^j"Z U d Z ddlZddlZddlZddlZddlmZmZ ddlm Z
ddl m Z m
Z
mZmZmZmZmZmZmZmZmZmZmZmZmZmZ ddl Z ddl m Z m Z m!Z!m"Z"m#Z#m$Z$ ddl%m&Z&m'Z'm(Z( e rddl)Z*ddl+Z, G d d
e Z- G d d e Z. ed
ee/e0f e1ed eee/df ee/df ed
de-f d Z2 edee/e0f e1ed eee/df ee/df ed
de-f e/d Z3 ede2e2d Z4 ede3e3d Z5de!de0fdZ6de.de3fdZ7de.de8fdZ9dKde2d e0de$fd Z:d Z;d ee e/f d ee/df d!ee jx d"e0de2f
d#Z=de
ded$ fd%Z> G d& d'e Z?de
dee? fd(Z@e
dee
ee jx f fd) ZAed*e/d+e1d,e/d-dde2f
d. ZBed*e/d+e1d,e/d-e/de3f
d/ ZBd*e/d+e1d,e/d-ee/ dee2e3f f
d0ZBd1d2dee4e
f fd3ZCd4d5de2fd6ZDd1d2d7ee-e.d8f dee4e2ef fd9ZEde!d!ee# dee j ee# f fd:ZGd4e$de0fd;ZHde j de2fd<ZId1e?d7d=deee4e2f e2ef fd>ZJde j de8fd?ZKde.d@e1ddfdAZLd1e?dee5ef fdBZMd1e?d7d=deee3e5f e2ef fdCZN G dD dE ZOdFeee"eOf deee" eeO f fdGZPeeee2e3f eee2e3e4e5f ee2e3f f f ZQeeRdH< G dI dJe ZSy)Lz+Helpers for interfacing array like objects. N)ABCabstractmethod)cache)
TYPE_CHECKINGAnyCallableDictListLiteralOptionalProtocolTupleType TypeAlias TypedDict TypeGuardUnioncastoverload ) ArrowCatList CNumericPtrDataType FeatureTypes
NumpyDType NumpyOrCupy) import_cupyimport_pyarrowlazy_isinstancec e Zd Zedd Zy)
_ArrayLikeArgc y N selfs ^/var/www/ramen.bs-engineer-server.com/venv/lib/python3.12/site-packages/xgboost/_data_utils.py__array_interface__z!_ArrayLikeArg.__array_interface__/ s 14 N)returnArrayInf)__name__
__module__ __qualname__propertyr( r$ r) r' r! r! . s
4 4r) r! c e Zd Zedd Zy)_CudaArrayLikeArgc y r# r$ r% s r' __cuda_array_interface__z*_CudaArrayLikeArg.__cuda_array_interface__4 s :=r) N)r* CudaArrayInf)r, r- r. r/ r3 r$ r) r' r1 r1 3 s
= =r) r1 r+ .datatypestrversionstridesshapemaskr4 r7 r8 r9 r: r; r< stream StringArrayoffsetsvaluesCudaStringArrayr7 r* c z t | d xr. t | j d xr | j j S )z)Whether the numpy array has object dtype.dtype hasobject)hasattrrE rF r7 s r' array_hasobjectrI W s8 g ! DJJ , ! JJ r) c t | r t d | j }d|v r_|d Z|d j }|d }|d d }|d dv r2|d d d
k( r*|r(t | }d |d< |f|d< |j d d ||d< |S )
z6Returns a dictionary storing the CUDA array interface..Arrayz3Wrapper type for communicating with numpy and cupy.N
_interfacer* c | j S r# )re r% s r' r( z7from_array_interface..Array.__array_interface__ s ?? "r) r^ c R t j | | _ t | j d | j d< | j d d | j d d f| j d< | j j dd }|t | | j d< y y )Nr; r7 r r r: )copyre tupleget)r& r^ r: s r' r( z7from_array_interface..Array.__array_interface__ s "ii 2DO',T__W-E'FDOOG $'*'*'DOOF # oo)))T:G"-27^ * #r) c | j S r# r( r% s r' r3 z.Array.__cuda_array_interface__ s ++ +r) c || _ y r# rl )r& r^ s r' r3 z.Array.__cuda_array_interface__ s
'0D $r) .c , | j }|J |d S )zShape of the input array.r; rl )r& aifs r' r; z)from_array_interface..Array.shape s# **C? "?w< r) c @ t j | j S )z Total size of the input array.)npprodr; r% s r' sizez(from_array_interface..Array.size s 774::& &r) )r, r- r. __doc__re r r+ __annotations__r/ r( setterr3 r intr; rq
signedintegerrs r$ r) r' Arrayrd ~ s A)-
HX&- #(); #
#
# #
<
