app

.ipynb_checkpoints/auto_evaluator_sl-checkpoint.py

@@ -18,9 +18,9 @@ import chemllama_mtr
 import datamodule_finetune_sl
 # from .model_finetune import CustomFinetuneModel
 import model_finetune_sl
-import utils_sol
+import utils_sl
 
-def auto_evaluator_level_2_sol(
+def auto_evaluator_level_2_sl(
     dir_model_mtr,
     # dir_model_mtr_ep_to_save:str,
     dir_model_ft_to_save:str,
@@ -68,7 +68,7 @@ def auto_evaluator_level_2_sol(
     name_model_ft = f"SolLlama_{solute_or_solvent}"
 
     # array_level_1, model_ft, data_loader_test
-    array_level_1 = auto_evaluator_level_1_sol(
+    array_level_1 = auto_evaluator_level_1_sl(
         model_mtr=model_mtr, 
         dir_model_ft=dir_model_ft, 
         name_model_ft=name_model_ft, 
@@ -93,7 +93,7 @@ def auto_evaluator_level_2_sol(
     
     # return array_level_2
 
-def auto_evaluator_level_1_sol(
+def auto_evaluator_level_1_sl(
     model_mtr, 
     dir_model_ft:str, 
     name_model_ft:str, 
@@ -141,7 +141,7 @@ def auto_evaluator_level_1_sol(
     batch_size_for_train = batch_size_pair[0]
     batch_size_for_valid = batch_size_pair[1]
 
-    data_module = datamodule_finetune_sol.CustomFinetuneDataModule(
+    data_module = datamodule_finetune_sl.CustomFinetuneDataModule(
         solute_or_solvent=solute_or_solvent,
         tokenizer=tokenizer,
         max_seq_length=max_length,
@@ -157,7 +157,7 @@ def auto_evaluator_level_1_sol(
     # Load model and optimizer for finetune
     learning_rate = lr
         
-    model_ft = model_finetune_sol.CustomFinetuneModel(
+    model_ft = model_finetune_sl.CustomFinetuneModel(
         model_mtr=model_mtr,
         steps_per_epoch=steps_per_epoch,
         warmup_epochs=1,
@@ -222,7 +222,7 @@ def auto_evaluator_level_1_sub_sol(
                            {dir_folder}/{name_model_ft}_ep_000
     """
     
-    array_loss_ranking = utils_sol.rank_value_sol(
+    array_loss_ranking = utils_sl.rank_value_sol(
         list_value=list_validation_loss, 
         # dataset_dict=dataset_dict,
         is_loss=True,
@@ -235,7 +235,7 @@ def auto_evaluator_level_1_sub_sol(
     list_level_1 = list()
     for ep in range(len(list_validation_loss)):
 
-        local_model_ft = utils_sol.load_model_ft_with_epoch(
+        local_model_ft = utils_sl.load_model_ft_with_epoch(
             class_model_ft=class_model_ft, 
             target_epoch=ep,
             dir_model_ft=dir_model_ft,
@@ -249,7 +249,7 @@ def auto_evaluator_level_1_sub_sol(
             result_pred.append(result[bat][0].squeeze())
             result_label.append(result[bat][1])
 
-        list_local_model_ft_result = utils_sol.model_evalulator_sol(
+        list_local_model_ft_result = utils_sl.model_evalulator_sol(
             array_predictions=np.vstack(result_pred),
             array_labels=np.vstack(result_label),
             # dataset_dict=dataset_dict, 
@@ -270,7 +270,7 @@ def auto_evaluator_level_1_sub_sol(
     # to get the metric_1 ranking
     array_level_1 = np.array(list_level_1)
     array_metric_1 = array_level_1[:, 2].astype('float32')
-    array_metric_1_ranking = utils_sol.rank_value_sol(list_value=array_metric_1,
+    array_metric_1_ranking = utils_sl.rank_value_sol(list_value=array_metric_1,
                                               # dataset_dict=dataset_dict,
                                               is_loss=False)
 

auto_evaluator_sl.py

@@ -18,9 +18,9 @@ import chemllama_mtr
 import datamodule_finetune_sl
 # from .model_finetune import CustomFinetuneModel
 import model_finetune_sl
-import utils_sol
+import utils_sl
 
