【解决】MindSpore报错RuntimeError: For ’Optimizer’, the argument group params must not be empty.

经验分享 Tech Blogs 模型训练-Model Training
1

1系统环境

硬件环境(Ascend/GPU/CPU): Ascend/GPU/CPU
MindSpore版本: 不限
执行模式（PyNative/ Graph）: 不限
Python版本: Python=3.7.5
操作系统平台: Linux

2 报错信息

2.1问题描述

使用MindSpore中的mindspore.nn.Optimizer出现报错

2.2报错信息

RuntimeError: For ’Optimizer’, the argument group params must not be empty.

2.3脚本代码（代码格式，可上传附件）

from mindspore import nn

net = Net()
conv_params = list(filter(lambda x: 'conv' in x.name, net.trainable_params()))
no_conv_params = list(filter(lambda x: 'conv' not in x.name, net.trainable_params()))
group_params = [{'params': conv_params, 'lr': 0.05},{'params': no_conv_params, 'lr': 0.01}]
optim = nn.Momentum(group_params, learning_rate=0.1, momentum=0.9, weight_decay=0.0)
conv_lr = optim.get_lr_parameter(conv_params)
print(conv_lr[0].asnumpy())

3 根因分析

**报错原因为：params里面不能为空，实例代码中的隐含条件是，Net应该是一个包含卷积操作的网络。本示例代码中没有展示网络定义。
以下代码，构造一个没包含params的Net，可以看到报错：

RuntimeError: For ’Optimizer’, the argument group params must not be empty.

import numpy as np
import mindspore.ops as ops
from mindspore import nn, Tensor, Parameter


class Net(nn.Cell):
    def __init__(self):
        super(Net, self).__init__()
        self.matmul = ops.MatMul()

    def construct(self, x):
        out = self.matmul(x, x)
        return out


net = Net()
conv_params = list(filter(lambda x: 'conv' in x.name, net.trainable_params()))
no_conv_params = list(filter(lambda x: 'conv' not in x.name, net.trainable_params()))
group_params = [{'params': conv_params, 'lr': 0.05}, {'params': no_conv_params, 'lr': 0.01}]
optim = nn.Momentum(group_params, learning_rate=0.1, momentum=0.9, weight_decay=0.0)
conv_lr = optim.get_lr_parameter(conv_params)
print(conv_lr[0].asnumpy())

运行结果：

1111×341 38 KB

4 解决方案

修改代码如下，解决报错：

import numpy as np
import mindspore.ops as ops
from mindspore import nn, Tensor, Parameter


class Net(nn.Cell):
    def __init__(self):
        super(Net, self).__init__()
        self.matmul = ops.MatMul()
        self.conv = nn.Conv2d(1, 6, 5, pad_mode="valid")
        self.param = Parameter(Tensor(np.array([1.0], np.float32)))

    def construct(self, x):
        x = self.conv(x)
        x = x * self.param
        out = self.matmul(x, x)
        return out


net = Net()
conv_params = list(filter(lambda x: 'conv' in x.name, net.trainable_params()))
no_conv_params = list(filter(lambda x: 'conv' not in x.name, net.trainable_params()))
group_params = [{'params': conv_params, 'lr': 0.05}, {'params': no_conv_params, 'lr': 0.01}]
optim = nn.Momentum(group_params, learning_rate=0.1, momentum=0.9, weight_decay=0.0)
conv_lr = optim.get_lr_parameter(conv_params)
print(conv_lr[0].asnumpy())

代码运行结果：

550×312 18.9 KB