Upload 5 files
Browse files- .gitignore +160 -0
- app.py +43 -0
- requirements.txt +18 -0
- run.py +102 -0
- train.py +454 -0
.gitignore
ADDED
|
@@ -0,0 +1,160 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Byte-compiled / optimized / DLL files
|
| 2 |
+
__pycache__/
|
| 3 |
+
*.py[cod]
|
| 4 |
+
*$py.class
|
| 5 |
+
|
| 6 |
+
# C extensions
|
| 7 |
+
*.so
|
| 8 |
+
|
| 9 |
+
# Distribution / packaging
|
| 10 |
+
.Python
|
| 11 |
+
build/
|
| 12 |
+
develop-eggs/
|
| 13 |
+
dist/
|
| 14 |
+
downloads/
|
| 15 |
+
eggs/
|
| 16 |
+
.eggs/
|
| 17 |
+
lib/
|
| 18 |
+
lib64/
|
| 19 |
+
parts/
|
| 20 |
+
sdist/
|
| 21 |
+
var/
|
| 22 |
+
wheels/
|
| 23 |
+
share/python-wheels/
|
| 24 |
+
*.egg-info/
|
| 25 |
+
.installed.cfg
|
| 26 |
+
*.egg
|
| 27 |
+
MANIFEST
|
| 28 |
+
|
| 29 |
+
# PyInstaller
|
| 30 |
+
# Usually these files are written by a python script from a template
|
| 31 |
+
# before PyInstaller builds the exe, so as to inject date/other infos into it.
|
| 32 |
+
*.manifest
|
| 33 |
+
*.spec
|
| 34 |
+
|
| 35 |
+
# Installer logs
|
| 36 |
+
pip-log.txt
|
| 37 |
+
pip-delete-this-directory.txt
|
| 38 |
+
|
| 39 |
+
# Unit test / coverage reports
|
| 40 |
+
htmlcov/
|
| 41 |
+
.tox/
|
| 42 |
+
.nox/
|
| 43 |
+
.coverage
|
| 44 |
+
.coverage.*
|
| 45 |
+
.cache
|
| 46 |
+
nosetests.xml
|
| 47 |
+
coverage.xml
|
| 48 |
+
*.cover
|
| 49 |
+
*.py,cover
|
| 50 |
+
.hypothesis/
|
| 51 |
+
.pytest_cache/
|
| 52 |
+
cover/
|
| 53 |
+
|
| 54 |
+
# Translations
|
| 55 |
+
*.mo
|
| 56 |
+
*.pot
|
| 57 |
+
|
| 58 |
+
# Django stuff:
|
| 59 |
+
*.log
|
| 60 |
+
local_settings.py
|
| 61 |
+
db.sqlite3
|
| 62 |
+
db.sqlite3-journal
|
| 63 |
+
|
| 64 |
+
# Flask stuff:
|
| 65 |
+
instance/
|
| 66 |
+
.webassets-cache
|
| 67 |
+
|
| 68 |
+
# Scrapy stuff:
|
| 69 |
+
.scrapy
|
| 70 |
+
|
| 71 |
+
# Sphinx documentation
|
| 72 |
+
docs/_build/
|
| 73 |
+
|
| 74 |
+
# PyBuilder
|
| 75 |
+
.pybuilder/
|
| 76 |
+
target/
|
| 77 |
+
|
| 78 |
+
# Jupyter Notebook
|
| 79 |
+
.ipynb_checkpoints
|
| 80 |
+
|
| 81 |
+
# IPython
|
| 82 |
+
profile_default/
|
| 83 |
+
ipython_config.py
|
| 84 |
+
|
| 85 |
+
# pyenv
|
| 86 |
+
# For a library or package, you might want to ignore these files since the code is
|
| 87 |
+
# intended to run in multiple environments; otherwise, check them in:
|
| 88 |
+
# .python-version
|
| 89 |
+
|
| 90 |
+
# pipenv
|
| 91 |
+
# According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
|
| 92 |
+
# However, in case of collaboration, if having platform-specific dependencies or dependencies
|
| 93 |
+
# having no cross-platform support, pipenv may install dependencies that don't work, or not
|
| 94 |
+
# install all needed dependencies.
|
| 95 |
+
#Pipfile.lock
|
| 96 |
+
|
| 97 |
+
# poetry
|
| 98 |
+
# Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
|
| 99 |
+
# This is especially recommended for binary packages to ensure reproducibility, and is more
|
| 100 |
+
# commonly ignored for libraries.
|
| 101 |
+
# https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
|
| 102 |
+
#poetry.lock
|
| 103 |
+
|
| 104 |
+
# pdm
|
| 105 |
+
# Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
|
| 106 |
+
#pdm.lock
|
| 107 |
+
# pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
|
| 108 |
+
# in version control.
|
| 109 |
+
# https://pdm.fming.dev/#use-with-ide
|
| 110 |
+
.pdm.toml
|
| 111 |
+
|
| 112 |
+
# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
|
| 113 |
+
__pypackages__/
|
| 114 |
+
|
| 115 |
+
# Celery stuff
|
| 116 |
+
celerybeat-schedule
|
| 117 |
+
celerybeat.pid
|
| 118 |
+
|
| 119 |
+
# SageMath parsed files
|
| 120 |
+
*.sage.py
|
| 121 |
+
|
| 122 |
+
# Environments
|
| 123 |
+
.env
|
| 124 |
+
.venv
|
| 125 |
+
env/
|
| 126 |
+
venv/
|
| 127 |
+
ENV/
|
| 128 |
+
env.bak/
|
| 129 |
+
venv.bak/
|
| 130 |
+
|
| 131 |
+
# Spyder project settings
|
| 132 |
+
.spyderproject
|
| 133 |
+
.spyproject
|
| 134 |
+
|
| 135 |
+
# Rope project settings
|
