dtennant / cl-visualizing-feature-transformation Goto Github PK

I am a freshman in self-supervised learning. I run this repo just following the readme python main_contrast.py --method MoCov2 --data_folder your/path/to/imagenet-1K/dataset --dataset imagenet --epochs 200 --input_res 224 --cosine --batch_size 256 --learning_rate 0.03 --mixnorm --mixnorm_target posneg --sep_alpha --pos_alpha 2.0 --neg_alpha 1.6 --mask_distribution beta --expolation_mask --alpha 0.999 --multiprocessing-distributed --world-size 1 --rank 0 --save_score and find the speed is much slower than the offcial moco and moco_v2. I don't know what I did wrong.

Hello, I want to know where the feature transformation method is located in the code. Second, I want to know how to add this to simsiam

Hi, thanks for sharing the code.
I just tried to reproduce the result of mocov2 on IN-100 following the README.

python main_contrast.py --method MoCov2 --data_folder your/path/to/imagenet-1K/dataset  --dataset imagenet100  --imagenet100path your/path/to/imagenet100.class  --epochs 200 --input_res 224 --cosine --batch_size 256 --learning_rate 0.03   --mixnorm --mixnorm_target posneg --sep_alpha --pos_alpha 2.0 --neg_alpha 1.6 --mask_distribution beta --expolation_mask --alpha 0.999 --multiprocessing-distributed --world-size 1 --rank 0 --save_score

The loss value goes up at the beginning of training (this is normal behavior since the queue is being filled), but it stays at the value 11.09 and does not drop.
I wonder if it's the expected behavior? or can you provide a training log of this case? Thanks in advance.