Question about 1-cycle policy

[Z-chocking]

Hi,thanks for your great work.
I am a newbie on Deep Learning,i have some question about the paper.
In your paper ,you said that :
We use the 1-cycle policy for the learning rate with max_lr = 3.5 × 10−4 , linear warm-up from max_lr/25 to max lr for the first 30% of iterations followed by cosine annealing to max_lr/75.

and the code in your work is:

I find that the learning rate is not increasing with the linear function but the cosine-like funtion.
I wrote a code following the setting in your work,and visualized the learning rate change.Here is the code and the result respectively

import torch
from torch.optim.lr_scheduler import CosineAnnealingLR, CosineAnnealingWarmRestarts,StepLR, OneCycleLR
import torch.nn as nn
from torchvision.models import resnet18
import matplotlib.pyplot as plt

if __name__ == '__main__':
    model = resnet18(pretrained=False)
    optimizer = torch.optim.SGD(model.parameters(), lr=0.1)
    mode = 'OneCycleLR'

    max_epoch = 100
    iters = 200

    scheduler = OneCycleLR(optimizer, max_lr=10, steps_per_epoch=iters, epochs=max_epoch, pct_start=0.3,
                               div_factor=10,final_div_factor=100,verbose=True,three_phase =False,
                               anneal_strategy='cos',
                                cycle_momentum = True,base_momentum=0.85, max_momentum=0.95)

    plt.figure()
    cur_lr_list = []
    for epoch in range(max_epoch):
        for batch in range(iters):
            optimizer.step()
            scheduler.step()
        cur_lr = optimizer.param_groups[-1]['lr']
        cur_lr_list.append(cur_lr)
        # print('Cur lr:', cur_lr)
    x_list = list(range(len(cur_lr_list)))
    plt.plot(x_list, cur_lr_list)
    plt.show()

I am so confused about this question,could you tell me the reason?Thanks.