soumyasanyal

local train_size = 70108;
local batch_size = 4;
local num_gradient_accumulation_steps = 4;
local num_epochs = 4;
local learning_rate = 1e-5;
local weight_decay = 0.1;
local warmup_ratio = 0.06;
local transformer_model = "roberta-large";
local max_pieces = 384;
local dataset_dir = "original/depth-3/";

soumyasanyal

import os, tarfile
from subprocess import call
from os.path import isfile, join, exists
from tqdm import tqdm

# list of variables used
# datapath contains data as follows:
# ../data/protein/CASP/
#   |- CASP11 (unzipped)
#   |- CASP12 (unzipped)

soumyasanyal

import pickle
import functools as f
from pymatgen import MPRester
from copy import deepcopy

API_KEY = "YOUR API KEY"
mpr = MPRester(API_KEY)

print('Querying MP database...')
entries = mpr.query({