Akukho TensorFlow ngaphandle Tensors

Imininingwane Overview Imininingwane mayelana Shapes Ukucaciswa Ukucubungula Shapes Ukubuyekezwa kwe-Dtypes Ukuhlobisa tf.convert_to ku_tensor Ukuhlobisa Tensors Ukuhlobisa Tensile Ukuhlobisa Tensors Tensors zihlanganisa i-array ye-multidimensional nge-type ye-uniform (eyaziwa njenge-a ) Uyakwazi ukubona konke okukhuselwa Ukusuka . dtype dtypes tf.dtypes Uma unemibuzo , ama-tensors zihlanganisa like . Ukubala np.arrays Zonke ama-tensors zihlukile njenge-Python ama-number ne-strings: ungakwazi ukuhlaziywa okuqukethwe kwe-tensor, kuphela ukwakha omusha. import tensorflow as tf import numpy as np 2024-08-15 03:05:18.327501: 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 03:05:18.348450: 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 03:05:18.354825: 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 Imininingwane Okokuqala, ukwakha amanye ama-tensors eziyinhloko. Ngiyazi "scalar" noma "rank-0" tensor. I-scalar inesibini elilodwa, futhi akukho "i-axes". # This will be an int32 tensor by default; see "dtypes" below. rank_0_tensor = tf.constant(4) print(rank_0_tensor) tf.Tensor(4, shape=(), dtype=int32) WARNING: All log messages before absl::InitializeLog() is called are written to STDERR I0000 00:00:1723691120.932442 176945 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 00:00:1723691120.936343 176945 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 00:00:1723691120.940040 176945 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 00:00:1723691120.943264 176945 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 00:00:1723691120.954872 176945 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 00:00:1723691120.958376 176945 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 00:00:1723691120.961894 176945 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 00:00:1723691120.964843 176945 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 00:00:1723691120.967730 176945 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 00:00:1723691120.971300 176945 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 00:00:1723691120.974711 176945 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 00:00:1723691120.977717 176945 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 00:00:1723691122.208679 176945 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 00:00:1723691122.210786 176945 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 00:00:1723691122.212791 176945 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 00:00:1723691122.214776 176945 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 00:00:1723691122.216798 176945 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 00:00:1723691122.218734 176945 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 00:00:1723691122.220650 176945 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 00:00:1723691122.222554 176945 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 00:00:1723691122.224486 176945 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 00:00:1723691122.226429 176945 