reduce learning rate linker

Configuration/config.ini

@@ -4,7 +4,7 @@ transformers = 4.16.2
 [DATASET_PARAMS]
 symbols_vocab_size = 26
 atom_vocab_size = 18
-max_len_sentence = 290
+max_len_sentence = 157
 max_atoms_in_sentence = 440
 max_atoms_in_one_type = 180
 

Linker/Linker.py

@@ -88,8 +88,8 @@ class Linker(Module):
 
         # Learning
         self.cross_entropy_loss = SinkhornLoss()
-        self.optimizer = AdamW(self.parameters(), lr=0.001)
-        self.scheduler = StepLR(self.optimizer, step_size=2, gamma=0.5)
+        self.optimizer = AdamW(self.parameters(), lr=0.0001)
+        self.scheduler = StepLR(self.optimizer, step_size=5, gamma=0.5)
         self.to(self.device)
 
     def load_weights(self, model_file):

NeuralProofNet/NeuralProofNet.py

@@ -42,9 +42,8 @@ class NeuralProofNet(Module):
 
         # Learning
         self.linker_loss = SinkhornLoss()
-        self.linker_optimizer = AdamW(self.linker.parameters(),
-                                      lr=0.001)
-        self.linker_scheduler = StepLR(self.linker_optimizer, step_size=2, gamma=0.5)
+        self.linker_optimizer = AdamW(self.linker.parameters(), lr=0.0001)
+        self.linker_scheduler = StepLR(self.linker_optimizer, step_size=5, gamma=0.5)
 
         self.to(self.device)
 

train_neuralproofnet.py

-import torch
 from Linker import *
 from NeuralProofNet.NeuralProofNet import NeuralProofNet
+from find_config import configurate_linker
 from utils import read_links_csv
+
+import torch
 torch.cuda.empty_cache()
 
+dataset = 'Datasets/gold_dataset_links.csv'
+model_tagger = "models/flaubert_super_98_V2_50e.pt"
 
-# region data
-file_path_axiom_links = 'Datasets/gold_dataset_links.csv'
-df_axiom_links = read_links_csv(file_path_axiom_links)
-# endregion
+configurate_linker(dataset, model_tagger, nb_sentences=1000000000)
 
+df_axiom_links = read_links_csv(dataset)
 
-# region model
-print("#" * 20)
-print("#" * 20)
-model_tagger = "models/flaubert_super_98_V2_50e.pt"
 neural_proof_net = NeuralProofNet(model_tagger)
 neural_proof_net.train_neuralproofnet(df_axiom_links, validation_rate=0.1, epochs=25, pretrain_linker_epochs=25, batch_size=16,
-                                      checkpoint=True, tensorboard=True)
-print("#" * 20)
-print("#" * 20)
-# endregion
\ No newline at end of file
+                                      checkpoint=True, tensorboard=True)
\ No newline at end of file

train_supertagger.py

@@ -11,21 +11,17 @@ bert_model = "flaubert/flaubert_base_cased"
 
 configurate_supertagger(dataset, index_to_super_path, bert_model, nb_sentences=1000000000)
 
-# region data
 df = read_supertags_csv(dataset)
 texts = df['X'].tolist()
 tags = df['Z'].tolist()
 
 index_to_super = load_obj(index_to_super_path)
-# endregion
 
-# region model
 tagger = SuperTagger()
 tagger.create_new_model(len(index_to_super),bert_model,index_to_super)
 ## If you want to upload a pretrained model
 # tagger.load_weights("models/model_check.pt")
 tagger.train(texts, tags, epochs=70, batch_size=16, validation_rate=0.1, 
             tensorboard=True, checkpoint=True)
-# endregion
 
 

utils.py

@@ -20,8 +20,12 @@ def read_links_csv(csv_path, nrows=float('inf'), chunksize=100):
     print("\n" + "#" * 20)
     print("Loading csv...")
 
+    rows = sum(1 for _ in open(csv_path, 'r', encoding="utf8")) - 1  # minus the header
     chunk_list = []
 
+    if rows > nrows:
+        rows = nrows
+
     with tqdm(total=rows, desc='Rows read: ') as bar:
         for chunk in pd.read_csv(csv_path, header=0, converters={'Y': pd.eval, 'Z': pd.eval}, 
                                 chunksize=chunksize, nrows=nrows):