| 136 |
+
.ropeproject
|
| 137 |
+
|
| 138 |
+
# mkdocs documentation
|
| 139 |
+
/site
|
| 140 |
+
|
| 141 |
+
# mypy
|
| 142 |
+
.mypy_cache/
|
| 143 |
+
.dmypy.json
|
| 144 |
+
dmypy.json
|
| 145 |
+
|
| 146 |
+
# Pyre type checker
|
| 147 |
+
.pyre/
|
| 148 |
+
|
| 149 |
+
# pytype static type analyzer
|
| 150 |
+
.pytype/
|
| 151 |
+
|
| 152 |
+
# Cython debug symbols
|
| 153 |
+
cython_debug/
|
| 154 |
+
|
| 155 |
+
# PyCharm
|
| 156 |
+
# JetBrains specific template is maintained in a separate JetBrains.gitignore that can
|
| 157 |
+
# be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
|
| 158 |
+
# and can be added to the global gitignore or merged into this file. For a more nuclear
|
| 159 |
+
# option (not recommended) you can uncomment the following to ignore the entire idea folder.
|
| 160 |
+
#.idea/
|
app.py
ADDED
|
@@ -0,0 +1,43 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import streamlit as st
|
| 2 |
+
import pandas as pd
|
| 3 |
+
import rdkit
|
| 4 |
+
import streamlit_ketcher
|
| 5 |
+
from streamlit_ketcher import st_ketcher
|
| 6 |
+
import run
|
| 7 |
+
|
| 8 |
+
# Page setup
|
| 9 |
+
st.set_page_config(page_title="DeepDAP", page_icon="🔋", layout="wide")
|
| 10 |
+
st.title("🔋DeepDAP")
|
| 11 |
+
|
| 12 |
+
# Connect to the Google Sheet
|
| 13 |
+
|
| 14 |
+
url1= r"https://docs.google.com/spreadsheets/d/1AKkZS04VF3osFT36aNHIb4iUbV8D1uNfsldcpHXogj0/gviz/tq?tqx=out:csv&sheet=dap"
|
| 15 |
+
df1 = pd.read_csv(url1, dtype=str, encoding='utf-8')
|
| 16 |
+
|
| 17 |
+
text_search = st.text_input("🔍Search papers or molecules", value="")
|
| 18 |
+
m1 = df1["Donor_Name"].str.contains(text_search)
|
| 19 |
+
m2 = df1["reference"].str.contains(text_search)
|
| 20 |
+
m3 = df1["Acceptor_Name"].str.contains(text_search)
|
| 21 |
+
df_search = df1[m1 | m2|m3]
|
| 22 |
+
if text_search:
|
| 23 |
+
st.write(df_search)
|
| 24 |
+
st.download_button( "⬇️Download edited files as .csv", df_search.to_csv(), "df_search.csv", use_container_width=True)
|
| 25 |
+
edited_df = st.data_editor(df1, num_rows="dynamic")
|
| 26 |
+
|
| 27 |
+
st.download_button(
|
| 28 |
+
"⬇️ Download edited files as .csv", edited_df.to_csv(), "edited_df.csv", use_container_width=True
|
| 29 |
+
)
|
| 30 |
+
|
| 31 |
+
molecule = st.text_input("👨🔬Molecule")
|
| 32 |
+
smile_code = st_ketcher(molecule)
|
| 33 |
+
st.markdown("🏆New SMILES of edited molecules: {smile_code }")
|
| 34 |
+
|
| 35 |
+
acceptor= st.text_input("🎈SMILES of acceptor")
|
| 36 |
+
|
| 37 |
+
donor = st.text_input("🎈SMILES of donor")
|
| 38 |
+
|
| 39 |
+
try:
|
| 40 |
+
pce = run.smiles_aas_test( str(acceptor ), str(donor) )
|
| 41 |
+
st.markdown("⚡PCE: ``{pce}``")
|
| 42 |
+
except:
|
| 43 |
+
st.markdown("⚡PCE: None ")
|
requirements.txt
ADDED
|
@@ -0,0 +1,18 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
altair
|
| 2 |
+
streamlit
|
| 3 |
+
streamlit-ketcher
|
| 4 |
+
torch
|
| 5 |
+
tqdm
|
| 6 |
+
transformers
|
| 7 |
+
pytorch_lightning
|
| 8 |
+
scipy
|
| 9 |
+
pandas
|
| 10 |
+
rdkit
|
| 11 |
+
scikit-learn
|
| 12 |
+
matplotlib
|
| 13 |
+
easydict
|
| 14 |
+
wandb
|
| 15 |
+
networkx
|
| 16 |
+
seaborn
|
| 17 |
+
|
| 18 |
+
|
run.py
ADDED
|
@@ -0,0 +1,102 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import os
|
| 2 |
+
import pandas as pd
|
| 3 |
+
|
| 4 |
+
import torch
|
| 5 |
+
from torch.nn import functional as F
|
| 6 |
+
from transformers import AutoTokenizer
|
| 7 |
+
|
| 8 |
+
from util.utils import *
|
| 9 |
+
|
| 10 |
+
from tqdm import tqdm
|
| 11 |
+
from train import markerModel
|
| 12 |
+
os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID"
|
| 13 |
+
os.environ["CUDA_VISIBLE_DEVICES"] = '0,1'
|
| 14 |
+
|
| 15 |
+
device_count = torch.cuda.device_count()
|
| 16 |
+
device_biomarker = torch.device('cuda' if torch.cuda.is_available() else "cpu")
|
| 17 |
+
|
| 18 |
+
device = torch.device('cpu')
|
| 19 |
+
d_model_name = 'DeepChem/ChemBERTa-10M-MTR'