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 00:00:1723691122.228329 176945 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 00:00:1723691122.230251 176945 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 00:00:1723691122.269036 176945 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 00:00:1723691122.271069 176945 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 00:00:1723691122.273006 176945 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 00:00:1723691122.274956 176945 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 00:00:1723691122.276917 176945 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 00:00:1723691122.278854 176945 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 00:00:1723691122.280754 176945 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 00:00:1723691122.282664 176945 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 00:00:1723691122.284613 176945 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 00:00:1723691122.287058 176945 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 00:00:1723691122.289508 176945 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 00:00:1723691122.291891 176945 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 A "vector" noma "rank-1" tensor kuyinto like a list of values. A vector has one axis: # Let's make this a float tensor. rank_1_tensor = tf.constant([2.0, 3.0, 4.0]) print(rank_1_tensor) tf.Tensor([2. 3. 4.], shape=(3,), dtype=float32) A "matrix" or "rank-2" tensor has two axes: # If you want to be specific, you can set the dtype (see below) at creation time rank_2_tensor = tf.constant([[1, 2], [3, 4], [5, 6]], dtype=tf.float16) print(rank_2_tensor) tf.Tensor( [[1. 2.] [3. 4.] [5. 6.]], shape=(3, 2), dtype=float16) A scalar, shape: [] A vector, shape: [3] A matrix, shape: [3, 2] Tensors may have more axes; here is a tensor with three axes: # There can be an arbitrary number of # axes (sometimes called "dimensions") rank_3_tensor = tf.constant([ [[0, 1, 2, 3, 4], [5, 6, 7, 8, 9]], [[10, 11, 12, 13, 14], [15, 16, 17, 18, 19]], [[20, 21, 22, 23, 24], [25, 26, 27, 28, 29]],]) print(rank_3_tensor) tf.Tensor( [[[ 0 1 2 3 4] [ 5 6 7 8 9]] [[10 11 12 13 14] [15 16 17 18 19]] [[20 21 22 23 24] [25 26 27 28 29]]], shape=(3, 2, 5), dtype=int32) Kwezindlela ezininzi ungayifundisa i-tensor enezinhloko ezingu-2. A 3-axis tensor, shape: [3, 2, 5] Ungayifaka i-tensor ku-NumPy array noma usebenzisa noma i Ukucaciswa: np.array tensor.numpy np.array(rank_2_tensor) array([[1., 2.], [3., 4.], [5., 6.]], dtype=float16) rank_2_tensor.numpy() array([[1., 2.], [3., 4.], [5., 6.]], dtype=float16) Tensors ngokuvamile zihlanganisa floats futhi ints, kodwa kukhona nezinye izinhlobo eziningi, kuhlanganise: Inombolo Complex Ukuhlobisa Ibhizinisi Class inikeza tensors ukuba 'rectangular' --- okuyinto, ngokulandelana nenkinga, zonke izakhiwo ubukhulu efanayo. Nokho, kukhona izinhlobo specialized of tensors that can handle different shapes: tf.Tensor Ukuhlobisa Tensile Ukuhlobisa Tensile Uyakwazi ukwenza imathemikhali esisodwa ku-tensors, kuhlanganise