torch VS tensorflow 同功能的api

1、裁剪

torch.clamp(a,-0.1,0.1) ==tf.clip()

2、复制

tf.tile(a,[1,2,1])和a.repeat(1,2,1)

3、增加维度

a.unsqueeze(0) == tf.expand_dim(a,axis=0)

4、L2归一化 

def normalize(x, axis=-1):
    x = 1. * x / (torch.norm(x, 2, axis, keepdim=True).expand_as(x).clamp_min(1e-12))
    return x
l2_normal_a =  normalize(a) == tf.l2_normal(a,axis=1)

5、交换维度

b.permute(1,0) == tf.transopose(b,[1,0])

6.类型转换

tf.cast(a,float16)==a.half()
a.int() a.double() a.float() a.bool()

6、常量

torch.tensor([[1,2,3],[4,5,6.0]]) == tf.constant ([[1,2,3],[4,5,6.0]])

7、torch.zeros_like ==tf. zeros_like   ones_like==ones_like  where=where

torch.mean=tf.reduce_mean()

8、torch.Tensor.detach() 与tf.stop_gradient() （待验证）

9、torch.nn.LogSoftmax 与 tf.nn.log_softmax()

torch.nn.LogSoftmax(dim=-1)(a)==tf.nn.log_softmax(a, axis=-1)