Bugfix/package config edgecases

MaskHIT/maskhit/configs/config_gram_stains.yml

@@ -0,0 +1,42 @@
+dataset:
+  meta_svs: !!str /pool2/users/jackm/Dartmouth_experiments/SlidePreprocessing/for_vit/meta/dartmouth_svs-2.pickle
+  meta_all: !!str /pool2/users/jackm/Dartmouth_experiments/SlidePreprocessing/for_vit/meta/dartmouth_meta-2.pickle
+
+  timestr_model: !!str 2024_2_6-vit
+  outcome: !!str morphology
+  outcome_type: !!str classification
+  study: !!str dartmouth
+  is_cancer: !!bool False
+  disease: !!str dartmouth
+  classes: !!str GN, GP
+
+patch:
+  num_patches: !!int 400
+  magnification: !!float 20.0
+
+model:
+  weighted_loss: !!bool True
+  resume: !!str 
+  resume_epoch: !!str BEST
+
+  regions_per_svs: !!int 64
+
+  # Weight Decays
+  wd_attn: !!float 1e-3
+  wd_fuse: !!float 1e-2
+  wd_loss: !!float 1e-2
+  wd_pred: !!float 1e-2
+
+  # Learning Rates
+  lr_attn: !!float 1e-3 
+  lr_fuse: !!float 1e-3
+  lr_loss: !!float 1e-3
+  lr_pred: !!float 1e-3
+
+  performance_measure: !!str f1
+
+  accumulation_steps: !!int 1
+  dropout: !!float 0.3
+  batch_size: !!int 8
+  override_logs: !!bool True
+  sample_patient: !!bool True

MaskHIT/maskhit/cross_validation.py

@@ -31,16 +31,14 @@
 
 args = opt.parse()
 
+print(f"[INFO] Running cross_validation.py")
 config_file = args.user_config_file
 config_file_default = args.default_config_file
-folds = [int(i) for i in args.folds.split(',')] if args.folds else list(range(config.dataset.num_folds))
-print(f"Testing on folds: {list(folds)}")
 
 # args_config = default_options()
-print(f"config_file: {config_file}")
 config = Config(config_file_default, config_file)
 folds = [int(i) for i in args.folds.split(',')] if args.folds else list(range(config.dataset.num_folds))
-print(f"Testing on folds: {list(folds)}")
+print(f"[INFO] Conducting Cross Validation on folds: {list(folds)}")
 study = config.dataset.study
 timestr = config.dataset.timestr_model
 

MaskHIT/maskhit/quick_test.py

@@ -15,10 +15,10 @@
 
 study = sys.argv[1]
 timestr = sys.argv[2]
-timestr_new = sys.argv[3]
+timestr_test = sys.argv[3]
 
 # the first three arguments passed from `cross_validation.py`
-print(f"{study}, {timestr}, {timestr_new}")
+print(f"{study}, {timestr}, {timestr_test}")
 
 current_directory = Path(__file__).parent
 os.chdir(current_directory)
@@ -44,11 +44,11 @@
     pattern = r'--timestr=[^\s]+'
     org_cmd = re.sub(pattern, '', org_cmd)
 
-    timestr_new += '-test'
     new_cmd = ' '.join([
         'python train.py', org_cmd,
-        f' --mode=test --test-type=test --resume-epoch=BEST --timestr={timestr_new}'
+        f' --resume={ckp} --mode=test --test-type=test --timestr={timestr_test}'
     ])
 
     print(f"executing following new_cmd: {new_cmd}")
     os.system(new_cmd)
+

MaskHIT/maskhit/train.py

@@ -233,7 +233,6 @@ def get_resume_checkpoint(checkpoints_name, epoch_to_resume):
     """
     files = glob.glob(
         os.path.join(args.checkpoints_folder, checkpoints_name, "*.pt"))
-
     checkpoint_to_resume = None
     for fname in files:
         if get_checkpoint_epoch(fname) == epoch_to_resume:
@@ -270,42 +269,44 @@ def prepare_data(meta_split, meta_file, vars_to_include=[]):
         - outcome: patient outcome variable, encoded for classification models e.g. 0, 1, 2 for three classes
 
     """
-
-    #TODO: Temp Duct-tape solution for existing datasets. Need update.
-    #Ideally, the target column name should be parameterized in a config file.
-    ids_to_add = []
-    for index, row in meta_split.iterrows():
-        if 'case_number' in row:
-            value_to_split = row['case_number']
-            split_value = value_to_split.split('.')[0]
-        elif 'barcode' in row:
-            split_value = row['barcode']
-            #TODO: Temp Duct-tape solution: Formatting the ID
-            split_value = "-".join(split_value.split('-')[:3])
-        else:
-            raise ValueError("Row does not contain 'case_number' or 'barcode'")
-
-        ids_to_add.append(split_value)
+    if config.dataset.study == "ibd_project":
+        #Moved specific preprocessing conditional to IBD dataset to run only for that study. 
+        #Other datasets check if affected and modify line 272 to add your study to this conditional.
+        #TODO: Temp Duct-tape solution for existing datasets. Need update.
+        #Ideally, the target column name should be parameterized in a config file.
+        ids_to_add = []
+        for index, row in meta_split.iterrows():
+            if 'case_number' in row:
+                value_to_split = row['case_number']
+                split_value = value_to_split.split('.')[0]
+            elif 'barcode' in row:
+                split_value = row['barcode']
+                #TODO: Temp Duct-tape solution: Formatting the ID
+                split_value = "-".join(split_value.split('-')[:3])
+            else:
+                raise ValueError("Row does not contain 'case_number' or 'barcode'")
+
+            ids_to_add.append(split_value)
 