ukuguqulwa, ukuguqulwa kwe-element-like, kanye nokuguqulwa kwe-matrix. a = tf.constant([[1, 2], [3, 4]]) b = tf.constant([[1, 1], [1, 1]]) # Could have also said `tf.ones([2,2], dtype=tf.int32)` print(tf.add(a, b), "\n") print(tf.multiply(a, b), "\n") print(tf.matmul(a, b), "\n") tf.Tensor( [[2 3] [4 5]], shape=(2, 2), dtype=int32) tf.Tensor( [[1 2] [3 4]], shape=(2, 2), dtype=int32) tf.Tensor( [[3 3] [7 7]], shape=(2, 2), dtype=int32) print(a + b, "\n") # element-wise addition print(a * b, "\n") # element-wise multiplication print(a @ b, "\n") # matrix multiplication tf.Tensor( [[2 3] [4 5]], shape=(2, 2), dtype=int32) tf.Tensor( [[1 2] [3 4]], shape=(2, 2), dtype=int32) tf.Tensor( [[3 3] [7 7]], shape=(2, 2), dtype=int32) I-Tensors isetshenziselwa zonke izinhlobo zokusebenza (noma "Ops"). c = tf.constant([[4.0, 5.0], [10.0, 1.0]]) # Find the largest value print(tf.reduce_max(c)) # Find the index of the largest value print(tf.math.argmax(c)) # Compute the softmax print(tf.nn.softmax(c)) tf.Tensor(10.0, shape=(), dtype=float32) tf.Tensor([1 0], shape=(2,), dtype=int64) tf.Tensor( [[2.6894143e-01 7.3105854e-01] [9.9987662e-01 1.2339458e-04]], shape=(2, 2), dtype=float32) Qaphela: Ngokuvamile, lapho umsebenzi we-TensorFlow ibheka i-Tensor njenge-input, isicelo esithathwe konke okwenziwe ku-Tensor usebenzisa i-tf.convert_to_tensor. Khona isibonelo ngezansi. Qaphela: Ngokuvamile, lapho umsebenzi we-TensorFlow ibheka i-Tensor njenge-input, isicelo esithathwe konke okwenziwe ku-Tensor usebenzisa i-tf.convert_to_tensor. Khona isibonelo ngezansi. tf.convert_to_tensor([1,2,3]) tf.reduce_max([1,2,3]) tf.reduce_max(np.array([1,2,3])) mayelana Shapes I-Tensors has shapes. Abaningi i-vocabulary: Umbala: Ubude (inombolo yama-elements) yezinhlayiyana ngamunye zezinhlayiyana ze-tensor. I-Rank: Inombolo ye-tensor axes. I-scalar inesibini ye-0, i-vector inesibini ye-1, futhi i-matrix inesibini ye-2 Axis noma Dimension: Ubukhulu obuningi we-tensor. Usayizi: Inani lwezinto ephelele ku-tensor, umkhiqizo yezinto ze-vector ye-shape. Qaphela: Nangona ungayifaka ku-"tensor ye-2-dimensional", i-tensor ye-rank-2 ikakhulukazi akubonisa indawo ye-2D. Qaphela: Nangona ungayifaka ku-"tensor ye-2-dimensional", i-tensor ye-rank-2 ikakhulukazi akubonisa indawo ye-2D. Tensors kanye Izinto zihlanganisa izakhiwo ezinhle ukufinyelela ngezinto: tf.TensorShape rank_4_tensor = tf.zeros([3, 2, 4, 5]) A rank-4 tensor, shape: [3, 2, 4, 5] print("Type of every element:", rank_4_tensor.dtype) print("Number of axes:", rank_4_tensor.ndim) print("Shape of tensor:", rank_4_tensor.shape) print("Elements along axis 0 of tensor:", rank_4_tensor.shape[0]) print("Elements along the last axis of tensor:", rank_4_tensor.shape[-1]) print("Total number of elements (3*2*4*5): ", tf.size(rank_4_tensor).numpy()) Type of every element: Number of axes: 4 Shape of tensor: (3, 2, 4, 5) Elements along axis 0 of tensor: 3 Elements along the last axis of tensor: 5 Total number of elements (3*2*4*5): 120 Ngiyazi ukuthi Waze Izinzuzo Amasipala izikhwama. Uma ufuna a Ukusebenzisa noma Uhlobo le-function. Lezi