Update 3 files

- /trainer.cli.py
- /trainer.py
- /model.py

model.py

@@ -55,6 +55,7 @@ class Model:
 
             logits = output[0].tolist()
             text = tokenizer.decode(logits)
+            
         return text
 
 

trainer.cli.py

@@ -33,6 +33,8 @@ if __name__ == '__main__':
     tokenizer.train(dataset.text, max_length=config.tokenizer.max_length)
     
     ids = tokenizer.c_encode(dataset.text)
+    
+    config.model.tokenizer = tokenizer
     config.model.params.vocab_size = tokenizer.vocab_size
     
     

trainer.py

@@ -12,6 +12,8 @@ class Trainer:
     def log(self, loss: float):
         print(f"Epoch: {self.epoch} / {self.num_epochs}\t\tBatch: {self.batch} / {self.num_batches}\t\tLoss: {round(loss, 4)}")
         
+        if batch % 20 == 0:
+            print(f'{model.generate_text(self.model.tokenizer, self.config.inference.seed_text, self.config.inference.n_predict)}')
 
 
     def train(self, batches):
@@ -29,4 +31,3 @@ class Trainer:
                 self.optimizer.step()
 
                 self.log(loss.item())
-                #Train.LogStep(infer_config, log_config, epoch, num_epochs, batch, num_batches, loss)