|
| 20 |
+
p_model_name = 'DeepChem/ChemBERTa-10M-MLM'
|
| 21 |
+
|
| 22 |
+
tokenizer = AutoTokenizer.from_pretrained(d_model_name)
|
| 23 |
+
prot_tokenizer = AutoTokenizer.from_pretrained(p_model_name)
|
| 24 |
+
|
| 25 |
+
#--biomarker Model
|
| 26 |
+
##-- hyper param config file Load --##
|
| 27 |
+
config = load_hparams('config/predict.json')
|
| 28 |
+
config = DictX(config)
|
| 29 |
+
model = markerModel.load_from_checkpoint(config.load_checkpoint,strict=False)
|
| 30 |
+
|
| 31 |
+
# model = BiomarkerModel.load_from_checkpoint('./biomarker_bindingdb_train8595_pretopre/3477h3wf/checkpoints/epoch=30-step=7284.ckpt').to(device_biomarker)
|
| 32 |
+
|
| 33 |
+
model.eval()
|
| 34 |
+
model.freeze()
|
| 35 |
+
|
| 36 |
+
if device_biomarker.type == 'cuda':
|
| 37 |
+
model = torch.nn.DataParallel(model)
|
| 38 |
+
|
| 39 |
+
def get_biomarker(drug_inputs, prot_inputs):
|
| 40 |
+
output_preds = model(drug_inputs, prot_inputs)
|
| 41 |
+
|
| 42 |
+
predict = torch.squeeze((output_preds)).tolist()
|
| 43 |
+
|
| 44 |
+
# output_preds = torch.relu(output_preds)
|
| 45 |
+
# predict = torch.tanh(output_preds)
|
| 46 |
+
# predict = predict.squeeze(dim=1).tolist()
|
| 47 |
+
|
| 48 |
+
return predict
|
| 49 |
+
|
| 50 |
+
|
| 51 |
+
def biomarker_prediction(smile_acc, smile_don):
|
| 52 |
+
try:
|
| 53 |
+
aas_input = smile_acc
|
| 54 |
+
|
| 55 |
+
|
| 56 |
+
das_input =smile_don
|
| 57 |
+
d_inputs = tokenizer(aas_input, padding='max_length', max_length=400, truncation=True, return_tensors="pt")
|
| 58 |
+
# d_inputs = tokenizer(smiles, truncation=True, return_tensors="pt")
|
| 59 |
+
drug_input_ids = d_inputs['input_ids'].to(device)
|
| 60 |
+
drug_attention_mask = d_inputs['attention_mask'].to(device)
|
| 61 |
+
drug_inputs = {'input_ids': drug_input_ids, 'attention_mask': drug_attention_mask}
|
| 62 |
+
|
| 63 |
+
p_inputs = prot_tokenizer(das_input, padding='max_length', max_length=400, truncation=True, return_tensors="pt")
|
| 64 |
+
# p_inputs = prot_tokenizer(aas_input, truncation=True, return_tensors="pt")
|
| 65 |
+
prot_input_ids = p_inputs['input_ids'].to(device)
|
| 66 |
+
prot_attention_mask = p_inputs['attention_mask'].to(device)
|
| 67 |
+
prot_inputs = {'input_ids': prot_input_ids, 'attention_mask': prot_attention_mask}
|
| 68 |
+
|
| 69 |
+
output_predict = get_biomarker(drug_inputs, prot_inputs)
|
| 70 |
+
|
| 71 |
+
return output_predict
|
| 72 |
+
|
| 73 |
+
except Exception as e:
|
| 74 |
+
print(e)
|
| 75 |
+
return {'Error_message': e}
|
| 76 |
+
|
| 77 |
+
|
| 78 |
+
def smiles_aas_test(smile_acc,smile_don):
|
| 79 |
+
|
| 80 |
+
batch_size = 1
|
| 81 |
+
try:
|
| 82 |
+
output_pred = biomarker_prediction((smile_acc), (smile_don))
|
| 83 |
+
|
| 84 |
+
datas = output_pred
|
| 85 |
+
|
| 86 |
+
## -- Export result data to csv -- ##
|
| 87 |
+
# df = pd.DataFrame(datas)
|
| 88 |
+
# df.to_csv('./results/predict_test.csv', index=None)
|
| 89 |
+
|
| 90 |
+
# print(df)
|
| 91 |
+
return datas
|
| 92 |
+
|
| 93 |
+
except Exception as e:
|
| 94 |
+
print(e)
|
| 95 |
+
return {'Error_message': e}
|
| 96 |
+
|
| 97 |
+
|
| 98 |
+
if __name__ == "__main__":
|
| 99 |
+
a = smiles_aas_test(smile_acc,smile_don)
|
| 100 |
+
|
| 101 |
+
|
| 102 |
+
|
train.py
ADDED
|
@@ -0,0 +1,454 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import os
|
| 2 |
+
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
|
| 3 |
+
from curses import delay_output
|
| 4 |
+
import gc, os
|
| 5 |
+
import numpy as np
|
| 6 |
+
import pandas as pd
|
| 7 |
+
import wandb
|
| 8 |
+
from scipy.stats import pearsonr
|
| 9 |
+
from util.utils import *
|
| 10 |
+
from util.attention_flow import *
|
| 11 |
+
|
| 12 |
+
import torch
|
| 13 |
+
import torch.nn as nn
|
| 14 |
+
|
| 15 |
+
import sklearn as sk
|
| 16 |
+
from torch.utils.data import Dataset, DataLoader
|
| 17 |
+
|
| 18 |
+
import pytorch_lightning as pl
|
| 19 |
+
from pytorch_lightning.loggers import WandbLogger, TensorBoardLogger