zihlanganisa, kodwa kungenziwa ebalulekile ekukhiqizeni i-graphs (ngemuva kwalokho). Tensor.ndim Tensor.shape Tensor Tensor tf.rank tf.shape tf.rank(rank_4_tensor) tf.shape(rank_4_tensor) Nangona ama-axis zihlanganisa ngokuvamile nge-indices yayo, kufanele uxhumane ngokuvamile igama lwezinye. Ngokuvamile ama-axis zihlanganisa kusuka ku-global kuya ku-local: Isisindo se-batch kuqala, esilandelayo ngezinqubo zendawo, futhi izici ze-location ngamunye. Ngokuvamile, ama-vectors ze-feature zihlanganisa izindawo ezingxenyeni ze-memory. Typical axis order Ukucaciswa I-indexing ye-single axis I-TensorFlow ivela ngezinsizakalo ezivamile ze-Python ye-indexing, efana ne , kanye nezinsizakalo eziyisisekelo ze-NumPy indexing. Ukubhalisa umbhalo noma umbhalo ku-Python indices ezivela ku-0 I-indices ye-negative ibekwe ngokushesha kusukela ekupheleni ama-colons, :, isetshenziselwa izikhwama: start:stop:step rank_1_tensor = tf.constant([0, 1, 1, 2, 3, 5, 8, 13, 21, 34]) print(rank_1_tensor.numpy()) [ 0 1 1 2 3 5 8 13 21 34] Ukubhalisa nge-scalar ukunciphisa isikhunta: print("First:", rank_1_tensor[0].numpy()) print("Second:", rank_1_tensor[1].numpy()) print("Last:", rank_1_tensor[-1].numpy()) First: 0 Second: 1 Last: 34 Ukubhalisa nge-a slice ukugcina i-axis: : print("Everything:", rank_1_tensor[:].numpy()) print("Before 4:", rank_1_tensor[:4].numpy()) print("From 4 to the end:", rank_1_tensor[4:].numpy()) print("From 2, before 7:", rank_1_tensor[2:7].numpy()) print("Every other item:", rank_1_tensor[::2].numpy()) print("Reversed:", rank_1_tensor[::-1].numpy()) Everything: [ 0 1 1 2 3 5 8 13 21 34] Before 4: [0 1 1 2] From 4 to the end: [ 3 5 8 13 21 34] From 2, before 7: [1 2 3 5 8] Every other item: [ 0 1 3 8 21] Reversed: [34 21 13 8 5 3 2 1 1 0] I-indexing ye-multi-axis Tensors okusezingeni eliphezulu zihlanganiswa ngokulandelana nezindices eziningi. Izinsizakalo efanayo ngokufanayo ne-single-axis isicelo ngamunye isixazululo ngokufanayo. print(rank_2_tensor.numpy()) [[1. 2.] [3. 4.] [5. 6.]] Ukubuyekeza inani elilodwa ngamunye ye-index, imiphumela kuyinto i-scalar. # Pull out a single value from a 2-rank tensor print(rank_2_tensor[1, 1].numpy()) You can index using any combination of integers and slices: # Get row and column tensors print("Second row:", rank_2_tensor[1, :].numpy()) print("Second column:", rank_2_tensor[:, 1].numpy()) print("Last row:", rank_2_tensor[-1, :].numpy()) print("First item in last column:", rank_2_tensor[0, -1].numpy()) print("Skip the first row:") print(rank_2_tensor[1:, :].numpy(), "\n") Second row: [3. 4.] Second column: [2. 4. 6.] Last row: [5. 6.] First item in last column: 2.0 Skip the first row: [[3. 4.] [5. 