Problems for loading state_dict.

[Cresynia]

Thanks for your code! And I met this problem when I evaluate the model. I want to know how to solve it or where I may make mistakes. Thanks!

[Cresynia]

here is the corresponding code.

[Cresynia]

I can evaluate it after I change the code: depth_decoder.load_state_dict({k: v for k, v in decoder_dict.items() if k in depth_decoder.state_dict()}) And the result is : I want to know if it is right. Thanks！

[myalos]

Sorry for late response, the value of a1 is supposed to be near 0.771. And from what you said, I feel that there are still some problems with my code. I will check it again later.

[Cresynia]

Thanks！

---Original--- From: @.> Date: Wed, Jan 31, 2024 08:14 AM To: @.>; Cc: @.**@.>; Subject: Re: [myalos/monodepth2nyu] Problems for loading state_dict. (Issue#3)

Sorry for late response, the value of a1 is supposed to be near 77.1. And from what you said, I feel that there are still some problems with my code. I will check it again later.

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>

[myalos]

sorry for late again, i was a little busy these day. And today i rerun my code, it started the training successfully, and i cut the training since it seems need a lot time. I run this code last year using only scale [0] getting the result 76.2. I didn't train the [0, 1, 2, 3] scale to reproduce the result in the paper since it is not environment friendly. To answer your question "I want to know if it is right.", I only know if you train this code using only scale [0], then you will get a1 76.2. And if train the scale [0, 1, 2, 3], i don't know what a1 result the model should get.

[Cresynia]

Happy New Year! Thanks for your reply! I'll try it.