在之前的一個問題中，我詢問了有關更快地將項目分配給數組的建議。從那時起，我取得了一些進展，例如，我擴展了推薦的版本來處理 3-D 數組，其目的是類似于神經網絡后續訓練數據的批量大小：import numpy as npimport timebatch_dim = 2first_dim = 5second_dim = 7depth_dim = 10upper_count = 5000toy_dict = {k:np.random.random_sample(size = depth_dim) for k in range(upper_count)}a = np.array(list(toy_dict.values()))def create_input_3d(orig_arr): print("Input shape:", orig_arr.shape) goal_arr = np.full(shape=(batch_dim, orig_arr.shape[1], orig_arr.shape[2], depth_dim), fill_value=1234, dtype=float) print("Goal shape:", goal_arr.shape) idx = np.indices(orig_arr.shape) print("Idx shape", idx.shape) goal_arr[idx[0], idx[1], idx[2]] = a[orig_arr[idx[0], idx[1], idx[2]]] return goal_arrorig_arr_three_dim = np.random.randint(0, upper_count, size=(batch_dim, first_dim, second_dim))orig_arr_three_dim.shape # (2,5,7)reshaped = create_input_3d(orig_arr_three_dim)然后，我決定創建一個自定義層以提高性能并即時進行轉換（減少內存）：import tensorflow as tffrom tensorflow import kerasimport numpy as np#custom layerclass CustLayer(keras.layers.Layer): def __init__(self, info_matrix, first_dim, second_dim, info_dim, batch_size): super(CustLayer, self).__init__() self.w = tf.Variable( initial_value=info_matrix, trainable=False, dtype=tf.dtypes.float32 ) self.info_dim = info_dim self.first_dim = first_dim self.second_dim = second_dim self.batch_size = batch_size由于高級索引（如我第一個發布的代碼中）不起作用，我回到了天真的 for 循環 - 這太慢了。我正在尋找的是一種使用第一個代碼片段中所示的高級索引的方法，并將其重新編程為 tf 兼容。這讓我以后能夠使用 GPU 進行學習。簡而言之：輸入的形狀為(batch_size, first_dim, second_dim)，返回的形狀為(batch_size, first_dim, second_dim, info_dim)，擺脫了緩慢的 for 循環。提前致謝。
1 回答
呼喚遠方
TA貢獻1856條經驗 獲得超11個贊
對于其他尋找答案的人來說，這就是我最終想出的：
import tensorflow as tf
from tensorflow import keras
import numpy as np
import time
class CustLayer(keras.layers.Layer):
def __init__(self, info_matrix, first_dim, second_dim, info_dim, batch_size):
super(CustLayer, self).__init__()
self.w = tf.Variable(
initial_value=info_matrix,
trainable=False,
dtype=tf.dtypes.float32
)
self.info_matrix = info_matrix
self.info_dim = info_dim
self.first_dim = first_dim
self.second_dim = second_dim
self.batch_size = batch_size
def my_numpy_func(self, x):
# x will be a numpy array with the contents of the input to the
# tf.function
shape = x.shape
goal_arr = np.zeros(shape=(shape[0], shape[1], shape[2], self.info_dim), dtype=np.float32)
# indices to expand
idx = np.indices(shape)
goal_arr[idx[0], idx[1], idx[2]] = self.info_matrix[x[idx[0], idx[1], idx[2]]]
shape_arr = np.array([shape[0], shape[1], shape[2]], dtype=np.int8)
#tf.print("Shape:", shape)
#tf.print("Shape_arr:", shape_arr)
#tf.print("Type:",type(shape_arr))
return goal_arr, shape_arr
@tf.function(input_signature=[tf.TensorSpec((None, 39, 25), tf.int64)])
def tf_function(self, input):
y, shape_arr = tf.numpy_function(self.my_numpy_func, [input], [tf.float32, tf.int8], "Nameless")
#tf.print("shape_arr", shape_arr)
y = tf.reshape(y, shape=(shape_arr[0], shape_arr[1], shape_arr[2], self.info_dim))
return y
def call(self, orig_arr):
return self.tf_function(orig_arr)
注意事項：在 GPU 上運行，但不能在 TPU 上運行。
添加回答
舉報
0/150
提交
取消