Thank you for your contribution to the MindOCR repo.
Before submitting this PR, please make sure:

• You have read the Contributing Guidelines on pull requests
• Your code builds clean without any errors or warnings
• You are using approved terminology
• You have added unit tests

Motivation

When resuming training, the correct logging should be [cur_epoch/(training_epochs+start_epochs)] to avoid confusion.

It is possible that opt.get_lr() is not a single scalar but a list of scalars. When users pass a list of dictionaries to the optimizer, for example:

[{  'params': group_base_params , 'lr': base_lr },   { 'params': group_small_params, 'lr': small_lr}]

The visualization should print each distinct lr.

Test Plan

1. Try to resume training when start_epoch >1
2. In mindocr/optim/param_grouping.py, create a new grouping function that returns a list of dictionaries containing different lrs, for example:

def grouping_params(params, filter_name, learning_rate):
    base_params, small_params = [],  []
    for param in params:
        if filter_name in param.name:
            base_params.append(param)
        else:
            small_params.append(param)
    return [{  'params': base_params, 'lr': learning_rate}, 
                { 'params': small_params, 'lr': learning_rate*0.1}]

Related Issues and PRs

NA