|
| 20 |
+
from pytorch_lightning.callbacks import ModelCheckpoint, EarlyStopping
|
| 21 |
+
from transformers import AutoConfig, AutoTokenizer, RobertaModel, BertModel
|
| 22 |
+
from sklearn.metrics import r2_score, mean_absolute_error,mean_squared_error
|
| 23 |
+
|
| 24 |
+
class markerDataset(Dataset):
|
| 25 |
+
def __init__(self, list_IDs, labels, df_dti, d_tokenizer, p_tokenizer):
|
| 26 |
+
'Initialization'
|
| 27 |
+
self.labels = labels
|
| 28 |
+
self.list_IDs = list_IDs
|
| 29 |
+
self.df = df_dti
|
| 30 |
+
|
| 31 |
+
self.d_tokenizer = d_tokenizer
|
| 32 |
+
self.p_tokenizer = p_tokenizer
|
| 33 |
+
|
| 34 |
+
|
| 35 |
+
|
| 36 |
+
def convert_data(self, acc_data, don_data):
|
| 37 |
+
|
| 38 |
+
|
| 39 |
+
d_inputs = self.d_tokenizer(acc_data, return_tensors="pt")
|
| 40 |
+
p_inputs = self.d_tokenizer(don_data, return_tensors="pt")
|
| 41 |
+
|
| 42 |
+
acc_input_ids = d_inputs['input_ids']
|
| 43 |
+
acc_attention_mask = d_inputs['attention_mask']
|
| 44 |
+
acc_inputs = {'input_ids': acc_input_ids, 'attention_mask': acc_attention_mask}
|
| 45 |
+
|
| 46 |
+
don_input_ids = p_inputs['input_ids']
|
| 47 |
+
don_attention_mask = p_inputs['attention_mask']
|
| 48 |
+
don_inputs = {'input_ids': don_input_ids, 'attention_mask': don_attention_mask}
|
| 49 |
+
|
| 50 |
+
return acc_inputs, don_inputs
|
| 51 |
+
|
| 52 |
+
def tokenize_data(self, acc_data, don_data):
|
| 53 |
+
|
| 54 |
+
tokenize_acc = ['[CLS]'] + self.d_tokenizer.tokenize(acc_data) + ['[SEP]']
|
| 55 |
+
|
| 56 |
+
tokenize_don = ['[CLS]'] + self.p_tokenizer.tokenize(don_data) + ['[SEP]']
|
| 57 |
+
|
| 58 |
+
return tokenize_acc, tokenize_don
|
| 59 |
+
|
| 60 |
+
def __len__(self):
|
| 61 |
+
'Denotes the total number of samples'
|
| 62 |
+
return len(self.list_IDs)
|
| 63 |
+
|
| 64 |
+
def __getitem__(self, index):
|
| 65 |
+
'Generates one sample of data'
|
| 66 |
+
index = self.list_IDs[index]
|
| 67 |
+
acc_data = self.df.iloc[index]['acceptor']
|
| 68 |
+
don_data = self.df.iloc[index]['donor']
|
| 69 |
+
|
| 70 |
+
d_inputs = self.d_tokenizer(acc_data, padding='max_length', max_length=400, truncation=True, return_tensors="pt")
|
| 71 |
+
p_inputs = self.p_tokenizer(don_data, padding='max_length', max_length=400, truncation=True, return_tensors="pt")
|
| 72 |
+
|
| 73 |
+
d_input_ids = d_inputs['input_ids'].squeeze()
|
| 74 |
+
d_attention_mask = d_inputs['attention_mask'].squeeze()
|
| 75 |
+
p_input_ids = p_inputs['input_ids'].squeeze()
|
| 76 |
+
p_attention_mask = p_inputs['attention_mask'].squeeze()
|
| 77 |
+
|
| 78 |
+
labels = torch.as_tensor(self.labels[index], dtype=torch.float)
|
| 79 |
+
|
| 80 |
+
dataset = [d_input_ids, d_attention_mask, p_input_ids, p_attention_mask, labels]
|
| 81 |
+
return dataset
|
| 82 |
+
|
| 83 |
+
|
| 84 |
+
class markerDataModule(pl.LightningDataModule):
|
| 85 |
+
def __init__(self, task_name, acc_model_name, don_model_name, num_workers, batch_size, traindata_rate = 1.0):
|
| 86 |
+
super().__init__()
|
| 87 |
+
self.batch_size = batch_size
|
| 88 |
+
self.num_workers = num_workers
|
| 89 |
+
self.task_name = task_name
|
| 90 |
+
|
| 91 |
+
self.traindata_rate = traindata_rate
|
| 92 |
+
|
| 93 |
+
self.d_tokenizer = AutoTokenizer.from_pretrained(acc_model_name)
|
| 94 |
+
self.p_tokenizer = AutoTokenizer.from_pretrained(don_model_name)
|
| 95 |
+
|
| 96 |
+
self.df_train = None
|
| 97 |
+
self.df_val = None
|
| 98 |
+
self.df_test = None
|
| 99 |
+
|
| 100 |
+
self.load_testData = True
|
| 101 |
+
|
| 102 |
+
self.train_dataset = None
|
| 103 |
+
self.valid_dataset = None
|
| 104 |
+
self.test_dataset = None
|
| 105 |
+
|
| 106 |
+
def get_task(self, task_name):
|
| 107 |
+
if task_name.lower() == 'OSC':
|
| 108 |
+
return './dataset/OSC/'
|
| 109 |
+
|
| 110 |
+
elif task_name.lower() == 'merge':
|
| 111 |
+
self.load_testData = False
|
| 112 |
+
return './dataset/MergeDataset'
|
| 113 |
+
|
| 114 |
+
def prepare_data(self):