r r8 r; rE rh ) r3 r rs emptyr; rq rE arrayr( )r^ ra ry arrcpouts r' from_array_interfacer { s (' ('T 'C 9 '0$
] 88q=88#))288Ii r; r: )
r r| r3 rq r(
isinstancerw ctypesr c_void_pvaluerr STREAM_PER_THREAD)r r; rE r r| r} addrlengths r' make_array_interfacer s
##$e#<..t51)) c3 {{3066
F v{* * | 4LE&M85(+hE'NE) Lr) zpa.DictionaryArrayc t | dd S )z"Is this an arrow dictionary array?z pyarrow.libDictionaryArray)r rH s r'
is_arrow_dictr s 4 0A BBr) c e Zd ZdZe d d Zed
d Zedej fd Zedej fd Z
de d fdZ ede
fd Zedefd
Zy )
DfCatAccessorz!Protocol for pandas cat accessor.r* c y r# r$ r% s r'
categorieszDfCatAccessor.categories s r) c y r# r$ r% s r' codeszDfCatAccessor.codes $'r) c y r# r$ r% s r' rE zDfCatAccessor.dtype s !$r) c y r# r$ r% s r' rB zDfCatAccessor.values r r) )pa.StringArraypa.IntegerArrayc y r# r$ r% s r' to_arrowzDfCatAccessor.to_arrow s 58r) c y r# r$ r% s r' r3 z&DfCatAccessor.__cuda_array_interface__ s 8;r) c y r# r$ r% s r' _columnzDfCatAccessor._column s !r) N)r* zpd.Index)r* pd.Series)r, r- r. rt r/ r r rq rE ndarrayrB r r r4 r3 r r r$ r) r' r r s +
' '
$rxx$ $
'
' '9 2 39 ;,; ;
!! !r) r c 6 t | d xr t | d S )Nr r )rG rH s r'
_is_df_catr s 4 & A74+AAr) c dd l } | j t j | j t j | j t j | j
t j
| j t j | j t j | j t j | j t j | j t j | j t j | j t j i }|S )Nr )
pyarrowint8rq int16int32int64uint8uint16uint32uint64float16float32float64)pamappings r' _arrow_npdtyper
s 277
BHH
BHH
BHH
BHH RYY RYY RYY
bjj
bjj
bjj +G Nr) addressr8 rs r> c y r# r$ r r8 rs r> s r' _arrow_buf_infr s UXr) c y r# r$ r s r' r r # s r) c @ || df|dd |fd |d}|S | df|dd |fd d}|S )NTr5 r= r6 r$ )r r8 rs r> jcuaifjaifs r' r r ) s[ dO W
$
D Kr) catsr c t s
t }| j dk( sJ | j }|\ }}}|j sJ t | dz dt dt f fd }|j |t j k( rgt | j s d}t |dt | dz | j |j }|j \ }}}|j |t j" k7 r t d t% |j& d
d } t% |j& d |j d }
|J | |
d } | |||ffS )
Nr r typr* c N t j | j dz z S )N )rq iinfobits)r off_lens r' get_n_bytesz)_arrow_cat_names_inf..get_n_bytesP s! "((3-,,122r) z Expecting a `pyarrow.Array`.z Got: .z?Arrow dictionary type offsets is required to be 32-bit integer. r< )
r r r ry r r r r r r r r r )r} r r< r7 r jmasks r' _arrow_array_infr j s
eRXX &.tE{m<== JD$ 5zm uzz*KK
E \\4( %j]
%
{{ /E(OE&M Lr) r r c B t | \ }}t | }|||dffS )zHGet the array interface representation of a string-based category array.N)r r )r r r cats_tmpjcodess r'
arrow_cat_infr s/
,D1FH
e
$F 6Hd+ ++r) c t | s,| j t j t j fv r#t j
}| j
|d } | j j st j | d } | |fS )z7Ensure the np array has correct type and is contiguous.Fr{ A) requirements)
rI rE rq r bool_r astypeflagsalignedrequire)r7 rE s r' _ensure_np_dtyper sf t
rzz288.D D
{{5u{- :: zz$S1 ;r) c Z t | j dk( xs | j d dk( S )Nr )r r; )r} s r' _is_flattenr s( u{{ q 7EKKNa$77r) c t | r t d | j }d|v r|d j |d< t t | S )z*Returns an array interface from the input.rK r< )rI rQ r( r r+ )r7 rT s r' array_interface_dictr sH t WXX # #D
~F|77V $ r) r c |j dt j }dt f fd } | r5 j }t | }|j }t | }||||ffS dt dt t j t f fd} | j \ }} t |t j \ }}