6.]] Here is an example with a 3-axis tensor: print(rank_3_tensor[:, :, 4]) tf.Tensor( [[ 4 9] [14 19] [24 29]], shape=(3, 2), dtype=int32) Selecting the last feature across all locations in each example in the batch Okufakiwe ukuze ufunde indlela yokusebenza kwe-indexing ukuguqulwa izinhlayiya ezithile ku-tensors akho. Umhlahlandlela we-Slicing Guide Ukucubungula Shapes Ukuguqulwa kwe-tensor kuyinto ezinhle kakhulu. # Shape returns a `TensorShape` object that shows the size along each axis x = tf.constant([[1], [2], [3]]) print(x.shape) (3, 1) # You can convert this object into a Python list, too print(x.shape.as_list()) [3, 1] Ungayifaka i-tensor ku-shape entsha. I- Ukusebenza ngokushesha futhi kulungile njengoba idatha esifundayo ayidinga ukuguqulwa. tf.reshape # You can reshape a tensor to a new shape. # Note that you're passing in a list reshaped = tf.reshape(x, [1, 3]) print(x.shape) print(reshaped.shape) (3, 1) (1, 3) I-Data ibhekwa emkhakheni yayo futhi i-tensor entsha ifakwe, ne-shape efakiwe, enikeza idatha efanayo. I-TensorFlow isetshenziselwa ukuguqulwa kwe-memory ye-C-style "row-major", lapho ukuguqulwa kwe-indicate ebusweni kulinganiswa ne-step eyodwa emkhakheni. print(rank_3_tensor) tf.Tensor( [[[ 0 1 2 3 4] [ 5 6 7 8 9]] [[10 11 12 13 14] [15 16 17 18 19]] [[20 21 22 23 24] [25 26 27 28 29]]], shape=(3, 2, 5), dtype=int32) Uma u-tensor ufakwe, ungakwazi ukubona ukuthi isigaba esekelwe ku-memory. # A `-1` passed in the `shape` argument says "Whatever fits". print(tf.reshape(rank_3_tensor, [-1])) tf.Tensor( [ 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29], shape=(30,), dtype=int32) Ngokuvamile, ukusetshenziswa okuhle kuphela Ukubambisana noma ukwahlukanisa izindandatho zangaphakathi (noma ukongeza / ukuthatha Ukubuyekezwa tf.reshape 1 Ukuze lokhu 3x2x5 tensor, ukuguqulwa ku (3x2)x5 noma 3x(2x5) zihlanganisa izinto ezinhle ukwenza, njengoba izikhwama zihlanganisa: print(tf.reshape(rank_3_tensor, [3*2, 5]), "\n") print(tf.reshape(rank_3_tensor, [3, -1])) tf.Tensor( [[ 0 1 2 3 4] [ 5 6 7 8 9] [10 11 12 13 14] [15 16 17 18 19] [20 21 22 23 24] [25 26 27 28 29]], shape=(6, 5), dtype=int32) tf.Tensor( [[ 0 1 2 3 4 5 6 7 8 9] [10 11 12 13 14 15 16 17 18 19] [20 21 22 23 24 25 26 27 28 29]], shape=(3, 10), dtype=int32) Some good reshapes. Reshaping "ukusebenza" noma iyiphi isakhiwo esitsha nge inani elifanayo yama-elements, kodwa akuyona yini elithakazelisayo uma ungacacacindezeleka inqubo ye-axes. Ukushintshwa kwama-axis akufanele; kufanele Ukuze lokhu. tf.reshape tf.transpose # Bad examples: don't do this # You can't reorder axes with reshape. print(tf.reshape(rank_3_tensor, [2, 3, 5]), "\n") # This is a mess print(tf.reshape(rank_3_tensor, [5, 6]), "\n") # This doesn't work at all try: tf.reshape(rank_3_tensor, [7, -1]) except Exception as e: print(f"{type(e).__name__}: {e}") tf.Tensor( [[[ 0 1 2 3 4] [ 5 6 7 8 9] [10 11 12 13 14]] [[15 16 17 18 19] [20 21 22 23 24] [25 26 27 28 29]]], shape=(2, 3, 5), dtype=int32) tf.Tensor( [[ 0 1 2 3 4 5] [ 6 7 8 9 10 11] [12 13 14 15 16 17] [18 19 20 21 22 23] [24 25 26 27 28 29]], shape=(5, 6), dtype=int32) InvalidArgumentError: { {function_node __wrapped__Reshape_device_/job:localhost/replica:0/task:0/device:GPU:0} } Input to reshape is a tensor with 30 values, but the requested shape requires a multiple of 7 [Op:Reshape] Some bad reshapes. Ungayisebenzisa izakhiwo ezingenalutho ezingenalutho. Noma isakhiwo ifakwe (umlinganiselo we-axis-length ayikho) noma isakhiwo ephelele (Ukuhla lwe-tensor akuyona). None None Ngaphandle kwe-tf.RaggedTensor, izakhiwo ezivela kuphela ku-TensorFlow API ye-symbolic, i-graph-building: Ukubuyekezwa I-Hard Functional API yokusebenza. Okuningi DTypes Ukuhlobisa Ukubuyekeza a I-Data Type isebenzisa i- imikhiqizo tf.Tensor Tensor.dtype Ukwakhiwa a ukusuka ku-object ye-Python, ungathola i-data type. tf.Tensor Uma akufanele, i-TensorFlow uye ukhethe uhlobo idatha enokuthumela idatha yakho. I-TensorFlow i-Python integer yakhelwe ku futhi i-Python i-floating point numbers ukuze Ngaphandle kwalokho, i-TensorFlow isetshenziselwa izinhlelo ezivamile ezisetshenziselwa i-NumPy ekuguqulwa ku-array. tf.int32 tf.float32 Ungathola ukusuka ku-type kuya ku-type. the_f64_tensor = tf.constant([2.2, 3.3, 4.4], dtype=tf.float64) the_f16_tensor = tf.cast(the_f64_tensor, dtype=tf.float16) # Now, cast to an uint8 and lose the decimal precision the_u8_tensor = tf.cast(the_f16_tensor, dtype=tf.uint8) print(the_u8_tensor) tf.Tensor([2 3 4], shape=(3,), dtype=uint8) Ukuhlobisa Ukubuyekezwa kuyinto umqondo owenziwe kusuka ku- Ngokuvamile, ngaphansi kwezimo ezithile, ama-tensors ezincinane zihlanganisa ngokuvamile ukuze zihlanganise ama-tensors ezincinane lapho zihlanganisa izinsizakalo ezihlanganisiwe. umphumela we-Equivalent in NumPy Isibonelo esisodwa futhi esithakazelisayo kuyinto lapho udinga ukuguqulwa noma ukongeza isilinganiso ku-scalar. Kulesi isilinganiso, isilinganiso esithunyelwe ukuba yintoni esifanayo ne-argument eyodwa. x = tf.constant([1, 2, 3]) y = tf.constant(2) z = tf.constant([2, 2, 2]) # All of these are the same computation print(tf.multiply(x, 2)) print(x * y) print(x * z) tf.Tensor([2 4 6], shape=(3,), dtype=int32) tf.Tensor([2 4 6], shape=(3,), dtype=int32) tf.Tensor([2 4 6], shape=(3,), dtype=int32) Ngaphezu kwalokho, isixazululo esikhulu se-1 ingahlukaniswa ukuze zihlanganise nezinye iziqu ze-arguments. Zonke iziqu ze-arguments zingahlukaniswa ngama-calculation efanayo. Kulokhu, i-matrix ye-3x1 iyahlukaniswa ngama-element ngokuvamile nge-matrix ye-1x4 ukukhiqiza i-matrix ye-3x4. Qaphela ukuthi i-leading 1 iyatholakala okungagunyaziwe: Umbala we-y . [4] # These are the same computations x = tf.reshape(x,[3,1]) y = tf.range(1, 5) print(x, "\n") print(y, "\n") print(tf.multiply(x, y)) tf.Tensor( [[1] [2] [3]], shape=(3, 1), dtype=int32) tf.Tensor([1 2 3 4], shape=(4,), dtype=int32) tf.Tensor( [[ 1 2 3 4] [ 2 4 6 8] [ 3 6 9 12]], shape=(3, 4), dtype=int32) A broadcasted add: a times a gives a [3, 1] [1, 4] [3,4] Ngiyazi i-operation efanayo ngaphandle kokudluliselwa: x_stretch = tf.constant([[1, 1, 1, 1], [2, 2, 2, 2], [3, 3, 3, 3]]) y_stretch = tf.constant([[1, 2, 3, 4], [1, 2, 3, 4], [1, 2, 3, 4]]) print(x_stretch * y_stretch) # Again, operator overloading tf.Tensor( [[ 1 2 3 4] [ 2 4 6 8] [ 3 6 9 12]], shape=(3, 4), dtype=int32) Ngokuvamile, ukudluliselwa kuyinto isikhathi futhi indawo efanelekayo, njengoba ukusebenza ukudluliselwa akufanele ukufakelwa tensors ku-memory. You can see what broadcasting looks like usebenzisa . tf.broadcast_to print(tf.broadcast_to(tf.constant([1, 2, 