pyg-nightly 2.6.0.dev20240318__py3-none-any.whl → 2.7.0.dev20250115__py3-none-any.whl

torch_geometric/explain/algorithm/base.py

@@ -50,7 +50,6 @@ class ExplainerAlgorithm(torch.nn.Module):
         r"""Checks if the explainer supports the user-defined settings provided
         in :obj:`self.explainer_config`, :obj:`self.model_config`.
         """
-        pass
 
     ###########################################################################
 

torch_geometric/explain/algorithm/graphmask_explainer.py

@@ -1,7 +1,6 @@
 import math
 from typing import List, Optional, Tuple, Union
 
-import numpy as np
 import torch
 import torch.nn.functional as F
 from torch import Tensor
@@ -162,7 +161,7 @@ class GraphMaskExplainer(ExplainerAlgorithm):
                 (torch.log(u) - torch.log(1 - u) + input_element) / beta)
 
             penalty = torch.sigmoid(input_element -
-                                    beta * np.math.log(-gamma / zeta))
+                                    beta * math.log(-gamma / zeta))
         else:
             s = torch.sigmoid(input_element)
             penalty = torch.zeros_like(input_element)

torch_geometric/explain/algorithm/pg_explainer.py

@@ -59,7 +59,7 @@ class PGExplainer(ExplainerAlgorithm):
         'edge_size': 0.05,
         'edge_ent': 1.0,
         'temp': [5.0, 2.0],
-        'bias': 0.0,
+        'bias': 0.01,
     }
 
     def __init__(self, epochs: int, lr: float = 0.003, **kwargs):

torch_geometric/explain/explanation.py

@@ -340,10 +340,10 @@ class HeteroExplanation(HeteroData, ExplanationMixin):
         """
         node_mask_dict = self.node_mask_dict
         for node_mask in node_mask_dict.values():
-            if node_mask.dim() != 2 or node_mask.size(1) <= 1:
+            if node_mask.dim() != 2:
                 raise ValueError(f"Cannot compute feature importance for "
                                  f"object-level 'node_mask' "
-                                 f"(got shape {node_mask_dict.size()})")
+                                 f"(got shape {node_mask.size()})")
 
         if feat_labels is None:
             feat_labels = {}

torch_geometric/graphgym/checkpoint.py

@@ -6,6 +6,7 @@ from typing import Any, Dict, List, Optional, Union
 import torch
 
 from torch_geometric.graphgym.config import cfg
+from torch_geometric.io import fs
 
 MODEL_STATE = 'model_state'
 OPTIMIZER_STATE = 'optimizer_state'
@@ -25,7 +26,7 @@ def load_ckpt(
     if not osp.exists(path):
         return 0
 
-    ckpt = torch.load(path)
+    ckpt = fs.torch_load(path)
     model.load_state_dict(ckpt[MODEL_STATE])
     if optimizer is not None and OPTIMIZER_STATE in ckpt:
         optimizer.load_state_dict(ckpt[OPTIMIZER_STATE])

torch_geometric/graphgym/logger.py

@@ -19,7 +19,7 @@ def set_printing():
     logging.root.handlers = []
     logging_cfg = {'level': logging.INFO, 'format': '%(message)s'}
     os.makedirs(cfg.run_dir, exist_ok=True)
-    h_file = logging.FileHandler('{}/logging.log'.format(cfg.run_dir))
+    h_file = logging.FileHandler(f'{cfg.run_dir}/logging.log')
     h_stdout = logging.StreamHandler(sys.stdout)
     if cfg.print == 'file':
         logging_cfg['handlers'] = [h_file]
@@ -40,7 +40,7 @@ class Logger:
         self._epoch_total = cfg.optim.max_epoch
         self._time_total = 0  # won't be reset
 
-        self.out_dir = '{}/{}'.format(cfg.run_dir, name)
+        self.out_dir = f'{cfg.run_dir}/{name}'
         os.makedirs(self.out_dir, exist_ok=True)
         if cfg.tensorboard_each_run:
             from tensorboardX import SummaryWriter
@@ -210,9 +210,9 @@ class Logger:
             }
 
         # print
-        logging.info('{}: {}'.format(self.name, stats))
+        logging.info(f'{self.name}: {stats}')
         # json
-        dict_to_json(stats, '{}/stats.json'.format(self.out_dir))
+        dict_to_json(stats, f'{self.out_dir}/stats.json')
         # tensorboard
         if cfg.tensorboard_each_run:
             dict_to_tb(stats, self.tb_writer, cur_epoch)

torch_geometric/graphgym/loss.py

@@ -10,7 +10,7 @@ def compute_loss(pred, true):
 
     Args:
         pred (torch.tensor): Unnormalized prediction
-        true (torch.tensor): Grou
+        true (torch.tensor): Ground truth labels
 
     Returns: Loss, normalized prediction score
 

torch_geometric/graphgym/utils/agg_runs.py

@@ -54,7 +54,7 @@ def agg_dict_list(dict_list):
         if key != 'epoch':
             value = np.array([dict[key] for dict in dict_list])
             dict_agg[key] = np.mean(value).round(cfg.round)
-            dict_agg['{}_std'.format(key)] = np.std(value).round(cfg.round)
+            dict_agg[f'{key}_std'] = np.std(value).round(cfg.round)
     return dict_agg
 
 
@@ -107,7 +107,7 @@ def agg_runs(dir, metric_best='auto'):
                     [stats[metric] for stats in stats_list])
                 best_epoch = \
                     stats_list[
-                        eval("performance_np.{}()".format(cfg.metric_agg))][
+                        eval(f"performance_np.{cfg.metric_agg}()")][
                         'epoch']
                 print(best_epoch)
 
@@ -190,7 +190,7 @@ def agg_batch(dir, metric_best='auto'):
             results[key] = pd.DataFrame(results[key])
             results[key] = results[key].sort_values(
                 list(dict_name.keys()), ascending=[True] * len(dict_name))
-            fname = osp.join(dir_out, '{}_best.csv'.format(key))
+            fname = osp.join(dir_out, f'{key}_best.csv')
             results[key].to_csv(fname, index=False)
 
     results = {'train': [], 'val': [], 'test': []}
@@ -213,7 +213,7 @@ def agg_batch(dir, metric_best='auto'):
             results[key] = pd.DataFrame(results[key])
             results[key] = results[key].sort_values(
                 list(dict_name.keys()), ascending=[True] * len(dict_name))
-            fname = osp.join(dir_out, '{}.csv'.format(key))
+            fname = osp.join(dir_out, f'{key}.csv')
             results[key].to_csv(fname, index=False)
 
     results = {'train': [], 'val': [], 'test': []}
@@ -245,7 +245,7 @@ def agg_batch(dir, metric_best='auto'):
             results[key] = pd.DataFrame(results[key])
             results[key] = results[key].sort_values(
                 list(dict_name.keys()), ascending=[True] * len(dict_name))
-            fname = osp.join(dir_out, '{}_bestepoch.csv'.format(key))
+            fname = osp.join(dir_out, f'{key}_bestepoch.csv')
             results[key].to_csv(fname, index=False)
 
-    print('Results aggregated across models saved in {}'.format(dir_out))
+    print(f'Results aggregated across models saved in {dir_out}')