kxqt
commented
6 months ago 我使用preprocess.ipynb脚本处理checkpoint之后，加载出现维度不匹配的情况。
RuntimeError: Error(s) in loading state_dict for DiffusionEngine:
size mismatch for model.diffusion_model.input_blocks.1.1.transformer_blocks.0.t_attn.to_k.weight: copying a param with shape torch.Size([320, 1024]) from checkpoint, the shape in current model is torch.Size([320, 2048]).
size mismatch for model.diffusion_model.input_blocks.1.1.transformer_blocks.0.t_attn.to_v.weight: copying a param with shape torch.Size([320, 1024]) from checkpoint, the shape in current model is torch.Size([320, 2048]).
size mismatch for model.diffusion_model.input_blocks.2.1.transformer_blocks.0.t_attn.to_k.weight: copying a param with shape torch.Size([320, 1024]) from checkpoint, the shape in current model is torch.Size([320, 2048]).
size mismatch for model.diffusion_model.input_blocks.2.1.transformer_blocks.0.t_attn.to_v.weight: copying a param with shape torch.Size([320, 1024]) from checkpoint, the shape in current model is torch.Size([320, 2048]).
size mismatch for model.diffusion_model.input_blocks.4.1.transformer_blocks.0.t_attn.to_k.weight: copying a param with shape torch.Size([640, 1024]) from checkpoint, the shape in current model is torch.Size([640, 2048]).
size mismatch for model.diffusion_model.input_blocks.4.1.transformer_blocks.0.t_attn.to_v.weight: copying a param with shape torch.Size([640, 1024]) from checkpoint, the shape in current model is torch.Size([640, 2048]).
size mismatch for model.diffusion_model.input_blocks.5.1.transformer_blocks.0.t_attn.to_k.weight: copying a param with shape torch.Size([640, 1024]) from checkpoint, the shape in current model is torch.Size([640, 2048]).
size mismatch for model.diffusion_model.input_blocks.5.1.transformer_blocks.0.t_attn.to_v.weight: copying a param with shape torch.Size([640, 1024]) from checkpoint, the shape in current model is torch.Size([640, 2048]).
size mismatch for model.diffusion_model.input_blocks.7.1.transformer_blocks.0.t_attn.to_k.weight: copying a param with shape torch.Size([1280, 1024]) from checkpoint, the shape in current model is torch.Size([1280, 2048]).
size mismatch for model.diffusion_model.input_blocks.7.1.transformer_blocks.0.t_attn.to_v.weight: copying a param with shape torch.Size([1280, 1024]) from checkpoint, the shape in current model is torch.Size([1280, 2048]).
size mismatch for model.diffusion_model.input_blocks.8.1.transformer_blocks.0.t_attn.to_k.weight: copying a param with shape torch.Size([1280, 1024]) from checkpoint, the shape in current model is torch.Size([1280, 2048]).
size mismatch for model.diffusion_model.input_blocks.8.1.transformer_blocks.0.t_attn.to_v.weight: copying a param with shape torch.Size([1280, 1024]) from checkpoint, the shape in current model is torch.Size([1280, 2048]).
size mismatch for model.diffusion_model.middle_block.1.transformer_blocks.0.t_attn.to_k.weight: copying a param with shape torch.Size([1280, 1024]) from checkpoint, the shape in current model is torch.Size([1280, 2048]).
size mismatch for model.diffusion_model.middle_block.1.transformer_blocks.0.t_attn.to_v.weight: copying a param with shape torch.Size([1280, 1024]) from checkpoint, the shape in current model is torch.Size([1280, 2048]).
size mismatch for model.diffusion_model.output_blocks.3.1.transformer_blocks.0.t_attn.to_k.weight: copying a param with shape torch.Size([1280, 1024]) from checkpoint, the shape in current model is torch.Size([1280, 2048]).
size mismatch for model.diffusion_model.output_blocks.3.1.transformer_blocks.0.t_attn.to_v.weight: copying a param with shape torch.Size([1280, 1024]) from checkpoint, the shape in current model is torch.Size([1280, 2048]).
size mismatch for model.diffusion_model.output_blocks.4.1.transformer_blocks.0.t_attn.to_k.weight: copying a param with shape torch.Size([1280, 1024]) from checkpoint, the shape in current model is torch.Size([1280, 2048]).
