Sadržaj Pregled 
 
 
 
 
 Uvod Struktura ekstrakta tenzora Unesite podatke u tenzore Daljnje čitanje i resursi Kada radite na ML aplikacijama kao što su detekcija objekata i NLP, ponekad je potrebno raditi sa pododjeljcima (rezima) tenzora. Na primjer, ako vaša arhitektura modela uključuje usmjeravanje, gde jedan sloj može kontrolisati koji primer obuke se usmjerava na sledeći sloj. U ovom slučaju, možete koristiti opcije rezanja tenzora da podelite tenzore i stavite ih zajedno u pravom redoslijedu. U NLP aplikacijama, možete koristiti tensor rezanje za izvršavanje maskiranje reči tokom obuke. Na primjer, možete generisati podatke o obuci iz liste rečenica tako što ćete odabrati indeks reči za maskiranje u svakoj rečenici, uzeti reč kao oznaku, a zatim zamijeniti odabranu reč sa tokenom maske. U ovom vodiču, naučit ćete kako koristiti TensorFlow API za: 
 
 
 Izvadite rezove iz tenzora Unesite podatke na specifične indekse u tenzor Ovaj vodič pretpostavlja da ste upoznati s indeksiranjem tenzora. Pročitajte odjeljke o indeksiranju i vodiča pre nego što počnete sa ovim vodičem. Tenzor Uslovi korišćenja NumPy postavljanje import tensorflow as tf
import numpy as np
 2024-08-15 02:59:22.387959: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:485] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered
2024-08-15 02:59:22.409014: E external/local_xla/xla/stream_executor/cuda/cuda_dnn.cc:8454] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
2024-08-15 02:59:22.415391: E external/local_xla/xla/stream_executor/cuda/cuda_blas.cc:1452] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
 Struktura ekstrakta tenzora Izvršite Numpy-like tensor slicing pomoću . tf.slice t1 = tf.constant([0, 1, 2, 3, 4, 5, 6, 7])

print(tf.slice(t1,
               begin=[1],
               size=[3]))
 tf.Tensor([1 2 3], shape=(3,), dtype=int32)
WARNING: All log messages before absl::InitializeLog() is called are written to STDERR
I0000 00:00:1723690765.029742  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690765.033507  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690765.037154  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690765.040869  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690765.052765  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690765.056359  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690765.059719  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690765.063138  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690765.066532  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690765.070040  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690765.073650  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690765.077235  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690766.300773  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690766.302797  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690766.304766  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690766.306851  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690766.308843  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690766.310723  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690766.312589  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690766.314574  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690766.316423  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690766.318291  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690766.320168  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690766.322150  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690766.360968  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690766.362946  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690766.364849  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690766.366867  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690766.368710  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690766.370591  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690766.372447  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690766.374435  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690766.376302  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690766.378677  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690766.380966  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1723690766.383365  169903 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
 Alternativno, možete koristiti više Pythonic sintaksu. Imajte na umu da su tensor rezovi ravnomjerno raspršeni po rasponu start-stop. print(t1[1:4])
 tf.Tensor([1 2 3], shape=(3,), dtype=int32)
 print(t1[-3:])
 tf.Tensor([5 6 7], shape=(3,), dtype=int32)
 Za 2-dimenzionalne tenzore, možete koristiti nešto poput: t2 = tf.constant([[0, 1, 2, 3, 4],
                  [5, 6, 7, 8, 9],
                  [10, 11, 12, 13, 14],
                  [15, 16, 17, 18, 19]])

print(t2[:-1, 1:3])
 tf.Tensor(
[[ 1  2]
 [ 6  7]
 [11 12]], shape=(3, 2), dtype=int32)
 Možete koristiti Takođe na višim dimenzionalnim tenzorima. tf.slice t3 = tf.constant([[[1, 3, 5, 7],
                   [9, 11, 13, 15]],
                  [[17, 19, 21, 23],
                   [25, 27, 29, 31]]
                  ])

print(tf.slice(t3,
               begin=[1, 1, 0],
               size=[1, 1, 2]))
 tf.Tensor([[[25 27]]], shape=(1, 1, 2), dtype=int32)
 Također možete koristiti da se ekstraktuju rezovi tenzora tako što će se "trčati" preko dimenzija tenzora. tf.strided_slice Koristite da izvuče specifične indekse iz jedne osi tenzora. tf.gather print(tf.gather(t1,
                indices=[0, 3, 6]))

