U
    E‹Ñh;!  ã                   @   s’  d dl Z d dlZd dl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 e jG dd„ dƒƒZG dd„ dƒZeedœd	d
„Zeeeef dœdd„Zee	jjdœdd„Ze	jjedœdd„Zeedœdd„Zeedœdd„Zeedœdd„Ze	jjdœdd„Zdd„ Zejeed f eeef eeejef  d!œd"d#„Z e	jjedœd$d%„Z!ed&œd'd(„Z"d)d*„ Z#ejd+œd,d-„Z$ejd+œd.d/„Z%dS )0é    N)ÚAnyÚCallableÚDictÚIterableÚTuple)Ú_Cc                   @   s*   e Zd ZU dZeed< eed< dd„ ZdS )ÚKernelz$Models a (function, source location)ÚfuncÚsourcec                 O   s   | j ||ŽS ©N)r	   )ÚselfÚargsÚkwargs© r   úB/tmp/user/2202/pip-unpacked-wheel-bf9o3bgk/torch/_library/utils.pyÚ__call__   s    zKernel.__call__N)Ú__name__Ú
__module__Ú__qualname__Ú__doc__r   Ú__annotations__Ústrr   r   r   r   r   r      s   
r   c                   @   s,   e Zd ZdZedœdd„Zddœdd„ZdS )	ÚRegistrationHandlez2Does something when someone calls .destroy() on it)Ú
on_destroyc                 C   s
   || _ d S r   ©Z_on_destroy)r   r   r   r   r   Ú__init__   s    zRegistrationHandle.__init__N©Úreturnc                 C   s   |   ¡  d S r   r   )r   r   r   r   Údestroy   s    zRegistrationHandle.destroy)r   r   r   r   r   r   r   r   r   r   r   r      s   r   )Ú
stacklevelr   c                 C   s&   t  t | ¡¡}|j› d|j› �}|S )zÃGet a string that represents the caller.

    Example: "/path/to/foo.py:42"

    Use stacklevel=1 to get the caller's source
    Use stacklevel=2 to get the caller's caller's source
    etc.
    ú:)ÚinspectÚgetframeinfoÚsysÚ	_getframeÚfilenameÚlineno)r   Úframer
   r   r   r   Ú
get_source!   s    	r(   )Úqualnamer   c                 C   s6   |   d¡}t|ƒdkr&td| › d�ƒ‚|d |d fS )Nz::é   zAExpected `qualname` to be of the form "namespace::name", but got zf. The qualname passed to the torch.library APIs must consist of a namespace and a name, e.g. aten::sinr   é   )ÚsplitÚlenÚ
ValueError)r)   Úsplitsr   r   r   Úparse_namespace/   s    

ÿr0   c                 C   sH   t | ƒ\}}d|kr$| d¡\}}nd}ttj|ƒ}t||ƒ}t||ƒS )NÚ.Údefault)r0   r,   ÚgetattrÚtorchÚops)r)   Ú	namespaceÚnameÚoverloadÚnsÚpacketr   r   r   Ú	lookup_op;   s    
r;   )Úopr   c                 C   s   t | tjjƒst‚| jdkS )N>   ÚprimsÚatenÚprim)Ú
isinstancer4   Ú_opsÚ
OpOverloadÚAssertionErrorr6   ©r<   r   r   r   Ú
is_builtinF   s    rE   )Úschemar   c                 C   sT   dd„ }t | tjjƒr|| ƒS ddlm} t | tƒr>| | ¡} t | |ƒsLt‚|| ƒS )zËCheck if the schema is functional.

    An operator is functional if:
    - it does not mutate any of its inputs
    - it does not return a view on any of its inputs
    - it has at least one return
    c                 S   sD   | j r
dS | j}t|ƒdko,tdd„ |D ƒƒ}|r6dS | js@dS dS )NFr   c                 s   s"   | ]}|j d k	o|j j V  qd S r   )Ú
alias_infoÚis_write)Ú.0Úrr   r   r   Ú	<genexpr>X   s    z>is_functional_schema.<locals>.is_functional.<locals>.<genexpr>T)Ú
is_mutableÚreturnsr-   Úany)rF   ZretsZis_non_mutating_viewr   r   r   Úis_functionalT   s    ÿz+is_functional_schema.<locals>.is_functionalr   )ÚFunctionSchema)r@   r4   r   rP   Ztorchgen.modelr   ÚparserC   )rF   rO   rP   r   r   r   Úis_functional_schemaK   s    	