3]), [3, 3])) tf.Tensor( [[1 2 3] [1 2 3] [1 2 3]], shape=(3, 3), dtype=int32) Ngokungafani nomathemikhali op, isibonelo, Ukusebenza okungenani okungenani okungenani okungenani okungenani okungenani okungenani okungenani okungenani okungenani okungenani. broadcast_to Ngaba kungenziwa ngakumbi kakhulu. Umbhali we-Jake VanderPlas Ukubonisa izintambo ezininzi zokuhamba (ngcono ku-NumPy). Isigaba Umhlahlandlela we-Python Data Science tf.convert_to ku_tensor Okuningi ops, like Waze Thola Arguments of Class Kodwa-ke, uzothola imiphumela emangalisayo, ama-Python ama-objects ebonakalayo njenge-tensors zihlanganisa. tf.matmul tf.reshape tf.Tensor Iningi, kodwa akuyona zonke, i-Ops Call ku-non-tensor arguments. Kukhona isiteshi se-conversions, futhi ama-object classes, njenge-NumPy's Ngena ngemva , izibuyekezo ze-Python, futhi Zonke izindlela zokusebenza ngokushesha. convert_to_tensor ndarray TensorShape tf.Variable Waze Ukuze uthole okwengeziwe, futhi uma unayo uhlobo yakho yakho ufuna ukuguqulwa ngokushesha ku-tensor. tf.register_tensor_conversion_function Ukuhlobisa Tensors I-tensor enezinhlayiyana eziningana nezinhlayiyana eziningana nezinhlayiyana zibizwa ngokuthi i-"ragged". Ukuze RAGED data. tf.ragged.RaggedTensor Ngokwesibonelo, This ayikwazi ukubonisa njenge-tensor ebonakalayo: A , shape: tf.RaggedTensor [4, None] ragged_list = [ [0, 1, 2, 3], [4, 5], [6, 7, 8], [9]] try: tensor = tf.constant(ragged_list) except Exception as e: print(f"{type(e).__name__}: {e}") ValueError: Can't convert non-rectangular Python sequence to Tensor. Ukwenza A Ukusebenzisa : tf.RaggedTensor tf.ragged.constant ragged_tensor = tf.ragged.constant(ragged_list) print(ragged_tensor) Umbala we-a will contain a few axes with unknown lengths: tf.RaggedTensor print(ragged_tensor.shape) (4, None) Ukuhlobisa I-A , okungenani ungakwazi ukubonisa idatha njenge-strings (i-variable-length byte array) ku-tensors. tf.string dtype I-strings iyona-atomic futhi ayikwazi ukufakelwa ngokufana ne-Python strings. Ubude we-string ayikho omunye ama-axis ye-tensor. Funda kabanzi izidingo zokusebenza kubo. tf.strings Ngiya ku-Scalar String Tensor: # Tensors can be strings, too here is a scalar string. scalar_string_tensor = tf.constant("Gray wolf") print(scalar_string_tensor) tf.Tensor(b'Gray wolf', shape=(), dtype=string) I-vector ye-strings ye- A vector of strings, shape: [3,] # If you have three string tensors of different lengths, this is OK. tensor_of_strings = tf.constant(["Gray wolf", "Quick brown fox", "Lazy dog"]) # Note that the shape is (3,). The string length is not included. print(tensor_of_strings) tf.Tensor([b'Gray wolf' b'Quick brown fox' b'Lazy dog'], shape=(3,), dtype=string) Ngaphezu kwalokho, i-print I-prefix ibonisa ukuthi dtype is a unicode string, kodwa i-byte-string. Ukuze uthole okwengeziwe mayelana nokusebenza ne-unicode text ku-TensorFlow. b tf.string I-Unicode Tutorial Uma uxhumane ama-unicode ama-characters, kukhona ama-utf-8 encoded. tf.constant("🥳👍") Izinhlelo ezithakazelisayo ze-strings zingatholakala ku- Ukubambisana . tf.strings tf.strings.split # You