t | } | j d } t j j t j | j" }
|
J |
dfd t% | fdd
dd }| |d }|j }t | }|| | |f}|||fS )
zCGet the array interface representation of pandas category accessor.r* c j } t j | t j xs$ t j | t j S # t
$ r Y yw xY w)NF)rE rq
issubdtypefloatingintegerr )rE r s r' is_primzpd_cat_inf..is_prim sL
== 4X
eRZZ8X X s A A A%$A%strarrc t | t j s>t | d r| j t
} nt j | t
} t j t }t j t j t j dgt j || g }| j j dk( r-| j D cg c] }|j! d }}n&| j D cg c]
}t# | }}dj% | }d|v rt' j( dt* |j- t j. |fS c c}w c c}w ) z5Convert a string-like array to an arrow string array.to_numpy)rE r SrY zzFound embedded NUL (\0) characters in string categories. Arrow used to strip these characters, but they are now preserved.)r rq r rG r objectasarray vectorizer cumsum concatenater} r rE kindtolistdecodestrjoinwarningswarn UserWarningr r )r lenarrrA sstr_listrB s r' npstr_to_arrow_strarrz)pd_cat_inf..npstr_to_arrow_strarr s &"**-vz* v 6F&9 c")) NNBHHaS 96&>J K
<< #39==?Ca)CHC(.
81A8H8" 6> MMX
~~bhh'// D8s 1F FrY NTr r5 )r7 r8 r; r: r9 r< r@ )replacerq nanboolrB r r r r r r r encoder r from_bufferc_char_pr r )r r r name_values_num jarr_values code_values
jarr_codesr name_offsets name_values_joffsetsbvaluesr jvaluesr r bufs ` r'
pd_cat_infr s^
MM"bff %ET y ++*?; ll )+6
J +(FFF0c 0eBJJO.D 08 !6dkk B L+&|RXX>OL!#L1H )G
//
%
%foog&>
?
E
EC ?? d k"$
G '/'BF,,K
!+
.F C 63 r) c Z t | }t t j | d }|S )zMake array interface str.rY )r rZ r[ r\ r] s r' array_interfacer s' $T*I$**Y/9M r) fieldc \ d| j v r | j d t d| yy)z(Make sure no missing value in meta data.r< Nz"Missing value is not allowed for: )r3 rQ )r7 r s r' check_cudf_metar s@ $/// ) )& 1 ==eWEFF
> 0r) c ddl }| j j }|j j |j
j k7 r t d |j dk( sJ |j dk( sJ |j d }|j j |j
j k( sJ d t |j j d d ddt }t |j j d d d|j! t }||d}||ffS ) z7String category index path for :py:func:`cudf_cat_inf`.r NzDUnexpected type for category index. It's neither numeric nor string.z Expected INT32 string offsets.r7 r r r@ ) pylibcudfr to_pylibcudfr idTypeIdSTRINGr
null_countr childrenINT32r r7 r3 r rs )r plcplc_col off_childr r r s r' _cudf_str_cat_infr+ s8 ll'')G||~cjj / // R
1 $$ $ >> q "1%I >> CJJ$4$4 4V6VV 4 ) //7:
E +11&9!<
G +2U CF G: r) c , t }|j | j |j xs& |j | j |j }|r t | }t | }||| |ffS t
| \ }}t | }|||fS )z4Obtain the cuda array interface for cuDF categories.)r r rE r r rW r+ ) r r r
is_num_idx cats_ainf
codes_ainfr r r s r' cudf_cat_infr0 6 s
B tzz2;;7 2==
BJJ<J -d3 .u5
*tUm33#D)KFC
&u
-F 63 r) c ~ e Zd ZdZdeej eg df f dee ddfdZ
de fdZ de fdZ
defd Zd d
Zy)
Categoriesa An internal storage class for categories returned by the DMatrix and the
Booster. This class is designed to be opaque. It is intended to be used exclusively
by XGBoost as an intermediate storage for re-coding categorical data.
The categories are saved along with the booster object. As a result, users don't
need to preserve this class for re-coding. Use the booster model IO instead if you
want to preserve the categories in a stable format.
