mae is same when training
SherlockHolmes221 opened this issue · comments
mae is same when training my own dataset
How many GPUs do you use?
I also have the same problem when using the jhu and NWPU datasets. I use one GPU.
I also have the same problem when using the jhu and NWPU datasets. I use one GPU.
Try to use my default command
Can you explain it in more detail? I don't quite understand what you said, thank you.
Can you explain it in more detail? I don't quite understand what you said, thank you.
Can you try to run with 4 GPUs, I haven't run with 1 GPU.
How many GPUs do you use?
1 GPU and only have one thus can not test on 4
Can you explain it in more detail? I don't quite understand what you said, thank you.
can not predict any person thus the mae not change
Can you explain it in more detail? I don't quite understand what you said, thank you.
Can you try to run with 4 GPUs, I haven't run with 1 GPU.
Okay, I'll give it a try, later. Thank you.
Can you explain it in more detail? I don't quite understand what you said, thank you.
can not predict any person thus the mae not change
May I ask if the same problem of Mae has been solved?
Can you explain it in more detail? I don't quite understand what you said, thank you.
can not predict any person thus the mae not change
May I ask if the same problem of Mae has been solved?
No, did not solve
I haven't tried to train on 1 GPU and may need to adjust some hyperparameters.
I haven't tried to train on 1 GPU and may need to adjust some hyperparameters.
Which parameters?
I also cannot reproduce the results using your repository.
I checked the training procedure using 1, 2, and 4 GPUs and all generated the same, unchanging MAE.
@SherlockHolmes221 did you solve it? Why did you mark it as completed without any answer?
Why did you mark it as completed without any answer? What are you avoiding??