Commit
·
21fa57e
1
Parent(s):
b24b48a
update
Browse files- classification/tokenizer_bbb.py +28 -23
- regression/tokenizer_bbb.py +28 -23
- tokenizer_bbb.py +28 -23
classification/tokenizer_bbb.py
CHANGED
|
@@ -39,6 +39,10 @@ class BBBTokenizer(PreTrainedTokenizer):
|
|
| 39 |
)
|
| 40 |
])
|
| 41 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 42 |
def generate_tab_features(self, smiles):
|
| 43 |
mol = Chem.MolFromSmiles(smiles)
|
| 44 |
|
|
@@ -80,29 +84,30 @@ class BBBTokenizer(PreTrainedTokenizer):
|
|
| 80 |
return_tensors: str = "pt",
|
| 81 |
**kwargs
|
| 82 |
):
|
| 83 |
-
if
|
| 84 |
-
|
| 85 |
-
|
| 86 |
-
|
| 87 |
-
|
| 88 |
-
|
| 89 |
-
|
| 90 |
-
|
| 91 |
-
|
| 92 |
-
|
| 93 |
-
|
| 94 |
-
|
| 95 |
-
|
| 96 |
-
|
| 97 |
-
|
| 98 |
-
|
| 99 |
-
|
| 100 |
-
|
| 101 |
-
|
| 102 |
-
|
| 103 |
-
|
| 104 |
-
|
| 105 |
-
|
|
|
|
| 106 |
|
| 107 |
data_list = []
|
| 108 |
tab, img, txt = [], [], []
|
|
|
|
| 39 |
)
|
| 40 |
])
|
| 41 |
|
| 42 |
+
self.feature_transformer_tab = None
|
| 43 |
+
self.feature_transformer_img = None
|
| 44 |
+
self.feature_transformer_txt = None
|
| 45 |
+
|
| 46 |
def generate_tab_features(self, smiles):
|
| 47 |
mol = Chem.MolFromSmiles(smiles)
|
| 48 |
|
|
|
|
| 84 |
return_tensors: str = "pt",
|
| 85 |
**kwargs
|
| 86 |
):
|
| 87 |
+
if self.transformer_tab_path is None and self.transformer_img_path is None and self.transformer_txt_path is None:
|
| 88 |
+
if task == 'classification':
|
| 89 |
+
model_dir = snapshot_download("SaeedLab/TITAN-BBB", allow_patterns=["normalize_cls_tabular.joblib"])
|
| 90 |
+
transformer_tab_path = os.path.join(model_dir, "normalize_cls_tabular.joblib")
|
| 91 |
+
model_dir = snapshot_download("SaeedLab/TITAN-BBB", allow_patterns=["normalize_cls_image.joblib"])
|
| 92 |
+
transformer_img_path = os.path.join(model_dir, "normalize_cls_image.joblib")
|
| 93 |
+
model_dir = snapshot_download("SaeedLab/TITAN-BBB", allow_patterns=["normalize_cls_text.joblib"])
|
| 94 |
+
transformer_txt_path = os.path.join(model_dir, "normalize_cls_text.joblib")
|
| 95 |
+
|
| 96 |
+
elif task == 'regression':
|
| 97 |
+
model_dir = snapshot_download("SaeedLab/TITAN-BBB", allow_patterns=["normalize_reg_tabular.joblib"])
|
| 98 |
+
transformer_tab_path = os.path.join(model_dir, "normalize_reg_tabular.joblib")
|
| 99 |
+
model_dir = snapshot_download("SaeedLab/TITAN-BBB", allow_patterns=["normalize_reg_image.joblib"])
|
| 100 |
+
transformer_img_path = os.path.join(model_dir, "normalize_reg_image.joblib")
|
| 101 |
+
model_dir = snapshot_download("SaeedLab/TITAN-BBB", allow_patterns=["normalize_reg_text.joblib"])
|
| 102 |
+
transformer_txt_path = os.path.join(model_dir, "normalize_reg_text.joblib")
|
| 103 |
+
|
| 104 |
+
else:
|
| 105 |
+
raise ValueError('task not defined')
|
| 106 |
+
return
|
| 107 |
+
|
| 108 |
+
self.feature_transformer_tab = joblib.load(transformer_tab_path)
|
| 109 |
+
self.feature_transformer_img = joblib.load(transformer_img_path)
|
| 110 |
+
