tentative to add random seeds

generic_ner.py

@@ -1,4 +1,21 @@
-import logging
+import torch
+import numpy as np
+import random
+import os
+
+# 1. Set random seeds
+seed = 2025
+torch.manual_seed(seed)
+torch.cuda.manual_seed_all(seed)
+np.random.seed(seed)
+random.seed(seed)
+os.environ["PYTHONHASHSEED"] = str(seed)
+
+# 2. Disable dropout & training randomness
+torch.use_deterministic_algorithms(True, warn_only=True)
+torch.backends.cudnn.deterministic = True
+
+torch.backends.cudnn.benchmark = False
 from transformers import Pipeline
 import numpy as np
 import torch

modeling_stacked.py

@@ -1,3 +1,21 @@
+import torch
+import numpy as np
+import random
+import os
+
+# 1. Set random seeds
+seed = 2025
+torch.manual_seed(seed)
+torch.cuda.manual_seed_all(seed)
+np.random.seed(seed)
+random.seed(seed)
+os.environ["PYTHONHASHSEED"] = str(seed)
+
+# 2. Disable dropout & training randomness
+torch.use_deterministic_algorithms(True, warn_only=True)
+torch.backends.cudnn.deterministic = True
+torch.backends.cudnn.benchmark = False
+
 from transformers.modeling_outputs import TokenClassifierOutput
 import torch
 import torch.nn as nn
@@ -98,8 +116,7 @@ class ExtendedMultitaskTimeModelForTokenClassification(PreTrainedModel):
             bert_kwargs.pop("head_mask", None)
 
         outputs = self.model(**bert_kwargs)
-        # token_output = self.dropout(outputs[0])  # (B, T, H)
-        token_output = outputs[0]  # (B, T, H)
+        token_output = self.dropout(outputs[0])  # (B, T, H)
         hidden_states = list(outputs.hidden_states) if output_hidden_states else None
 
         # Pass through additional transformer layers