size mismatch for model.diffusion_model.output_blocks.4.1.transformer_blocks.0.t_attn.to_v.weight: copying a param with shape torch.Size([1280, 1024]) from checkpoint, the shape in current model is torch.Size([1280, 2048]).
size mismatch for model.diffusion_model.output_blocks.5.1.transformer_blocks.0.t_attn.to_k.weight: copying a param with shape torch.Size([1280, 1024]) from checkpoint, the shape in current model is torch.Size([1280, 2048]).
size mismatch for model.diffusion_model.output_blocks.5.1.transformer_blocks.0.t_attn.to_v.weight: copying a param with shape torch.Size([1280, 1024]) from checkpoint, the shape in current model is torch.Size([1280, 2048]).
size mismatch for model.diffusion_model.output_blocks.6.1.transformer_blocks.0.t_attn.to_k.weight: copying a param with shape torch.Size([640, 1024]) from checkpoint, the shape in current model is torch.Size([640, 2048]).
size mismatch for model.diffusion_model.output_blocks.6.1.transformer_blocks.0.t_attn.to_v.weight: copying a param with shape torch.Size([640, 1024]) from checkpoint, the shape in current model is torch.Size([640, 2048]).
size mismatch for model.diffusion_model.output_blocks.7.1.transformer_blocks.0.t_attn.to_k.weight: copying a param with shape torch.Size([640, 1024]) from checkpoint, the shape in current model is torch.Size([640, 2048]).
size mismatch for model.diffusion_model.output_blocks.7.1.transformer_blocks.0.t_attn.to_v.weight: copying a param with shape torch.Size([640, 1024]) from checkpoint, the shape in current model is torch.Size([640, 2048]).
size mismatch for model.diffusion_model.output_blocks.8.1.transformer_blocks.0.t_attn.to_k.weight: copying a param with shape torch.Size([640, 1024]) from checkpoint, the shape in current model is torch.Size([640, 2048]).
size mismatch for model.diffusion_model.output_blocks.8.1.transformer_blocks.0.t_attn.to_v.weight: copying a param with shape torch.Size([640, 1024]) from checkpoint, the shape in current model is torch.Size([640, 2048]).
size mismatch for model.diffusion_model.output_blocks.9.1.transformer_blocks.0.t_attn.to_k.weight: copying a param with shape torch.Size([320, 1024]) from checkpoint, the shape in current model is torch.Size([320, 2048]).
size mismatch for model.diffusion_model.output_blocks.9.1.transformer_blocks.0.t_attn.to_v.weight: copying a param with shape torch.Size([320, 1024]) from checkpoint, the shape in current model is torch.Size([320, 2048]).
size mismatch for model.diffusion_model.output_blocks.10.1.transformer_blocks.0.t_attn.to_k.weight: copying a param with shape torch.Size([320, 1024]) from checkpoint, the shape in current model is torch.Size([320, 2048]).
size mismatch for model.diffusion_model.output_blocks.10.1.transformer_blocks.0.t_attn.to_v.weight: copying a param with shape torch.Size([320, 1024]) from checkpoint, the shape in current model is torch.Size([320, 2048]).
size mismatch for model.diffusion_model.output_blocks.11.1.transformer_blocks.0.t_attn.to_k.weight: copying a param with shape torch.Size([320, 1024]) from checkpoint, the shape in current model is torch.Size([320, 2048]).
size mismatch for model.diffusion_model.output_blocks.11.1.transformer_blocks.0.t_attn.to_v.weight: copying a param with shape torch.Size([320, 1024]) from checkpoint, the shape in current model is torch.Size([320, 2048]).另外，请问是用8张A40在SynthText数据上训练100k iter需要48h吗？在LAION-OCR数据上训练100k iter需要另外48h？ 我还发现paper里面的batchsize是64，而config/train.yaml里面似乎是单卡batchsize为64，所以如果我想复现您的结果，需要改成总batchsize为64吗？ 感谢！
sokoloveai
ZYM-PKU
在训练一开始加载512-inpainting-ema.ckpt时，发现pertrained weight和model很多权重没有成功加载。请问这个是正常的吗？
另外想请教一下，训练的时候用了什么GPU，需要训练多少时间。感谢！