self.feature_transformer_txt = joblib.load(transformer_txt_path)
|
| 111 |
|
| 112 |
data_list = []
|
| 113 |
tab, img, txt = [], [], []
|
regression/tokenizer_bbb.py
CHANGED
|
@@ -39,6 +39,10 @@ class BBBTokenizer(PreTrainedTokenizer):
|
|
| 39 |
)
|
| 40 |
])
|
| 41 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 42 |
def generate_tab_features(self, smiles):
|
| 43 |
mol = Chem.MolFromSmiles(smiles)
|
| 44 |
|
|
@@ -80,29 +84,30 @@ class BBBTokenizer(PreTrainedTokenizer):
|
|
| 80 |
return_tensors: str = "pt",
|
| 81 |
**kwargs
|
| 82 |
):
|
| 83 |
-
if
|
| 84 |
-
|
| 85 |
-
|
| 86 |
-
|
| 87 |
-
|
| 88 |
-
|
| 89 |
-
|
| 90 |
-
|
| 91 |
-
|
| 92 |
-
|
| 93 |
-
|
| 94 |
-
|
| 95 |
-
|
| 96 |
-
|
| 97 |
-
|
| 98 |
-
|
| 99 |
-
|
| 100 |
-
|
| 101 |
-
|
| 102 |
-
|
| 103 |
-
|
| 104 |
-
|
| 105 |
-
|
|
|
|
| 106 |
|
| 107 |
data_list = []
|
| 108 |
tab, img, txt = [], [], []
|
|
|
|
| 39 |
)
|
| 40 |
])
|
| 41 |
|
| 42 |
+
self.feature_transformer_tab = None
|
| 43 |
+
self.feature_transformer_img = None
|
| 44 |
+
self.feature_transformer_txt = None
|
| 45 |
+
|
| 46 |
def generate_tab_features(self, smiles):
|
| 47 |
mol = Chem.MolFromSmiles(smiles)
|
| 48 |
|
|
|
|
| 84 |
return_tensors: str = "pt",
|
| 85 |
**kwargs
|
| 86 |
):
|
| 87 |
+
if self.transformer_tab_path is None and self.transformer_img_path is None and self.transformer_txt_path is None:
|
| 88 |
+
if task == 'classification':
|
| 89 |
+
model_dir = snapshot_download("SaeedLab/TITAN-BBB", allow_patterns=["normalize_cls_tabular.joblib"])
|
| 90 |
+
transformer_tab_path = os.path.join(model_dir, "normalize_cls_tabular.joblib")
|
| 91 |
+
model_dir = snapshot_download("SaeedLab/TITAN-BBB", allow_patterns=["normalize_cls_image.joblib"])
|
| 92 |
+
transformer_img_path = os.path.join(model_dir, "normalize_cls_image.joblib")
|
| 93 |
+
model_dir = snapshot_download("SaeedLab/TITAN-BBB", allow_patterns=["normalize_cls_text.joblib"])
|
| 94 |
+
transformer_txt_path = os.path.join(model_dir, "normalize_cls_text.joblib")
|
| 95 |
+
|
| 96 |
+
elif task == 'regression':
|
| 97 |
+
model_dir = snapshot_download("SaeedLab/TITAN-BBB", allow_patterns=["normalize_reg_tabular.joblib"])
|
| 98 |
+
transformer_tab_path = os.path.join(model_dir, "normalize_reg_tabular.joblib")
|
| 99 |
+
model_dir = snapshot_download("SaeedLab/TITAN-BBB", allow_patterns=["normalize_reg_image.joblib"])
|
| 100 |
+
transformer_img_path = os.path.join(model_dir, "normalize_reg_image.joblib")
|
| 101 |
+
model_dir = snapshot_download("SaeedLab/TITAN-BBB", allow_patterns=["normalize_reg_text.joblib"])
|
| 102 |
+
transformer_txt_path = os.path.join(model_dir, "normalize_reg_text.joblib")
|
| 103 |
+
|
| 104 |
+
else:
|
| 105 |
+
raise ValueError('task not defined')
|
| 106 |
+
return
|
| 107 |
+
|
| 108 |
+
self.feature_transformer_tab = joblib.load(transformer_tab_path)
|
| 109 |
+
self.feature_transformer_img = joblib.load(transformer_img_path)
|
| 110 |
+
self.feature_transformer_txt = joblib.load(transformer_txt_path)
|
| 111 |
