fixed model configs not thoroughly passed to the feature extractor

modeling/classify.py

@@ -21,11 +21,7 @@ def __init__(self, model_stem, logger_name=None):
         model_checkpoint = f"{model_stem}.pt"
         model_config = yaml.safe_load(open(model_config_file))
         self.training_labels = train.pretraining_binned_label(model_config)
-        self.featurizer = data_loader.FeatureExtractor(
-            img_enc_name=model_config["img_enc_name"],
-            pos_enc_dim=model_config.get("pos_enc_dim", 0),
-            pos_length=model_config.get("pos_length", 0),
-            pos_unit=model_config.get("pos_unit", 0))
+        self.featurizer = data_loader.FeatureExtractor(**model_config)
         label_count = len(FRAME_TYPES) + 1
         if 'bins' in model_config:
             label_count = len(model_config['bins'].keys()) + 1

modeling/data_loader.py

@@ -56,26 +56,25 @@ class FeatureExtractor(object):
     on a weighed sum of the two.
     """
     img_encoder: backbones.ExtractorModel
-    pos_enc_dim: int
     pos_length: int
     pos_unit: int
     pos_abs_th_front: int
     pos_abs_th_end: int
     pos_vec_coeff: float
     sinusoidal_embeddings: ClassVar[Dict[Tuple[int, int], torch.Tensor]] = {}
 
-    def __init__(self, img_enc_name: str,
-                 pos_enc_dim: int = 512,
+    def __init__(self, 
+                 img_enc_name: str,
                  pos_length: int = 6000000,
                  pos_unit: int = 60000,
                  pos_abs_th_front: int = 3,
                  pos_abs_th_end: int = 10,
-                 pos_vec_coeff: float = 0.5):
+                 pos_vec_coeff: float = 0.5, 
+                 **kwargs):  # to catch unexpected arguments
         """
         Initializes the FeatureExtractor object.
 
         :param img_enc_name: a name of backbone model (e.g. CNN) to use for image vector extraction
-        :param pos_enc_dim: dimension of positional embedding, when not given use 512
         :param pos_length: "width" of positional encoding matrix, actual number of matrix columns is calculated by 
                              pos_length / pos_unit (with default values, that is 100 minutes)
         :param pos_unit: unit of positional encoding in milliseconds (e.g., 60000 for minutes, 1000 for seconds)
@@ -87,7 +86,6 @@ def __init__(self, img_enc_name: str,
             raise ValueError("A image vector model must be specified")
         else:
             self.img_encoder: backbones.ExtractorModel = backbones.model_map[img_enc_name]()
-        self.pos_enc_dim = pos_enc_dim
         self.pos_unit = pos_unit
         self.pos_abs_th_front = pos_abs_th_front
         self.pos_abs_th_end = pos_abs_th_end

modeling/train.py

@@ -1,11 +1,12 @@
 import argparse
+import copy
 import json
 import logging
+import os
 import platform
 import shutil
 import time
 from pathlib import Path
-import copy
 from typing import Union
 
 import numpy as np
@@ -119,12 +120,7 @@ def prepare_datasets(indir, train_guids, validation_guids, configs):
         pre_bin_size = len(FRAME_TYPES) + 1
     train_vimg = valid_vimg = 0
 
-    extractor = data_loader.FeatureExtractor(
-        img_enc_name=configs.get('img_enc_name'),
-        pos_unit=configs['pos_unit'] if configs and 'pos_unit' in configs else 3600000,
-        pos_enc_dim=configs['pos_enc_dim'] if 'pos_enc_dim' in configs else 512,
-        pos_length=configs.get('pos_length')
-    )
+    extractor = data_loader.FeatureExtractor(**config)
 
     for j in Path(indir).glob('*.json'):
         guid = j.with_suffix("").name
@@ -152,6 +148,8 @@ def prepare_datasets(indir, train_guids, validation_guids, configs):
 
 
 def k_fold_train(indir, outdir, config_file, configs, train_id=time.strftime("%Y%m%d-%H%M%S")):
+    os.makedirs(outdir, exist_ok=True)
+
     # need to implement "whitelist"?
     guids = get_guids(indir)
     configs = load_config(configs) if not isinstance(configs, dict) else configs

test/test_pos_enc.py

@@ -15,7 +15,6 @@ class TestPosAbsTh(unittest.TestCase):
     def prep_extractor(th_front, th_end, cols=100):
         extractor = data_loader.FeatureExtractor(
             img_enc_name="mock_model_name",
-            pos_enc_dim=256,
             pos_length=6000000,
             pos_unit=60000,
             pos_abs_th_front=th_front,