update pretrainer to not use distributed sampler (Trainer uses accelerate)

update with 12L and 20L i4096 gc95M models, multitask and quantiz code

move dicts to init

update kwargs for pretrainer

fix: incorrect condition control flow (#261)

Update pretrainer for transformers==4.28.0

add in silico perturbation module

Add data collator for cell classification and example for cell classification