|
| 112 |
data_list = []
|
| 113 |
tab, img, txt = [], [], []
|
tokenizer_bbb.py
CHANGED
|
@@ -39,6 +39,10 @@ class BBBTokenizer(PreTrainedTokenizer):
|
|
| 39 |
)
|
| 40 |
])
|
| 41 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 42 |
def generate_tab_features(self, smiles):
|
| 43 |
mol = Chem.MolFromSmiles(smiles)
|
| 44 |
|
|
@@ -80,29 +84,30 @@ class BBBTokenizer(PreTrainedTokenizer):
|
|
| 80 |
return_tensors: str = "pt",
|
| 81 |
**kwargs
|
| 82 |
):
|
| 83 |
-
if
|
| 84 |
-
|
| 85 |
-
|
| 86 |
-
|
| 87 |
-
|
| 88 |
-
|
| 89 |
-
|
| 90 |
-
|
| 91 |
-
|
| 92 |
-
|
| 93 |
-
|
| 94 |
-
|
| 95 |
-
|
| 96 |
-
|
| 97 |
-
|
| 98 |
-
|
| 99 |
-
|
| 100 |
-
|
| 101 |
-
|
| 102 |
-
|
| 103 |
-
|
| 104 |
-
|
| 105 |
-
|
|
|
|
| 106 |
|
| 107 |
data_list = []
|
| 108 |
tab, img, txt = [], [], []
|
|
|
|
| 39 |
)
|
| 40 |
])
|
| 41 |
|
| 42 |
+
self.feature_transformer_tab = None
|
| 43 |
+
self.feature_transformer_img = None
|
| 44 |
+
self.feature_transformer_txt = None
|
| 45 |
+
|
| 46 |
def generate_tab_features(self, smiles):
|
| 47 |
mol = Chem.MolFromSmiles(smiles)
|
| 48 |
|
|
|
|
| 84 |
return_tensors: str = "pt",
|
| 85 |
**kwargs
|
| 86 |
):
|
| 87 |
+
if self.transformer_tab_path is None and self.transformer_img_path is None and self.transformer_txt_path is None:
|
| 88 |
+
if task == 'classification':
|
| 89 |
+
model_dir = snapshot_download("SaeedLab/TITAN-BBB", allow_patterns=["normalize_cls_tabular.joblib"])
|
| 90 |
+
transformer_tab_path = os.path.join(model_dir, "normalize_cls_tabular.joblib")
|
| 91 |
+
model_dir = snapshot_download("SaeedLab/TITAN-BBB", allow_patterns=["normalize_cls_image.joblib"])
|
| 92 |
+
transformer_img_path = os.path.join(model_dir, "normalize_cls_image.joblib")
|
| 93 |
+
model_dir = snapshot_download("SaeedLab/TITAN-BBB", allow_patterns=["normalize_cls_text.joblib"])
|
| 94 |
+
transformer_txt_path = os.path.join(model_dir, "normalize_cls_text.joblib")
|
| 95 |
+
|
| 96 |
+
elif task == 'regression':
|
| 97 |
+
model_dir = snapshot_download("SaeedLab/TITAN-BBB", allow_patterns=["normalize_reg_tabular.joblib"])
|
| 98 |
+
transformer_tab_path = os.path.join(model_dir, "normalize_reg_tabular.joblib")
|
| 99 |
+
model_dir = snapshot_download("SaeedLab/TITAN-BBB", allow_patterns=["normalize_reg_image.joblib"])
|
| 100 |
+
transformer_img_path = os.path.join(model_dir, "normalize_reg_image.joblib")
|
| 101 |
+
model_dir = snapshot_download("SaeedLab/TITAN-BBB", allow_patterns=["normalize_reg_text.joblib"])
|
| 102 |
+
transformer_txt_path = os.path.join(model_dir, "normalize_reg_text.joblib")
|
| 103 |
+
|
| 104 |
+
else:
|
| 105 |
+
raise ValueError('task not defined')
|
| 106 |
+
return
|
| 107 |
+
|
| 108 |
+
self.feature_transformer_tab = joblib.load(transformer_tab_path)
|
| 109 |
+
self.feature_transformer_img = joblib.load(transformer_img_path)
|
| 110 |
+
self.feature_transformer_txt = joblib.load(transformer_txt_path)
|
| 111 |
|
| 112 |
data_list = []
|
| 113 |
tab, img, txt = [], [], []
|