rR   )Útypr   c              	   C   sh   | t  t j ¡ ¡kpf| t  t  t j ¡ ¡¡kpf| t  t  t j ¡ ¡¡kpf| t  t  t  t j ¡ ¡¡¡kS r   )r   ZListTypeÚ
TensorTypeÚgetÚOptionalType©rS   r   r   r   Úis_tensorlist_like_typen   s    ÿþürX   c                 C   s"   | t j ¡ kp | t  t j ¡ ¡kS r   )r   rT   rU   rV   rW   r   r   r   Úis_tensor_like_typex   s    rY   rD   c                 C   sî   | j dkrdS | j}t|jƒdks&dS |jd jdkr:dS |jd jj}t|ƒdkrXdS tt|ƒƒ}t|jƒdk rvdS |jd }|jdkrŽdS |jj	sšdS |jj}t|ƒdkr²dS |tt|ƒƒkrÆdS |jdd… D ]}|jdk	rÔ dS qÔdS )aN  Check if an op is an inplace aten op, i.e. it mutates and returns the first arg.

    TODO: torchgen/model.py's FunctionSchema.parse is the source of truth for this,
    but not all PyTorch builds have torchgen (due to the yaml dependency being weird).
    Figure this out.

    Example: add_(Tensor(a!) x, Tensor y) -> Tensor(a)
    r>   Fr+   r   NT)
r6   Ú_schemar-   rM   rG   Z	after_setÚnextÚiterÚ	argumentsrH   )r<   rF   Z	alias_setÚlocZ	first_argÚargr   r   r   Úmutates_and_returns_first_arg|   s6    	



r`   c                 C   sŠ   g }i }t t| jƒƒD ]f}| j| }|jrT|j|krF||j ||j< q||j||j< q|t|ƒk rp| || ¡ q| |j¡ qt|ƒ|fS r   )Úranger-   r]   Ú
kwarg_onlyr7   Údefault_valueÚappendÚtuple)rF   r   r   Únew_argsZ
new_kwargsÚiÚinfor   r   r   Úfill_defaults¢   s    

ri   .)rF   r   r   r   c                 c   s~   t | jƒt |ƒt |ƒ kst‚tt | jƒƒD ]L}| j| }|jr\|j|kr,|||j fV  q,|t |ƒkrjq,||| fV  q,dS )zÝzips schema.arguments and (args, kwargs) together.

    Assumes that (args, kwargs) were the inputs to some torch._ops.OpOverload:
    that is, kwargs must be keyword-only arguments and default values may be omitted.
    N)r-   r]   rC   ra   rb   r7   )rF   r   r   rg   rh   r   r   r   Ú
zip_schema´   s    

rj   c                 C   sD   t | tjjƒst‚t| ƒrdS | j}|js.dS t|j	ƒdkr@dS dS )NFr   T)
r@   r4   rA   rB   rC   rE   rZ   rL   r-   rM   )r<   rF   r   r   r   Úcan_generate_trivial_fake_implÌ   s    rk   r   c                   C   s   t tddƒS )zðIf an op was defined in C++ and extended from Python using the
    torch.library APIs, returns if we require that there have been a
    m.set_python_module("mylib.ops") call from C++ that associates
    the C++ op with a python module.
    ZREQUIRES_SET_PYTHON_MODULET)r3   Ú_utils_internalr   r   r   r   Úrequires_set_python_moduleÜ   s    rm   c                 O   s~   t | tjjjƒst‚g }tjj || ¡ f¡\}}|D ]6}t |tj	ƒr6tj
 |¡ tj
jj¡r6| t|ƒ¡ q6|  ||||¡S r   )r@   r4   ÚutilsZ_python_dispatchZTorchDispatchModerC   Z_pytreeZtree_flattenÚvaluesZTensorr   Z_dispatch_keysZhasZDispatchKeyÚPythonrd   ÚtypeZ__torch_dispatch__)Z	curr_modeZop_overloadr   r   Zoverload_typesZargs_flattenedÚ_Úar   r   r   Úhandle_dispatch_modeå   s    ÿrt   ©rF   c                 C   s   t dd„ | jD ƒƒS )Nc                 s   s   | ]}|j V  qd S r   )rb   )rI   rs   r   r   r   rK   ø   s     z&has_kwarg_only_args.<locals>.<genexpr>)rN   r]   ru   r   r   r   Úhas_kwarg_only_args÷   s    rv   c                 C   s2   | j D ]&}t|jƒs t|jƒs q|js(q dS dS )NTF)r]   rY   rq   rX   rb   )rF   rs   r   r   r   Úhas_kwarg_only_tensorsû   s    
rw   )&Zdataclassesr!   r#   Útypingr   r   r   r   r   r4   Ztorch._utils_internalrl   r   Z	dataclassr   r   Úintr   r(   r0   rA   rB   r;   ÚboolrE   rR   rX   rY   r`   ri   rP   ZArgumentrj   rk   rm   rt   rv   rw   r   r   r   r   Ú<module>   s8   

#
& 
 
þ	