r  nansumr  median	nanmedianr  
dim_valuesr  rf   r  logical_not_r  repeatr  zero_r  mul_div_logical_and_logical_or_logical_xor_r  add_sub_r  rounddecimalsr  r  
__rshift__r  
__lshift__r  zeror  r[  r  fillr  relu_r  	index_put_unsafe_index_putr  masked_fill_r  
index_put_r  aliasr  r  bmmr  r  r
  r  r  r  r  r  r   max_pool2d_with_indices_backwardr  max_pool2d_with_indicesr   max_unpool2dr*  r+  max_unpool3dr.  max_pool3d_with_indicesr6   max_pool3d_with_indices_backwardr8  r<  r=  rE  grid_sampler_2d_backwardrL  rQ  rR  rT  r  r1  r_  select_scatterra  slice_scatterre  rl   rh  rn  gatherrr  r  r  r  r  r  scatter_addr  scatter_add_r  r  r   r  r   value_reducer  scatter_r  #_scaled_dot_product_flash_attentionr  ,_scaled_dot_product_flash_attention_backwardr  '_scaled_dot_product_efficient_attentionr  0_scaled_dot_product_efficient_attention_backwardr  scatter_reducetwotwo_outr  scatter_reduce_r  multinomialr  r  r  r  r  r  typesSymIntr  r   r  values_stabler  argsortr  r  _thnn_fused_lstm_cellr
  r!  r*  r,  r-  r.  argminr/  r0  topkr5  r   r>  r<  rB  pixel_shufflerG  rX  	bucketize
Tensor_outr[  _upsample_bilinear2d_aar\  r^  rb  r  rg  searchsortedrk  	polygammarl  rr  rs  special_airy_aispecial_bessel_y0special_bessel_y1special_modified_bessel_i0special_modified_bessel_i1special_modified_bessel_k0special_modified_bessel_k1!special_scaled_modified_bessel_k0!special_scaled_modified_bessel_k1special_chebyshev_polynomial_tn_scalar_outspecial_chebyshev_polynomial_uspecial_hermite_polynomial_hspecial_hermite_polynomial_hespecial_laguerre_polynomial_ltorch._refs.nn.functionaltorch._refs.specialr  r*   r*   r*   r+   <module>   sP	   $
		6








	







	

	















)


"

2
&
*
7
(
"
%


	
;

/
Z&5
 ?'$,




e
	
'
!
M
,HT
N


.

*&c 














$
#h	







!
T	
]>	
6G+
!7
/

ge( 

	,$/	








	[	
-'	
4


"
7'
"


"	






9