-    meta_split['id_patient'] = ids_to_add
+        meta_split['id_patient'] = ids_to_add
 
-    #TODO: This whole block should be in another script to be run before train.py
-    if 'id_patient' not in meta_split.columns:
-        patient_ids = []
-        # iterating over the meta_split dataframe
-        for index, row in meta_split.iterrows():
-            #TODO: Debug: This code is basically repeating what we have done in MaskHIT_Prep
+        # TODO: This whole block should be in another script to be run before train.py
+        if 'id_patient' not in meta_split.columns:
+            patient_ids = []
+            # iterating over the meta_split dataframe
+            for index, row in meta_split.iterrows():
+                #TODO: Debug: This code is basically repeating what we have done in MaskHIT_Prep
 
-            # obtaining the paths of the files to the related slide
-            #TODO: Debug: temp fix but need to check with other datasets to see if literal_eval is required and why.
-            #file_names = ast.literal_eval(row['Path'])
-            file_names = str(row['path'])
+                # obtaining the paths of the files to the related slide
+                #TODO: Debug: temp fix but need to check with other datasets to see if literal_eval is required and why.
+                #file_names = ast.literal_eval(row['Path'])
+                file_names = str(row['path'])
 
-            patient_id = file_names[0].split('/')[5].split(' ')[0]
-            patient_ids.append(patient_id) # adding patient id to the list
-        meta_split['id_patient'] = patient_ids # adding column to the meta_split dataframe
-        # formatting rows in meta_file of the id patients so they match that of meta_split df
-        meta_file['id_patient'] = meta_file['id_patient'].apply(lambda x: pd.Series(x.split(' ')[0]))
+                patient_id = file_names[0].split('/')[5].split(' ')[0]
+                patient_ids.append(patient_id) # adding patient id to the list
+            meta_split['id_patient'] = patient_ids # adding column to the meta_split dataframe
+            # formatting rows in meta_file of the id patients so they match that of meta_split df
+            meta_file['id_patient'] = meta_file['id_patient'].apply(lambda x: pd.Series(x.split(' ')[0]))
 
 
     vars_to_keep = ['id_patient']
@@ -343,6 +344,8 @@ def prepare_data(meta_split, meta_file, vars_to_include=[]):
 
 
 def main():
+
+    print(f"[INFO] Running train.py")
 
     if len(args.timestr):
         TIMESTR = args.timestr
@@ -354,7 +357,14 @@ def main():
         model_name = f"{TIMESTR}-{args.fold}"
 
     # if we want to resume previous training
-    if config.model.resume:
+    if args.resume:
+        checkpoint_to_resume = get_resume_checkpoint(args.resume,
+                                                     config.model.resume_epoch)
+        if args.resume_train:
+            # use the model name
+            model_name = config.model.resume
+            TIMESTR = model_name.split('-')[0]
+    elif config.model.resume:
         checkpoint_to_resume = get_resume_checkpoint(config.model.resume,
                                                      config.model.resume_epoch)
         if args.resume_train:
@@ -393,7 +403,7 @@ def main():
     if config.dataset.meta_all is not None:
         meta_all = pd.read_pickle(config.dataset.meta_all)
         #Debug: Checking input data
-        print("Debug: meta_all:\n", meta_all.head(2))
+        # print("Debug: meta_all:\n", meta_all.head(2))
 
         #TODO: mode=extract is not expected in the train_options. Need doc.
         if args.mode == 'extract':
@@ -466,14 +476,15 @@ def main():
                             meta_file=meta_svs, 
                             vars_to_include=vars_to_include)
 
-    print("df_test")
-    print(df_test)
+    print(f"[INFO] Train folds are : {train_folds}")
+    print(f"[INFO] Validation fold is: {val_fold}")
+    print(f"[INFO] Fold {test_fold} saved for testing")
 
     if config.dataset.outcome_type == 'classification':
         num_classes = len(df_train[config.dataset.outcome].unique().tolist())
+
     else:
         num_classes = 1
-    print(f"num_classes: {num_classes}")
 
     if config.model.weighted_loss:
         weight = df_train.shape[0] / df_train[
@@ -499,8 +510,6 @@ def main():
 
     data_dict = {"train": df_train, "val": df_test}
 
-    df_test.to_csv('fold0.csv')
-
     # Simply call main_worker function
     # print(f"Validation Folds: {df_test.fold.unique()}")
     if args.mode == 'test':

MaskHIT/maskhit/trainer/fitter.py

@@ -676,6 +676,19 @@ def fit(self, data_dict, procedure='train'):
         self.es = EarlyStopping(patience=self.args.patience, mode='max')
 
         self.get_logger()
+
+        if self.config.dataset.outcome_type == 'classification':
+            self.writer['meta'].info(f"Running classification between classes: {self.config.dataset.classes}")
+        self.writer['meta'].info(f"Number of classes: {self.num_classes}")
+        self.writer['meta'].info('Training patients:')
+        self.writer['meta'].info('\t'+ data_dict['train'][['id_patient']].to_string()) 
+
+        if procedure == 'train':
+            self.writer['meta'].info('Validation patients:')
+            self.writer['meta'].info('\t'+ data_dict['val'][['id_patient']].to_string()) 
+        elif procedure == 'test':
+            self.writer['meta'].info('Testing patients:')
+            self.writer['meta'].info('\t'+ data_dict['val'][['id_patient']].to_string()) 
 
         # creating an instance of the model
         model = HybridModel(in_dim=self.args.num_features,

MaskHIT/maskhit/utils/config.py

@@ -116,4 +116,4 @@ def _nested_dicts_to_objects(self, config_dict: Dict) -> Dict:
                 result[key] = obj
             else:
                 result[key] = value
-        return result
+        return result