|
| 115 |
+
# Use this method to do things that might write to disk or that need to be done only from
|
| 116 |
+
# a single process in distributed settings.
|
| 117 |
+
dataFolder = './dataset/OSC'
|
| 118 |
+
|
| 119 |
+
self.df_train = pd.read_csv(dataFolder + '/train.csv')
|
| 120 |
+
self.df_val = pd.read_csv(dataFolder + '/val.csv')
|
| 121 |
+
|
| 122 |
+
## -- Data Lenght Rate apply -- ##
|
| 123 |
+
traindata_length = int(len(self.df_train) * self.traindata_rate)
|
| 124 |
+
validdata_length = int(len(self.df_val) * self.traindata_rate)
|
| 125 |
+
|
| 126 |
+
self.df_train = self.df_train[:traindata_length]
|
| 127 |
+
self.df_val = self.df_val[:validdata_length]
|
| 128 |
+
|
| 129 |
+
if self.load_testData is True:
|
| 130 |
+
self.df_test = pd.read_csv(dataFolder + '/test.csv')
|
| 131 |
+
|
| 132 |
+
def setup(self, stage=None):
|
| 133 |
+
if stage == 'fit' or stage is None:
|
| 134 |
+
self.train_dataset = markerDataset(self.df_train.index.values, self.df_train.Label.values, self.df_train,
|
| 135 |
+
self.d_tokenizer, self.p_tokenizer)
|
| 136 |
+
self.valid_dataset = markerDataset(self.df_val.index.values, self.df_val.Label.values, self.df_val,
|
| 137 |
+
self.d_tokenizer, self.p_tokenizer)
|
| 138 |
+
|
| 139 |
+
if self.load_testData is True:
|
| 140 |
+
self.test_dataset = markerDataset(self.df_test.index.values, self.df_test.Label.values, self.df_test,
|
| 141 |
+
self.d_tokenizer, self.p_tokenizer)
|
| 142 |
+
|
| 143 |
+
def train_dataloader(self):
|
| 144 |
+
return DataLoader(self.train_dataset, batch_size=self.batch_size, shuffle=True, num_workers=self.num_workers)
|
| 145 |
+
|
| 146 |
+
def val_dataloader(self):
|
| 147 |
+
return DataLoader(self.valid_dataset, batch_size=self.batch_size, num_workers=self.num_workers)
|
| 148 |
+
|
| 149 |
+
def test_dataloader(self):
|
| 150 |
+
return DataLoader(self.test_dataset, batch_size=self.batch_size, num_workers=self.num_workers)
|
| 151 |
+
|
| 152 |
+
|
| 153 |
+
class markerModel(pl.LightningModule):
|
| 154 |
+
def __init__(self, acc_model_name, don_model_name, lr, dropout, layer_features, loss_fn = "smooth", layer_limit = True, d_pretrained=True, p_pretrained=True):
|
| 155 |
+
super().__init__()
|
| 156 |
+
self.lr = lr
|
| 157 |
+
self.loss_fn = loss_fn
|
| 158 |
+
self.criterion = torch.nn.MSELoss()
|
| 159 |
+
self.criterion_smooth = torch.nn.SmoothL1Loss()
|
| 160 |
+
# self.sigmoid = nn.Sigmoid()
|
| 161 |
+
|
| 162 |
+
#-- Pretrained Model Setting
|
| 163 |
+
acc_config = AutoConfig.from_pretrained("seyonec/SMILES_BPE_PubChem_100k_shard00")
|
| 164 |
+
if d_pretrained is False:
|
| 165 |
+
self.d_model = RobertaModel(acc_config)
|
| 166 |
+
print('acceptor model without pretraining')
|
| 167 |
+
else:
|
| 168 |
+
self.d_model = RobertaModel.from_pretrained(acc_model_name, num_labels=2,
|
| 169 |
+
output_hidden_states=True,
|
| 170 |
+
output_attentions=True)
|
| 171 |
+
|
| 172 |
+
don_config = AutoConfig.from_pretrained("seyonec/SMILES_BPE_PubChem_100k_shard00")
|
| 173 |
+
|
| 174 |
+
if p_pretrained is False:
|
| 175 |
+
self.p_model = RobertaModel(don_config)
|
| 176 |
+
print('donor model without pretraining')
|
| 177 |
+
else:
|
| 178 |
+
self.p_model = RobertaModel.from_pretrained(don_model_name,
|
| 179 |
+
output_hidden_states=True,
|
| 180 |
+
output_attentions=True)
|
| 181 |
+
|
| 182 |
+
#-- Decoder Layer Setting
|
| 183 |
+
layers = []
|
| 184 |
+
firstfeature = self.d_model.config.hidden_size + self.p_model.config.hidden_size
|
| 185 |
+
for feature_idx in range(0, len(layer_features) - 1):
|
| 186 |
+
layers.append(nn.Linear(firstfeature, layer_features[feature_idx]))
|
| 187 |
