remove epoch_sampling_fraction from hash

datamol-io · Jul 16, 2023 · 4b82ba3 · 4b82ba3
1 parent 7c68305
commit 4b82ba3
Show file tree

Hide file tree

Showing 2 changed files with 11 additions and 5 deletions.
diff --git a/expts/neurips2023_configs/debug/config_large_gcn_debug.yaml b/expts/neurips2023_configs/debug/config_large_gcn_debug.yaml
@@ -88,7 +88,7 @@ datamodule:
         sample_size: 2000 # use sample_size for test
         task_level: graph
         splits_path: graphium/data/neurips2023/large-dataset/pcba_1328_random_splits.pt  # Download with `wget https://storage.googleapis.com/graphium-public/datasets/neurips_2023/Large-dataset/pcba_1328_random_splits.pt`
-        epoch_sampling_fraction: 0.5
+        epoch_sampling_fraction: 1.0
 
       pcqm4m_g25:
         df: null
@@ -103,7 +103,7 @@ datamodule:
         label_normalization:
           normalize_val_test: True
           method: "normal"
-        epoch_sampling_fraction: 0.5
+        epoch_sampling_fraction: 1.0
 
       pcqm4m_n4:
         df: null
@@ -119,7 +119,7 @@ datamodule:
         label_normalization:
           normalize_val_test: True
           method: "normal"
-        epoch_sampling_fraction: 0.5
+        epoch_sampling_fraction: 1.0
 
     # Featurization
     prepare_dict_or_graph: pyg:graph
@@ -338,7 +338,7 @@ predictor:
     # weight_decay: 1.e-7
   torch_scheduler_kwargs:
     module_type: WarmUpLinearLR
-    max_num_epochs: &max_epochs 200
+    max_num_epochs: &max_epochs 100
     warmup_epochs: 10
     verbose: False
   scheduler_kwargs:

diff --git a/graphium/data/datamodule.py b/graphium/data/datamodule.py
@@ -1928,9 +1928,15 @@ def get_data_hash(self):
         Get a hash specific to a dataset and smiles_transformer.
         Useful to cache the pre-processed data.
         """
+        args = deepcopy(self.task_specific_args)
+        # pop epoch_sampling_fraction out when creating hash 
+        # so that the data cache does not need to be regenerated
+        # when epoch_sampling_fraction has changed.
+        for task in self.task_specific_args.keys():
+            args[task].pop("epoch_sampling_fraction")
         hash_dict = {
             "smiles_transformer": self.smiles_transformer,
-            "task_specific_args": self.task_specific_args,
+            "task_specific_args": args,
         }
         data_hash = get_md5_hash(hash_dict)
         return data_hash