import os
import sys
sys.path.append("..") # add top folder to path
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import torch
import impepdom
import pickle
hla_allele = 'HLA-A01:01'
fold_idx = [0, 1, 2]
hla_a01_01 = impepdom.PeptideDataset(hla_allele, padding='after2', toy=True)
peploader = hla_a01_01.get_peptide_dataloader(fold_idx)
mlp = impepdom.MultilayerPerceptron()
folder = impepdom.train_nn(mlp,
peploader,
hla_allele,
fold_idx,
save_results=True)
predictions = mlp(torch.tensor(hla_a01_01.get_fold([3])[0]).float())
print(folder)
pred_path = os.path.join(folder, 'pred-3')
outfile = open(pred_path, 'wb')
pickle.dump(predictions, outfile)
outfile.close()
'HLA-B07:02', # >>> Khoi 2 <<<
'HLA-A24:02',
'HLA-B27:05', # >>> Michael
'HLA-A68:01' # <<< Michael
]
hla_alleles_test = ['HLA-A01:01']
impepdom.time_tracker.reset_timer() # start counting time
for i, hla_allele in enumerate(hla_alleles_test): # change allele list here
print(impepdom.time_tracker.now() +
'working with allele {0} out of {1}'.format(
i + 1, len(hla_alleles_test))) # change allele list here
dataset = impepdom.PeptideDataset(hla_allele=hla_allele,
padding='flurry',
toy=False)
best_config = impepdom.hyperparam_search(
model_type='MultilayerPerceptron',
dataset=dataset,
max_epochs=15,
batch_sizes=[32],
learning_rates=[1e-3],
dropout_input_list=[0.75, 0.65],
dropout_hidden_list=[0.50, 0.45],
conv_flags=[False],
num_conv_layers_list=[2],
conv_filt_sz_list=[5],
conv_stride_list=[1],
)
import os
import sys
sys.path.append("..") # add top folder to path
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import pickle
from sklearn.metrics import roc_auc_score
import torch
import impepdom
model = impepdom.models.MultilayerPerceptron(num_hidden_layers=2,
hidden_layer_size=100)
dataset = impepdom.PeptideDataset(hla_allele='HLA-A01:01',
padding='flurry',
toy=True)
folder, baseline_metrics, _ = impepdom.run_experiment(model,
dataset,
train_fold_idx=[1, 2, 3],
val_fold_idx=[0],
learning_rate=2e-3,
num_epochs=5,
batch_size=32)
trained_model, train_history = impepdom.load_trained_model(model, folder)
impepdom.plot_train_history(train_history, baseline_metrics)