-def auto_evaluator_level_2_sol(
+def auto_evaluator_level_2_sl(
     dir_model_mtr,
     # dir_model_mtr_ep_to_save:str,
     dir_model_ft_to_save:str,
@@ -68,7 +68,7 @@ def auto_evaluator_level_2_sol(
     name_model_ft = f"SolLlama_{solute_or_solvent}"
 
     # array_level_1, model_ft, data_loader_test
-    array_level_1 = auto_evaluator_level_1_sol(
+    array_level_1 = auto_evaluator_level_1_sl(
         model_mtr=model_mtr, 
         dir_model_ft=dir_model_ft, 
         name_model_ft=name_model_ft, 
@@ -93,7 +93,7 @@ def auto_evaluator_level_2_sol(
     
     # return array_level_2
 
-def auto_evaluator_level_1_sol(
+def auto_evaluator_level_1_sl(
     model_mtr, 
     dir_model_ft:str, 
     name_model_ft:str, 
@@ -141,7 +141,7 @@ def auto_evaluator_level_1_sol(
     batch_size_for_train = batch_size_pair[0]
     batch_size_for_valid = batch_size_pair[1]
 
-    data_module = datamodule_finetune_sol.CustomFinetuneDataModule(
+    data_module = datamodule_finetune_sl.CustomFinetuneDataModule(
         solute_or_solvent=solute_or_solvent,
         tokenizer=tokenizer,
         max_seq_length=max_length,
@@ -157,7 +157,7 @@ def auto_evaluator_level_1_sol(
     # Load model and optimizer for finetune
     learning_rate = lr
         
-    model_ft = model_finetune_sol.CustomFinetuneModel(
+    model_ft = model_finetune_sl.CustomFinetuneModel(
         model_mtr=model_mtr,
         steps_per_epoch=steps_per_epoch,
         warmup_epochs=1,
@@ -222,7 +222,7 @@ def auto_evaluator_level_1_sub_sol(
                            {dir_folder}/{name_model_ft}_ep_000
     """
     
-    array_loss_ranking = utils_sol.rank_value_sol(
+    array_loss_ranking = utils_sl.rank_value_sol(
         list_value=list_validation_loss, 
         # dataset_dict=dataset_dict,
         is_loss=True,
@@ -235,7 +235,7 @@ def auto_evaluator_level_1_sub_sol(
     list_level_1 = list()
     for ep in range(len(list_validation_loss)):
 
-        local_model_ft = utils_sol.load_model_ft_with_epoch(
+        local_model_ft = utils_sl.load_model_ft_with_epoch(
             class_model_ft=class_model_ft, 
             target_epoch=ep,
             dir_model_ft=dir_model_ft,
@@ -249,7 +249,7 @@ def auto_evaluator_level_1_sub_sol(
             result_pred.append(result[bat][0].squeeze())
             result_label.append(result[bat][1])
 
-        list_local_model_ft_result = utils_sol.model_evalulator_sol(
+        list_local_model_ft_result = utils_sl.model_evalulator_sol(
             array_predictions=np.vstack(result_pred),
             array_labels=np.vstack(result_label),
             # dataset_dict=dataset_dict, 
@@ -270,7 +270,7 @@ def auto_evaluator_level_1_sub_sol(
     # to get the metric_1 ranking
     array_level_1 = np.array(list_level_1)
     array_metric_1 = array_level_1[:, 2].astype('float32')
-    array_metric_1_ranking = utils_sol.rank_value_sol(list_value=array_metric_1,
+    array_metric_1_ranking = utils_sl.rank_value_sol(list_value=array_metric_1,
                                               # dataset_dict=dataset_dict,
                                               is_loss=False)