[MNT] solve the conflict

learnware/config.py

@@ -63,11 +63,13 @@ LEARNWARE_FOLDER_POOL_PATH = os.path.join(LEARNWARE_POOL_PATH, "learnwares")
 
 DATABASE_PATH = os.path.join(ROOT_DIRPATH, "database")
 STDOUT_PATH = os.path.join(ROOT_DIRPATH, "stdout")
+CACHE_PATH = os.path.join(ROOT_DIRPATH, "cache")
 
 # TODO: Delete them later
 os.makedirs(ROOT_DIRPATH, exist_ok=True)
 os.makedirs(DATABASE_PATH, exist_ok=True)
 os.makedirs(STDOUT_PATH, exist_ok=True)
+os.makedirs(CACHE_PATH, exist_ok=True)
 
 semantic_config = {
     "Data": {
@@ -123,6 +125,7 @@ _DEFAULT_CONFIG = {
     "root_path": ROOT_DIRPATH,
     "package_path": PACKAGE_DIRPATH,
     "stdout_path": STDOUT_PATH,
+    "cache_path": CACHE_PATH,
     "logging_level": logging.INFO,
     "logging_outfile": None,
     "semantic_specs": semantic_config,

learnware/market/heterogeneous/organizer/__init__.py

@@ -21,7 +21,6 @@ class HeteroMapTableOrganizer(EasyOrganizer):
         os.makedirs(hetero_folder_path, exist_ok=True)
         self.market_mapping_path = os.path.join(hetero_folder_path, "model.bin")
         self.hetero_specs_path = os.path.join(hetero_folder_path, "hetero_specifications")
-        self.training_args.update({"cache_dir": hetero_folder_path})
         os.makedirs(self.hetero_specs_path, exist_ok=True)
 
         if os.path.exists(self.market_mapping_path):
@@ -42,7 +41,7 @@ class HeteroMapTableOrganizer(EasyOrganizer):
                     self._update_learnware_by_ids(self.get_learnware_ids(check_status=BaseChecker.USABLE_LEARWARE))
         else:
             logger.warning(f"No market mapping to reload!")
-            self.market_mapping = HeteroMap(cache_dir=hetero_folder_path)
+            self.market_mapping = HeteroMap()
 
     def reset(self, market_id, rebuild=False, auto_update=False, auto_update_limit=100, **training_args):
         self.auto_update = auto_update

learnware/market/heterogeneous/organizer/hetero_map/__init__.py

@@ -40,8 +40,7 @@ class HeteroMap(nn.Module):
         temperature=10,
         base_temperature=10,
         activation="relu",
-        device="cpu",
-        cache_dir=None,
+        device="cuda:0",
         **kwargs,
     ):
         """
@@ -73,8 +72,6 @@ class HeteroMap(nn.Module):
             Activation function for transformer layer, by default "relu"
         device : str, optional
             Device to run the model on, by default "cuda:0"
-        cache_dir : str, optional
-            The cache directory, by default None
         """
         super(HeteroMap, self).__init__()
 
@@ -88,12 +85,11 @@ class HeteroMap(nn.Module):
             "ffn_dim": ffn_dim,
             "projection_dim": projection_dim,
             "activation": activation,
-            "cache_dir": cache_dir,
         }
         self.model_args.update(kwargs)
 
         if feature_tokenizer is None:
-            feature_tokenizer = FeatureTokenizer(cache_dir=cache_dir, **kwargs)
+            feature_tokenizer = FeatureTokenizer(**kwargs)
 
         self.feature_tokenizer = feature_tokenizer
 

learnware/market/heterogeneous/organizer/hetero_map/feature_extractor.py

@@ -8,6 +8,8 @@ import torch.nn.init as nn_init
 from torch import Tensor, nn
 from transformers import BertTokenizerFast
 
+from .....config import C as conf
+
 
 class WordEmbedding(nn.Module):
     """Encode tokens drawn from column names"""
@@ -62,7 +64,6 @@ class FeatureTokenizer:
     def __init__(
         self,
         disable_tokenizer_parallel=True,
-        cache_dir=None,
         **kwargs,
     ):
         """
@@ -71,6 +72,8 @@ class FeatureTokenizer:
         disable_tokenizer_parallel : bool, optional
             true if use extractor for collator function in torch.DataLoader
         """
+        cache_dir = conf["cache_path"]
+        os.makedirs(cache_dir, exist_ok=True)
         self.tokenizer = BertTokenizerFast.from_pretrained("bert-base-uncased", cache_dir=cache_dir)
         self.tokenizer.__dict__["model_max_length"] = 512
         if disable_tokenizer_parallel:  # disable tokenizer parallel