change default max runtime to 300 and maximum to 1h

print for debugging

reinstate exact ppl

remove exact PPL because too long

add bigger ESM models

dynamic max job duration

remove GPU decorators from entire pipeline, only for GPU-heavy subtasks

et time limit at 1000

change emoji

mention companion spaces

add citation

insert link to ESM's paper

description text

restore return statement

introduce approximate PPL through mask batching

fix number of layers in 650M loading

fix 650M inclusion

include 650M model

fix locally and implement PPL

pass model and tokenizer as input to pipeline and add batch_size

same

fix selection of default ESM model

fix spaces decorator

remove old esm package dependency

replace fair-esm model access with huggingface hub, modularize and simplify post-processing

add spaces decorator

fix biopython import

remove gradio version from requirements

first try ESM app

initial commit