# This is similar to doing

t1[::3]
 tf.Tensor([0 3 6], shape=(3,), dtype=int32)
<tf.Tensor: shape=(3,), dtype=int32, numpy=array([0, 3, 6], dtype=int32)>
 ne zahtijeva da indeksi budu ravnomjerno raspodijeljeni. tf.gather alphabet = tf.constant(list('abcdefghijklmnopqrstuvwxyz'))

print(tf.gather(alphabet,
                indices=[2, 0, 19, 18]))
 tf.Tensor([b'c' b'a' b't' b's'], shape=(4,), dtype=string)
 Da biste izvukli rezove iz više osi tenzora, koristite Ovo je korisno kada želite da skupite elemente matrice, a ne samo njegove redove ili kolone. tf.gather_nd t4 = tf.constant([[0, 5],
                  [1, 6],
                  [2, 7],
                  [3, 8],
                  [4, 9]])

print(tf.gather_nd(t4,
                   indices=[[2], [3], [0]]))
 tf.Tensor(
[[2 7]
 [3 8]
 [0 5]], shape=(3, 2), dtype=int32)
 t5 = np.reshape(np.arange(18), [2, 3, 3])

print(tf.gather_nd(t5,
                   indices=[[0, 0, 0], [1, 2, 1]]))
 tf.Tensor([ 0 16], shape=(2,), dtype=int64)
 # Return a list of two matrices

print(tf.gather_nd(t5,
                   indices=[[[0, 0], [0, 2]], [[1, 0], [1, 2]]]))
 tf.Tensor(
[[[ 0  1  2]
  [ 6  7  8]]

 [[ 9 10 11]
  [15 16 17]]], shape=(2, 2, 3), dtype=int64)
 # Return one matrix

print(tf.gather_nd(t5,
                   indices=[[0, 0], [0, 2], [1, 0], [1, 2]]))
 tf.Tensor(
[[ 0  1  2]
 [ 6  7  8]
 [ 9 10 11]
 [15 16 17]], shape=(4, 3), dtype=int64)
 Unesite podatke u tenzore Koristite za umetanje podataka na određene rezove/indekse tensora. Imajte na umu da je tensor u koji umetnete vrijednosti nula-inicijalizovan. tf.scatter_nd t6 = tf.constant([10])
indices = tf.constant([[1], [3], [5], [7], [9]])
data = tf.constant([2, 4, 6, 8, 10])

print(tf.scatter_nd(indices=indices,
                    updates=data,
                    shape=t6))
 tf.Tensor([ 0  2  0  4  0  6  0  8  0 10], shape=(10,), dtype=int32)
 Metode poput koji zahtijevaju nulte inicijalizovane tenzore slični su rijetkim inicijalizatorima tenzora. i da oponaša ponašanje sparse tensor ops. tf.scatter_nd tf.gather_nd tf.scatter_nd Razmotrite primer u kojem gradite spor tenzor koristeći ove dve metode u kombinaciji. # Gather values from one tensor by specifying indices

new_indices = tf.constant([[0, 2], [2, 1], [3, 3]])
t7 = tf.gather_nd(t2, indices=new_indices)
 # Add these values into a new tensor

t8 = tf.scatter_nd(indices=new_indices, updates=t7, shape=tf.constant([4, 5]))

print(t8)
 tf.Tensor(
[[ 0  0  2  0  0]
 [ 0  0  0  0  0]
 [ 0 11  0  0  0]
 [ 0  0  0 18  0]], shape=(4, 5), dtype=int32)
 Ovo je slično: t9 = tf.SparseTensor(indices=[[0, 2], [2, 1], [3, 3]],
                     values=[2, 11, 18],
                     dense_shape=[4, 5])

