Check the discrepancy between two resnet18 implementations

[snoop2head]

Performance between these code blocks are different.
https://github.com/snoop2head/ELimNet/blob/799a5c25199b925dc3bf035d95c81a5a9765ba0c/src/modules/resnet18.py#L9-L52

https://github.com/snoop2head/ELimNet/blob/799a5c25199b925dc3bf035d95c81a5a9765ba0c/train_resnet_elim.py#L31-L59

[snoop2head]

Checked that those two codes' results are identical regardless of number of layers.

• resnet_18 and resnet_18_elim_2 is code ran with train.py
• resnet_18(no yaml) and resnet_18_elim_2(no yaml) is code ran with train_resnet_elim.py

It's either the previous experimentation during P-Stage on Resnet18 is wrong: val f1 score of 0.71 is too low.
Or the change of FP32 in previous experiment to FP16 in current experiment affecting the model's performance.