Can you give me some advice on training the model？

[BayMaxBHL]

After setting up the environment, I used NYU dataset for training, but the training results were very strange. The loss function converged slowly, rmse kept increasing, and delta kept decreasing.

[BayMaxBHL]

[BayMaxBHL]

[BayMaxBHL]

[BayMaxBHL]

[BayMaxBHL]

[BayMaxBHL]

The only changes I made to the code were to write a script to generate the CSV with a batch-size of 8.

[haifengwu205]

After setting up the environment, I used NYU dataset for training, but the training results were very strange. The loss function converged slowly, rmse kept increasing, and delta kept decreasing.
Can you share your code, I can't run it with the current

[haifengwu205]

@BayMaxBHL Hello，can you run the test code? I run it with a size mismatch, as follows:
size mismatch for net.coords: copying a param with shape torch.Size([8, 480, 640, 2]) from checkpoint, the shape in current model is torch.Size([16, 480, 640, 2]).

[BayMaxBHL]

@haifengwu205 确实是用不了，我这不晒出来的结果就是不收敛嘛。rmse还卡卡往上涨，人都麻了。

[hutingz]

@haifengwu205 确实是用不了，我这不晒出来的结果就是不收敛嘛。rmse还卡卡往上涨，人都麻了。

请问这个代码用的是多大数据的NYU呀

[macromogic]

Hi. I'm also trying to train the model myself and it does not converge as well. Does anyone have some solution so far? Thanks in advance!

[zhaorui-tan]

It's a similar situation to me. I used NYU dataset, and the result crashed as hell. It seems the model learned a completed wrong thing.

[jianqiaowang-wjq]

Why did I not output any results after using the train.sh script?