Pytorch训练杂记

最新推荐文章于 2024-07-24 15:22:35 发布

Mondayismyday

最新推荐文章于 2024-07-24 15:22:35 发布

阅读量70

点赞数

文章标签： pytorch 深度学习人工智能

本文链接：https://blog.csdn.net/weixin_43999401/article/details/129105511

版权

传参

*args 不确定参数个数，以元组形式传入
**args 不确定参数个数，以字典形式传入
e.g.

def create_model(opt, step=0, **opt_kwargs):
    if local_config is not None:
        opt['path']['pretrain_model_G'] = os.path.join(local_config.checkpoint_path, os.path.basename(opt['path']['results_root'] + '.pth'))

    for k, v in opt_kwargs.items():
        opt[k] = v

    model = opt['model']
    model = model.split("-")[0]

    M = find_model_using_name(model)

    m = M(opt, step)
    logger.info('Model [{:s}] is created.'.format(m.__class__.__name__))
    return m

functional.partial

官方文档
functools.partial(func, /, *args, **keywords)

def partial(func, /, *args, **keywords):
    def newfunc(*fargs, **fkeywords):
        newkeywords = {**keywords, **fkeywords}
        return func(*args, *fargs, **newkeywords)
    newfunc.func = func
    newfunc.args = args
    newfunc.keywords = keywords
    return newfunc

在这里插入图片描述

RRDB_block_f = functools.partial(RRDB, nf=nf, gc=gc)

def find_model_using_name(model_name):
    # Given the option --model [modelname],
    # the file "models/modelname_model.py"
    # will be imported.
    model_filename = "models." + model_name + "_model"
    modellib = importlib.import_module(model_filename)

    # In the file, the class called ModelNameModel() will
    # be instantiated. It has to be a subclass of torch.nn.Module,
    # and it is case-insensitive.
    model = None
    target_model_name = model_name.replace('_', '') + 'Model'
    for name, cls in modellib.__dict__.items():
        if name.lower() == target_model_name.lower():
            model = cls

    if model is None:
        print(
            "In %s.py, there should be a subclass of torch.nn.Module with class name that matches %s." % (
            model_filename, target_model_name))
        exit(0)

    return model

Pytorch固定部分参数

参考
 参考1
参考2
1.requires_grad设置为False
2.过滤优化器的参数

for k,v in model2.named_parameters():
    if k in Layer1pre.keys():
        v.requires_grad = False

params = filter(lambda p: p.requires_grad, model.parameters())
criterion = nn.CrossEntropyLoss()
# optimizer = torch.optim.Adam(model2.parameters(), lr=learning_rate)
optimizer = torch.optim.Adam(params, lr=learning_rate)

冻结之后再打开

for k,v in model.named_parameters():
        v.requires_grad=True  # 固定层打开
optimizer = optim.Adam(model.parameters(),lr=0.01)
scheduler = torch.optim.lr_scheduler.StepLR(optimizer,step_size=10, gamma=0.5)

或者

if epoch==50:
    for parameter in deeplabv3.classifier.parameters():
        if parameter.requires_grad==False:
            parameter.requires_grad = True
            optimizer.add_param_group({'params': parameter})