Computes and stores the average and current value

ResNet Variants Parameters ---------- block : Block Class for the residual block. Options are BasicBlockV1, BottleneckV1. laye

Jigswa + linear + l2norm

From: https://hips.seas.harvard.edu/blog/2013/03/03/the-alias-method-efficient-sampling-with-many-discrete-outcomes/

blur a single image on CPU

Folder datasets which returns the index of the image (for memory_bank)

Dummy RGB transfer.

Convert RGB PIL image to ndarray YDbDr.

trainer for contrastive pretraining

Gaussian Blur version 2

Jigsaw style crop

trainer for Linear evaluation

Split-Attention Conv2d

transform a group of images independently

Base Memory Class

base class for MoCo-style memory cache

class for BaseTrainer

ResNet Bottleneck

Memory bank for two modalities, e.g. in CMC

MoCo-style memory for two modalities, e.g. in CMC

CMC model with Multiple linear/mlp projection heads

CMC model with a single linear/mlp projection head

blur a batch of images on CPU or GPU

Convert RGB PIL image to ndarray Lab.

Convert RGB PIL image to ndarray YCbCr.

Convert RGB PIL image to ndarray YIQ.

Convert RGB PIL image to ndarray YPbPr.

Memory bank for single modality

Single Modal (e.g., RGB) MoCo-style cache

RGB model with Multiple linear/mlp projection heads

RGB model with a single linear/mlp projection head

rotation