During training process, there is a size error

princeton-vl / RAFT-Stereo

MIT License

697 stars 136 forks source link

During training process, there is a size error #42

Open hillaric opened 2 years ago

hillaric commented 2 years ago

When I run "python3 train_stereo.py --batch_size 8 --train_iters 22 --valid_iters 32 --spatial_scale -0.2 0.4 --saturation_range 0 1.4 --n_downsample 2 --num_steps 200000 --mixed_precision --train_datasets middlebury_F --restore_ckpt models/raftstereo-sceneflow.pth --num_steps 5000"，here is the error:

lahavlipson commented 2 years ago

I can't reproduce this issue. From the root directory, I can obtain the file information for a Middlebury GT using the following command:

(raftstereo) llipson@node803 RAFT-Stereo$ identify datasets/Middlebury/MiddEval3/trainingF/PianoL/disp0GT.pfm
datasets/Middlebury/MiddEval3/trainingF/PianoL/disp0GT.pfm PFM 2828x1924 2828x1924+0+0 32-bit Grayscale Gray 20.7561MiB 0.160u 0:00.194

Do you get the same result?

hillaric commented 2 years ago

yes, I got the same result, but I still have the same error.

lahavlipson commented 2 years ago

Are you setting test_mode=True in the forward pass? I get the same error in this case.