print(t9)
 SparseTensor(indices=tf.Tensor(
[[0 2]
 [2 1]
 [3 3]], shape=(3, 2), dtype=int64), values=tf.Tensor([ 2 11 18], shape=(3,), dtype=int32), dense_shape=tf.Tensor([4 5], shape=(2,), dtype=int64))
 # Convert the sparse tensor into a dense tensor

t10 = tf.sparse.to_dense(t9)

print(t10)
 tf.Tensor(
[[ 0  0  2  0  0]
 [ 0  0  0  0  0]
 [ 0 11  0  0  0]
 [ 0  0  0 18  0]], shape=(4, 5), dtype=int32)
 Za umetanje podataka u tenzor sa prethodno postojećim vrednostima, koristite . tf.tensor_scatter_nd_add t11 = tf.constant([[2, 7, 0],
                   [9, 0, 1],
                   [0, 3, 8]])

# Convert the tensor into a magic square by inserting numbers at appropriate indices

t12 = tf.tensor_scatter_nd_add(t11,
                               indices=[[0, 2], [1, 1], [2, 0]],
                               updates=[6, 5, 4])

print(t12)
 tf.Tensor(
[[2 7 6]
 [9 5 1]
 [4 3 8]], shape=(3, 3), dtype=int32)
 Slično tome, upotreba da oduzme vrednosti iz tenzora sa prethodno postojećim vrednostima. tf.tensor_scatter_nd_sub # Convert the tensor into an identity matrix

t13 = tf.tensor_scatter_nd_sub(t11,
                               indices=[[0, 0], [0, 1], [1, 0], [1, 1], [1, 2], [2, 1], [2, 2]],
                               updates=[1, 7, 9, -1, 1, 3, 7])

print(t13)
 tf.Tensor(
[[1 0 0]
 [0 1 0]
 [0 0 1]], shape=(3, 3), dtype=int32)
 Koristite kopiranje elementarnih minimalnih vrijednosti iz jednog tenzora u drugi. tf.tensor_scatter_nd_min t14 = tf.constant([[-2, -7, 0],
                   [-9, 0, 1],
                   [0, -3, -8]])

t15 = tf.tensor_scatter_nd_min(t14,
                               indices=[[0, 2], [1, 1], [2, 0]],
                               updates=[-6, -5, -4])

print(t15)
 tf.Tensor(
[[-2 -7 -6]
 [-9 -5  1]
 [-4 -3 -8]], shape=(3, 3), dtype=int32)
 Slično tome, upotreba kopirati maksimalne vrednosti u vidu elemenata iz jednog tenzora u drugi. tf.tensor_scatter_nd_max t16 = tf.tensor_scatter_nd_max(t14,
                               indices=[[0, 2], [1, 1], [2, 0]],
                               updates=[6, 5, 4])

print(t16)
 tf.Tensor(
[[-2 -7  6]
 [-9  5  1]
 [ 4 -3 -8]], shape=(3, 3), dtype=int32)
 Daljnje čitanje i resursi U ovom vodiču naučili ste kako da koristite opcije rezanja tensora dostupne s TensorFlowom kako biste izvršili finiju kontrolu nad elementima u vašim tensorima. 
 
 
 Pogledajte opcije rezanja dostupne s TensorFlow NumPy kao što su tf.experimental.numpy.take_along_axis i tf.experimental.numpy.take. Također pogledajte Tensor vodič i varijabilni vodič. 
 
 Prvotno objavljen na web stranici TensorFlow, ovaj članak se pojavljuje ovdje pod novim naslovom i licenciran je pod CC BY 4.0. Prvotno objavljen na web stranici TensorFlow, ovaj članak se pojavljuje ovdje pod novim naslovom i licenciran je pod CC BY 4.0. Tenzorski tokovi Tenzorski tokovi

Walkthroughs, tutorials, guides, and tips. This story will teach you how to do something new or how to do something better.

Ovaj audio je proizveden na originalnom jeziku priče!

Predugo; Citati

Make resilience your competitive advantage

Tensor rezanje i umetanje podataka učinjeno lako sa TensorFlow

About Author

KOMENTARI

HANG TAGS

OVAJ ČLANAK JE PREDSTAVLJEN U

Related Stories

Light-Mode

Classic

Newspaper

Dark-Mode

Neon Noir

Minty

HN StartUps