+
firstfeature = layer_features[feature_idx]
|
| 188 |
+
|
| 189 |
+
if feature_idx is len(layer_features)-2:
|
| 190 |
+
layers.append(nn.ReLU())
|
| 191 |
+
else:
|
| 192 |
+
layers.append(nn.ReLU())
|
| 193 |
+
|
| 194 |
+
if dropout > 0:
|
| 195 |
+
layers.append(nn.Dropout(dropout))
|
| 196 |
+
|
| 197 |
+
layers.append(nn.Linear(firstfeature, layer_features[-1]))
|
| 198 |
+
|
| 199 |
+
self.decoder = nn.Sequential(*layers)
|
| 200 |
+
|
| 201 |
+
self.save_hyperparameters()
|
| 202 |
+
|
| 203 |
+
def forward(self, acc_inputs, don_inputs):
|
| 204 |
+
|
| 205 |
+
d_outputs = self.d_model(acc_inputs['input_ids'], acc_inputs['attention_mask'])
|
| 206 |
+
p_outputs = self.p_model(don_inputs['input_ids'], don_inputs['attention_mask'])
|
| 207 |
+
|
| 208 |
+
outs = torch.cat((d_outputs.last_hidden_state[:, 0], p_outputs.last_hidden_state[:, 0]), dim=1)
|
| 209 |
+
outs = self.decoder(outs)
|
| 210 |
+
|
| 211 |
+
return outs
|
| 212 |
+
|
| 213 |
+
def attention_output(self, acc_inputs, don_inputs):
|
| 214 |
+
|
| 215 |
+
d_outputs = self.d_model(acc_inputs['input_ids'], acc_inputs['attention_mask'])
|
| 216 |
+
p_outputs = self.p_model(don_inputs['input_ids'], don_inputs['attention_mask'])
|
| 217 |
+
|
| 218 |
+
outs = torch.cat((d_outputs.last_hidden_state[:, 0], p_outputs.last_hidden_state[:, 0]), dim=1)
|
| 219 |
+
outs = self.decoder(outs)
|
| 220 |
+
|
| 221 |
+
return d_outputs['attentions'], p_outputs['attentions'], outs
|
| 222 |
+
|
| 223 |
+
def training_step(self, batch, batch_idx):
|
| 224 |
+
|
| 225 |
+
acc_inputs = {'input_ids': batch[0], 'attention_mask': batch[1]}
|
| 226 |
+
|
| 227 |
+
don_inputs = {'input_ids': batch[2], 'attention_mask': batch[3]}
|
| 228 |
+
|
| 229 |
+
labels = batch[4]
|
| 230 |
+
|
| 231 |
+
output = self(acc_inputs, don_inputs)
|
| 232 |
+
logits = output.squeeze(dim=1)
|
| 233 |
+
|
| 234 |
+
if self.loss_fn == 'MSE':
|
| 235 |
+
loss = self.criterion(logits, labels)
|
| 236 |
+
else:
|
| 237 |
+
loss = self.criterion_smooth(logits, labels)
|
| 238 |
+
|
| 239 |
+
self.log("train_loss", loss, on_step=False, on_epoch=True, logger=True)
|
| 240 |
+
# print("train_loss", loss)
|
| 241 |
+
return {"loss": loss}
|
| 242 |
+
|
| 243 |
+
def validation_step(self, batch, batch_idx):
|
| 244 |
+
acc_inputs = {'input_ids': batch[0], 'attention_mask': batch[1]}
|
| 245 |
+
don_inputs = {'input_ids': batch[2], 'attention_mask': batch[3]}
|
| 246 |
+
labels = batch[4]
|
| 247 |
+
|
| 248 |
+
output = self(acc_inputs, don_inputs)
|
| 249 |
+
logits = output.squeeze(dim=1)
|
| 250 |
+
|
| 251 |
+
|
| 252 |
+
if self.loss_fn == 'MSE':
|
| 253 |
+
loss = self.criterion(logits, labels)
|
| 254 |
+
else:
|
| 255 |
+
loss = self.criterion_smooth(logits, labels)
|
| 256 |
+
|
| 257 |
+
self.log("valid_loss", loss, on_step=False, on_epoch=True, logger=True)
|
| 258 |
+
# print("valid_loss", loss)
|
| 259 |
+
return {"logits": logits, "labels": labels}
|
| 260 |
+
|
| 261 |
+
def validation_step_end(self, outputs):
|
| 262 |
+
return {"logits": outputs['logits'], "labels": outputs['labels']}
|
| 263 |
+
|
| 264 |
+
def validation_epoch_end(self, outputs):
|
| 265 |
+
preds = self.convert_outputs_to_preds(outputs)
|
| 266 |
+
labels = torch.as_tensor(torch.cat([output['labels'] for output in outputs], dim=0), dtype=torch.int)
|
| 267 |
+
|
| 268 |
+
mae, mse, r2,r = self.log_score(preds, labels)
|
| 269 |
+
|
| 270 |
+
self.log("mae", mae, on_step=False, on_epoch=True, logger=True)
|
| 271 |
+
self.log("mse", mse, on_step=False, on_epoch=True, logger=True)
|
| 272 |
+
|
| 273 |
+
self.log("r2", r2, on_step=False, on_epoch=True, logger=True)
|