can use split to split a string into a set of tensors print(tf.strings.split(scalar_string_tensor, sep=" ")) tf.Tensor([b'Gray' b'wolf'], shape=(2,), dtype=string) # ...but it turns into a `RaggedTensor` if you split up a tensor of strings, # as each string might be split into a different number of parts. print(tf.strings.split(tensor_of_strings)) Three strings split, shape: [3, None] Yini : tf.strings.to_number text = tf.constant("1 10 100") print(tf.strings.to_number(tf.strings.split(text, " "))) tf.Tensor([ 1. 10. 100.], shape=(3,), dtype=float32) Nangona ungasebenzisa Ukuguqulwa kwe-string tensor ku-number, ungakwazi ukuguqulwa ku-byte, bese ku-number. tf.cast byte_strings = tf.strings.bytes_split(tf.constant("Duck")) byte_ints = tf.io.decode_raw(tf.constant("Duck"), tf.uint8) print("Byte strings:", byte_strings) print("Bytes:", byte_ints) Byte strings: tf.Tensor([b'D' b'u' b'c' b'k'], shape=(4,), dtype=string) Bytes: tf.Tensor([ 68 117 99 107], shape=(4,), dtype=uint8) # Or split it up as unicode and then decode it unicode_bytes = tf.constant("アヒル 🦆") unicode_char_bytes = tf.strings.unicode_split(unicode_bytes, "UTF-8") unicode_values = tf.strings.unicode_decode(unicode_bytes, "UTF-8") print("\nUnicode bytes:", unicode_bytes) print("\nUnicode chars:", unicode_char_bytes) print("\nUnicode values:", unicode_values) Unicode bytes: tf.Tensor(b'\xe3\x82\xa2\xe3\x83\x92\xe3\x83\xab \xf0\x9f\xa6\x86', shape=(), dtype=string) Unicode chars: tf.Tensor([b'\xe3\x82\xa2' b'\xe3\x83\x92' b'\xe3\x83\xab' b' ' b'\xf0\x9f\xa6\x86'], shape=(5,), dtype=string) Unicode values: tf.Tensor([ 12450 12498 12523 32 129414], shape=(5,), dtype=int32) Waze i-dtype isetshenziselwa zonke ama-byte yedatha e-TensorFlow. Module iqukethe umsebenzi ukuguqulwa idatha ku-byte kanye kusuka ku-byte, kuhlanganise ukucubungula izithombe kanye nokuguqulwa kwe-csv. tf.string tf.io Ukuhlobisa Tensile Ngezinye izikhathi, idatha yakho kuyinto emangalisayo, njenge-embedding indawo eningi kakhulu. I-TensorFlow isekelwe kanye nezinsizakalo ezihlobene ukugcina idatha e-sparse ngokushesha. tf.sparse.SparseTensor A , shape: tf.SparseTensor [3, 4] # Sparse tensors store values by index in a memory-efficient manner sparse_tensor = tf.sparse.SparseTensor(indices=[[0, 0], [1, 2]], values=[1, 2], dense_shape=[3, 4]) print(sparse_tensor, "\n") # You can convert sparse tensors to dense print(tf.sparse.to_dense(sparse_tensor)) SparseTensor(indices=tf.Tensor( [[0 0] [1 2]], shape=(2, 2), dtype=int64), values=tf.Tensor([1 2], shape=(2,), dtype=int32), dense_shape=tf.Tensor([3 4], shape=(2,), dtype=int64)) tf.Tensor( [[1 0 0 0] [0 0 2 0] [0 0 0 0]], shape=(3, 4), dtype=int32) Okwakhiwa kwebhizinisi le-TensorFlow, le ncwadi ifumaneke lapha ngaphansi kwebhizinisi elisha futhi isetshenziswe ngaphansi kwe-CC BY 4.0. Isampula se-codes ifakwe ngaphansi kwe-Apache 2.0 License. Okwakhiwa kwebhizinisi le-TensorFlow, le ncwadi ifumaneke lapha ngaphansi kwebhizinisi elisha futhi isetshenziswe ngaphansi kwe-CC BY 4.0. Isampula se-codes ifakwe ngaphansi kwe-Apache 2.0 License. Iwebhusayithi ye-TensorFlow