比较与torch.nn.utils.clip_grad_value_的差异

查看源文件

torch.nn.utils.clip_grad_value_

torch.nn.utils.clip_grad_value_(parameters, clip_value)

更多内容详见torch.nn.utils.clip_grad_value_

mindspore.ops.clip_by_value

mindspore.ops.clip_by_value(x, clip_value_min=None, clip_value_max=None)

更多内容详见mindspore.ops.clip_by_value

差异对比

PyTorch中梯度是Tensor的属性,可以通过设置requires_grad=True使Tensor带有梯度。由于框架机制的不同,在MindSpore中,梯度和权重是互相独立的Tensor。因此在梯度裁剪时,MindSpore需要先获取梯度Tensor再进行裁剪。

PyTorch:实现梯度裁剪可以直接传入带有梯度的Tensor。

MindSpore:由于框架机制不同,实现梯度裁剪,需要先获取梯度,再对梯度进行裁剪。可以使用mindspore.grad等方法获取梯度,详情请参考梯度求导

分类

子类

PyTorch

MindSpore

差异

参数

参数1

parameters

x

梯度机制不同,PyTorch传入Tensor即可对梯度裁剪,MindSpore需要传入梯度Tensor,如何获取梯度请参考梯度求导

参数2

clip_value

clip_value_min

PyTorch裁剪范围为[-clip_value, clip_value],MindSpore裁剪范围为[clip_value_min, clip_value_max]

参数3

-

clip_value_max

PyTorch裁剪范围为[-clip_value, clip_value],MindSpore裁剪范围为[clip_value_min, clip_value_max]

代码示例

由于机制不同,MindSpore实现梯度裁剪需要先使用mindspore.grad等方法获取梯度(更多获取梯度的方法,请参考梯度求导),再对梯度进行裁剪,示例代码如下。

import numpy as np
import torch
import mindspore as ms
from mindspore.common.initializer import initializer, Zero

data = np.array([0.2, 0.5, 0.2], dtype=np.float32)
label = np.array([1, 0], dtype=np.float32)
label_pt = np.array([0], dtype=np.float32)

# PyTorch
class Net1(torch.nn.Module):
    def __init__(self):
        super(Net1, self).__init__()
        self.dense = torch.nn.Linear(3, 2)
        self.apply(self._init_weights)

    def _init_weights(self, module):
        if isinstance(module, torch.nn.Linear):
            module.weight.data.zero_()
            module.bias.data.zero_()

    def forward(self, x):
        x = self.dense(x)
        return x

net1 = Net1()
loss_fun = torch.nn.CrossEntropyLoss()
out = net1(torch.tensor(data))
out = torch.unsqueeze(out, 0)
loss = loss_fun(out, torch.tensor(label_pt, dtype=torch.long))
loss.backward()
grads = [p.grad for p in net1.parameters() if p.grad is not None]
print(grads)
# [tensor([[-0.1000, -0.2500, -0.1000],
#         [ 0.1000,  0.2500,  0.1000]]), tensor([-0.5000,  0.5000])]
torch.nn.utils.clip_grad_value_(net1.parameters(), clip_value=0.1)
print(grads)
# [tensor([[-0.1000, -0.1000, -0.1000],
#         [ 0.1000,  0.1000,  0.1000]]), tensor([-0.1000,  0.1000])]

# MindSpore
class Net2(ms.nn.Cell):
    def __init__(self):
        super(Net2, self).__init__()
        self.dense = ms.nn.Dense(3, 2)
        self.apply(self._init_weights)

    def _init_weights(self, cell):
        if isinstance(cell, ms.nn.Dense):
            cell.weight.set_data(initializer(Zero(), cell.weight.shape, cell.weight.dtype))
            cell.bias.set_data(initializer(Zero(), cell.bias.shape, cell.bias.dtype))

    def construct(self, x):
        return self.dense(x)

net2 = Net2()
loss_fn = ms.nn.CrossEntropyLoss()

def forward_fn(data, label):
    logits = ms.ops.squeeze(net2(data))
    loss = loss_fn(logits, label)
    return loss, logits

grad_fn = ms.grad(forward_fn, grad_position=None, weights=net2.trainable_params(), has_aux=True)
grads = grad_fn(ms.ops.unsqueeze(ms.Tensor(data), dim=0), ms.Tensor(label))
print(grads)
# ((Tensor(shape=[2, 3], dtype=Float32, value=
# [[-1.00000001e-01, -2.50000000e-01, -1.00000001e-01],
#  [ 1.00000001e-01,  2.50000000e-01,  1.00000001e-01]]), Tensor(shape=[2], dtype=Float32, value= [-5.00000000e-01,  5.00000000e-01])), (Tensor(shape=[2], dtype=Float32, value= [ 0.00000000e+00,  0.00000000e+00]),))
grads = ms.ops.clip_by_value(grads, clip_value_min=-0.1, clip_value_max=0.1)
print(grads)
# ((Tensor(shape=[2, 3], dtype=Float32, value=
# [[-1.00000001e-01, -1.00000001e-01, -1.00000001e-01],
#  [ 1.00000001e-01,  1.00000001e-01,  1.00000001e-01]]), Tensor(shape=[2], dtype=Float32, value= [-1.00000001e-01,  1.00000001e-01])), (Tensor(shape=[2], dtype=Float32, value= [ 0.00000000e+00,  0.00000000e+00]),))