None of the inputs have requires_grad=True. Gradients will be None

[TigerHH6866]

because max_grad_norm is set, clip_grad_norm is enabled. consider set to 0 / max_grad_normが設定されているためclip_grad_normが有効になります。0に設定して無効にしたほうがいいかもしれません
running training / 学習開始
num train images * repeats / 学習画像の数×繰り返し回数: 1050
num reg images / 正則化画像の数: 0
num batches per epoch / 1epochのバッチ数: 525
num epochs / epoch数: 12
batch size per device / バッチサイズ: 2
gradient accumulation steps / 勾配を合計するステップ数 = 1
total optimization steps / 学習ステップ数: 6300
steps: 0%| | 0/6300 [00:00<?, ?it/s]
epoch 1/12
/root/miniconda3/lib/python3.10/site-packages/torch/utils/checkpoint.py:31: UserWarning: None of the inputs have requires_grad=True. Gradients will be None
warnings.warn("None of the inputs have requires_grad=True. Gradients will be None")

first one: max_grad_norm or clip_grad_norm where to set?
seccond one: None of the inputs have requires_grad=True. Gradients will be None
this seem to make result worng. how to set right??

[araleza]

I get both of those messages, and my training works fine. I don't think they're a problem, although it would be nice if they were removed via code so as not to worry people. Just ignore them for now.

[TigerHH6866]

ok thx