田运杰 comments

Results 40 comments of


                                            田运杰

Experimental hardware device and the pre-training model

How about the evaluation speed please?

ConvMAE flops for classification ?

Could you provide the flops during finetuning for classification please?

Using for Tree skeleton

应该可以

Accessing Baidu Datasets and Re-evaluation Code outside China

https://drive.google.com/file/d/12ka227h__2B1r7-Ztd_6U41XXoYskFkQ/view?usp=sharing https://drive.google.com/file/d/1VYbKyDdobyG_4fTyjTMKrhT5m4dBAJO5/view?usp=sharing

Accessing Baidu Datasets and Re-evaluation Code outside China

https://drive.google.com/file/d/11Om7YJOyEenbFFtt6ATMnFrhSxSM1o1k/view?usp=sharing https://drive.google.com/file/d/1C9osP7b2exs4TaWZrTieZDUtvhrMIAKK/view?usp=sharing

RuntimeError: Expected a 'cuda' device type for generator but found 'cpu'

It looks like you didn't put the model or tensor on the GPU device. Can you provide more details？

RuntimeError: Expected a 'cuda' device type for generator but found 'cpu'

Based on this information above I still can’t determine where the problem is. And I did not encounter this error when I tried the code... Maybe it's a problem with...

default masking ratios for Zoomed-In, Distort, Blur,... are 0.75

We do not conduct masking for the newly proposed tasks. I will check the code recently.

default masking ratios for Zoomed-In, Distort, Blur,... are 0.75

In fact, the masking ratio is not used for the 5 designed tasks. Could you provide the log files of pre-training and fine-tuning?

default masking ratios for Zoomed-In, Distort, Blur,... are 0.75

Is this zoomed-in task? Could you provide the fine-tuning log file? Pre-trained models are in HUAWEI, which are prohibited to take out. But I could pre-train again with lab. machines,...