| 274 |
+
|
| 275 |
+
def test_step(self, batch, batch_idx):
|
| 276 |
+
acc_inputs = {'input_ids': batch[0], 'attention_mask': batch[1]}
|
| 277 |
+
don_inputs = {'input_ids': batch[2], 'attention_mask': batch[3]}
|
| 278 |
+
labels = batch[4]
|
| 279 |
+
|
| 280 |
+
output = self(acc_inputs, don_inputs)
|
| 281 |
+
logits = output.squeeze(dim=1)
|
| 282 |
+
|
| 283 |
+
if self.loss_fn == 'MSE':
|
| 284 |
+
loss = self.criterion(logits, labels)
|
| 285 |
+
else:
|
| 286 |
+
loss = self.criterion_smooth(logits, labels)
|
| 287 |
+
|
| 288 |
+
self.log("test_loss", loss, on_step=False, on_epoch=True, logger=True)
|
| 289 |
+
return {"logits": logits, "labels": labels}
|
| 290 |
+
|
| 291 |
+
def test_step_end(self, outputs):
|
| 292 |
+
return {"logits": outputs['logits'], "labels": outputs['labels']}
|
| 293 |
+
|
| 294 |
+
def test_epoch_end(self, outputs):
|
| 295 |
+
preds = self.convert_outputs_to_preds(outputs)
|
| 296 |
+
labels = torch.as_tensor(torch.cat([output['labels'] for output in outputs], dim=0), dtype=torch.int)
|
| 297 |
+
|
| 298 |
+
mae, mse, r2,r = self.log_score(preds, labels)
|
| 299 |
+
|
| 300 |
+
self.log("mae", mae, on_step=False, on_epoch=True, logger=True)
|
| 301 |
+
self.log("mse", mse, on_step=False, on_epoch=True, logger=True)
|
| 302 |
+
self.log("r2", r2, on_step=False, on_epoch=True, logger=True)
|
| 303 |
+
self.log("r", r, on_step=False, on_epoch=True, logger=True)
|
| 304 |
+
def configure_optimizers(self):
|
| 305 |
+
|
| 306 |
+
param_optimizer = list(self.named_parameters())
|
| 307 |
+
|
| 308 |
+
no_decay = ["bias", "gamma", "beta"]
|
| 309 |
+
optimizer_grouped_parameters = [
|
| 310 |
+
{
|
| 311 |
+
"params": [p for n, p in param_optimizer if not any(nd in n for nd in no_decay)],
|
| 312 |
+
"weight_decay_rate": 0.0001
|
| 313 |
+
},
|
| 314 |
+
{
|
| 315 |
+
"params": [p for n, p in param_optimizer if any(nd in n for nd in no_decay)],
|
| 316 |
+
"weight_decay_rate": 0.0
|
| 317 |
+
},
|
| 318 |
+
]
|
| 319 |
+
optimizer = torch.optim.AdamW(
|
| 320 |
+
optimizer_grouped_parameters,
|
| 321 |
+
lr=self.lr,
|
| 322 |
+
)
|
| 323 |
+
return optimizer
|
| 324 |
+
|
| 325 |
+
def convert_outputs_to_preds(self, outputs):
|
| 326 |
+
logits = torch.cat([output['logits'] for output in outputs], dim=0)
|
| 327 |
+
return logits
|
| 328 |
+
|
| 329 |
+
def log_score(self, preds, labels):
|
| 330 |
+
y_pred = preds.detach().cpu().numpy()
|
| 331 |
+
y_label = labels.detach().cpu().numpy()
|
| 332 |
+
|
| 333 |
+
mae = mean_absolute_error(y_label, y_pred)
|
| 334 |
+
mse = mean_squared_error(y_label, y_pred)
|
| 335 |
+
r2=r2_score(y_label, y_pred)
|
| 336 |
+
r = pearsonr(y_label, y_pred)
|
| 337 |
+
print(f'\nmae : {mae}')
|
| 338 |
+
print(f'mse : {mse}')
|
| 339 |
+
print(f'r2 : {r2}')
|
| 340 |
+
print(f'r : {r}')
|
| 341 |
+
|
| 342 |
+
return mae, mse, r2, r
|
| 343 |
+
|
| 344 |
+
|
| 345 |
+
def main_wandb(config=None):
|
| 346 |
+
try:
|
| 347 |
+
if config is not None:
|
| 348 |
+
wandb.init(config=config, project=project_name)
|
| 349 |
+
else:
|
| 350 |
+
wandb.init(settings=wandb.Settings(console='off'))
|
| 351 |
+
|
| 352 |
+
config = wandb.config
|
| 353 |
+
pl.seed_everything(seed=config.num_seed)
|
| 354 |
+
|
| 355 |
+
dm = markerDataModule(config.task_name, config.d_model_name, config.p_model_name,
|
| 356 |
+
config.num_workers, config.batch_size, config.prot_maxlength, config.traindata_rate)
|
| 357 |
+
dm.prepare_data()
|
| 358 |
+
dm.setup()
|
| 359 |
+
|
| 360 |
+
model_type = str(config.pretrained['chem'])+"To"+str(config.pretrained['prot'])
|
| 361 |
+
#model_logger = WandbLogger(project=project_name)
|
| 362 |
+