.. versionadded:: 3.1.0
.. warning::
This class is internal.
.. code-block:: python
Xy = xgboost.QuantileDMatrix(X, y, enable_categorical=True)
booster = xgboost.train({}, Xy)
categories = booster.get_categories() # Get categories
# Use categories as a reference for re-coding
Xy_new = xgboost.QuantileDMatrix(
X_new, y_new, feature_types=categories, enable_categorical=True, ref=Xy
)
# Categories will be part of the `model.json`.
booster.save_model("model.json")
handleN arrow_arraysr* c 0 |\ | _ | _ || _ y r# )_handle_free
_arrow_arrays)r& r3 r4 s r' __init__zCategories.__init__h s $* dj )r) c H | j t d | j S )ao Get the categories in the dataset. The results are stored in a list of
(feature name, arrow array) pairs, with one array for each categorical
feature. If a feature is numerical, then the corresponding column in the list is
None. A value error will be raised if this container was created without the
`export_to_arrow` option.
zHThe `export_to_arrow` option of the `get_categories` method is required.)r8 rQ r% s r' r zCategories.to_arrowu s0 %
!!!r) c 2 | j j du S )z$Returns True if there's no category.Nr6 r r% s r' r| zCategories.empty s ||!!T))r) c ^ | j j sJ | j j S )z*Internal method for retrieving the handle.r< r% s r'
get_handlezCategories.get_handle s% ||!!!!||!!!r) c R | j j y | j y r# )r6 r r7 r% s r' __del__zCategories.__del__ s << %
r) )r* N)r, r- r. rt r r r r r r r9 r r
r| rw r> r@ r$ r) r' r2 r2 H sm > *fooxD'99: * |, *
*
",
" *t *"C "
r) r2
feature_typesc > t | t r| }d} | |fS d}| |fS )zGet the optional reference categories from the `feature_types`. This is used by
various `DMatrix` where the `feature_types` is reused for specifying the reference
categories.
N)r r2 )rA ref_categoriess r' get_ref_categoriesrD s7 -,&
. (( . ((r) AifTypec j e Zd ZdZdee dedee ddfdZ de
fdZ e e
deeef fd Zy)
TransformedDfzInternal class for storing transformed dataframe.
Parameters
----------
ref_categories :
Optional reference categories used for re-coding.
aitfs :
Array interface for each column.
rC aitfstemporary_buffersr* Nc || _ |(|j |j }|| _ nd | _ || _ || _ y r# )rC r> ref_aifrH rI )r& rC rH rI ro s r' r9 zTransformedDf.__init__ sK - %.*C*C*E*Q ++-C*-DL DL
!2 r) c | j :| j | j d}t t j
| d }|S t t j
| j d }|S )z6Return a byte string for JSON encoded array interface.)rC columnsrY )rC rK rH rZ r[ r\ )r& ref_infinfs r' r z TransformedDf.array_interface s\ */3||
SG
7+W5C
4::.8C
r) c y)z"Return the shape of the dataframe.Nr$ r% s r' r; zTransformedDf.shape s r) )r, r- r. rt r r2 rE r
r r9 rZ r r/ r rw r; r$ r) r' rG rG sr
3 ,3 3 ; 3
3 1uS#X 1 1r) rG )F)Trt rh r r[ r abcr r functoolsr fcachetypingr r r r r
r r r
r r r r r r r r numpyrq _typingr r r r r r compatr r r pandaspdr r r! r1 rw r
r r+ r4 r? rC rI rW rZ r` r r numberr r r r r r r r r r r r r r r r r+ r0 r2 rD rE ru rG r$ r) r' r[ s/ 1
# % & A @5H 5
> >
c4i 1:E#s(O,sCx j$
56
c4i 1:E#s(O,sCx j$
56
8x'PQ <<H
( t $5 , ( 0 U >H > >+ >D
{C c? ?
DC C *> ? C
"H "8BS BY} 5 B
S$ryy/ 12 (
XC X# XS X$ X8 X
X
&)36
&)3;C=
8\ !"4$( / $(E+s:J4K $(N
F,
, 13DD E, ;% '(,
#J/
2::x
+ +, 8{ 8t 8 rzz h M
M +M
5h& '5 89M `"** G+ GC GD G! M ! eOU4J.K ! H
+
5 . /5 @A $H HV
)E,
": ;<
)
8L !8J#7 78
)"
h $%
(L+F G (L( )
+
)1C )1r)