RepVGG

RepVGG copied to clipboard

Thanks for your excellent work.

hopef

Congratulations you guys make a big progress in deep learning nerual nets. I found an interesting thing in the model initialization, the diff is so large when BN module initialized...

2h4dl

Pre-trained model on segmentation

5

Nice work. Could you share your pre-trained model for segmentation, .e.g RepVGG-B1g2-fast. Looking forward to your replay~

githubBingoChen

Hi Thank you for sharing your amazing work. I have a few questions and concerns: 1. Is it true that the accuracy of the training model and the re-parameterized model...

hosytuyen

多分支存在非线性的合并

1

请问如果多分支内部有激活函数、门控等非线性操作的话也可以合并吗？提前感谢解答！

piaohe20221128

LKAMING97

a mistake in code

1

‘config.py’ line178 config.DATA.TRAIN_DATA_SIZE = args.train_data_path I think it should be 'config.DATA.TRAIN_DATA_PATH = args.train_data_path'

Halance

I want to do "jit.trace " to scripting repvggplus. But there is a error in "class SEBlock": ``` x = F.avg_pool2d(inputs, kernel_size=inputs.size(3)) TypeError: avg_pool2d(): argument 'kernel_size' must be tuple of...

nistarlwc

关于多卡训练卡死的问题

1

对代码进行了修改以训练自己数据，发现多卡训练总是不能进行下去，单卡是没问题。调试发现是卡在了下面这一步。 `msg = model.load_state_dict(checkpoint, strict=False)`  请问是为什么呢？ 训练命令为： `CUDA_VISIBLE_DEVICES=0,1 python -m torch.distributed.run --nproc_per_node 2 --master_port 12345 main.py`

SEUvictor

anyone use repvgg backbone in insightface face recognition task?

1

xiakj