checkpoint_callback = ModelCheckpoint(f"{config.task_name}_{model_type}_{config.lr}_{config.num_seed}", save_top_k=1, monitor="mae", mode="max")
|
| 363 |
+
|
| 364 |
+
trainer = pl.Trainer(
|
| 365 |
+
max_epochs=config.max_epoch,
|
| 366 |
+
precision=16,
|
| 367 |
+
#logger=model_logger,
|
| 368 |
+
callbacks=[checkpoint_callback],
|
| 369 |
+
accelerator='cpu',log_every_n_steps=40
|
| 370 |
+
)
|
| 371 |
+
|
| 372 |
+
|
| 373 |
+
if config.model_mode == "train":
|
| 374 |
+
model = markerModel(config.d_model_name, config.p_model_name,
|
| 375 |
+
config.lr, config.dropout, config.layer_features, config.loss_fn, config.layer_limit, config.pretrained['chem'], config.pretrained['prot'])
|
| 376 |
+
model.train()
|
| 377 |
+
trainer.fit(model, datamodule=dm)
|
| 378 |
+
|
| 379 |
+
model.eval()
|
| 380 |
+
trainer.test(model, datamodule=dm)
|
| 381 |
+
|
| 382 |
+
else:
|
| 383 |
+
model = markerModel.load_from_checkpoint(config.load_checkpoint)
|
| 384 |
+
|
| 385 |
+
model.eval()
|
| 386 |
+
trainer.test(model, datamodule=dm)
|
| 387 |
+
|
| 388 |
+
except Exception as e:
|
| 389 |
+
print(e)
|
| 390 |
+
|
| 391 |
+
|
| 392 |
+
def main_default(config):
|
| 393 |
+
try:
|
| 394 |
+
config = DictX(config)
|
| 395 |
+
pl.seed_everything(seed=config.num_seed)
|
| 396 |
+
|
| 397 |
+
dm = markerDataModule(config.task_name, config.d_model_name, config.p_model_name,
|
| 398 |
+
config.num_workers, config.batch_size, config.traindata_rate)
|
| 399 |
+
|
| 400 |
+
dm.prepare_data()
|
| 401 |
+
dm.setup()
|
| 402 |
+
model_type = str(config.pretrained['chem'])+"To"+str(config.pretrained['prot'])
|
| 403 |
+
# model_logger = TensorBoardLogger("./log", name=f"{config.task_name}_{model_type}_{config.num_seed}")
|
| 404 |
+
checkpoint_callback = ModelCheckpoint(f"{config.task_name}_{model_type}_{config.lr}_{config.num_seed}", save_top_k=1, monitor="mse", mode="max")
|
| 405 |
+
|
| 406 |
+
trainer = pl.Trainer(
|
| 407 |
+
max_epochs=config.max_epoch,
|
| 408 |
+
precision= 32,
|
| 409 |
+
# logger=model_logger,
|
| 410 |
+
callbacks=[checkpoint_callback],
|
| 411 |
+
accelerator='cpu',log_every_n_steps=40
|
| 412 |
+
)
|
| 413 |
+
|
| 414 |
+
|
| 415 |
+
if config.model_mode == "train":
|
| 416 |
+
model = markerModel(config.d_model_name, config.p_model_name,
|
| 417 |
+
config.lr, config.dropout, config.layer_features, config.loss_fn, config.layer_limit, config.pretrained['chem'], config.pretrained['prot'])
|
| 418 |
+
|
| 419 |
+
model.train()
|
| 420 |
+
|
| 421 |
+
trainer.fit(model, datamodule=dm)
|
| 422 |
+
|
| 423 |
+
model.eval()
|
| 424 |
+
trainer.test(model, datamodule=dm)
|
| 425 |
+
|
| 426 |
+
else:
|
| 427 |
+
model = markerModel.load_from_checkpoint(config.load_checkpoint)
|
| 428 |
+
|
| 429 |
+
model.eval()
|
| 430 |
+
trainer.test(model, datamodule=dm)
|
| 431 |
+
except Exception as e:
|
| 432 |
+
print(e)
|
| 433 |
+
|
| 434 |
+
|
| 435 |
+
if __name__ == '__main__':
|
| 436 |
+
using_wandb = False
|
| 437 |
+
|
| 438 |
+
if using_wandb == True:
|
| 439 |
+
#-- hyper param config file Load --##
|
| 440 |
+
config = load_hparams('config/config_hparam.json')
|
| 441 |
+
project_name = config["name"]
|
| 442 |
+
|
| 443 |
+
main_wandb(config)
|
| 444 |
+
|
| 445 |
+
##-- wandb Sweep Hyper Param Tuning --##
|
| 446 |
+
# config = load_hparams('config/config_sweep_bindingDB.json')
|
| 447 |
+
# project_name = config["name"]
|
| 448 |
+
# sweep_id = wandb.sweep(config, project=project_name)
|
| 449 |
+
# wandb.agent(sweep_id, main_wandb)
|
| 450 |
+
|
| 451 |
+
else:
|
| 452 |
+
config = load_hparams('config/config_hparam.json')
|
| 453 |
+
|
| 454 |
+
main_default(config)
|