fix typos

train.py

@@ -566,7 +566,7 @@ def t1_from_scratch(args):
         test_dataset, collate_fn=deft_collater, batch_size=args.bs, pin_memory=pin_memory)
 
     model, optimizer, prev_epoch, config_dict = t1_model_from_params(
-        None, args.d, attention_type, position_type, tokenizer, max_seq_size, mask_token, pad_token, device, checkpoint_path=args.checkpoint, height=args.height, depth=args.depth, , attention_scaling=attention_scaling)
+        None, args.d, attention_type, position_type, tokenizer, max_seq_size, mask_token, pad_token, device, checkpoint_path=args.checkpoint, height=args.height, depth=args.depth, attention_scaling=attention_scaling)
     run_name = t1_run_name_from_params(args)
 
     if args.logdir is None:
@@ -713,7 +713,7 @@ def finetune_t2(args):
         test_dataset, collate_fn=deft_collater, batch_size=args.bs, pin_memory=pin_memory)
 
     model, optimizer, prev_epoch, config_dict = t2_model_from_params(
-        args.model, args.d, attention_type, position_type, tokenizer, max_seq_size, mask_token, pad_token, device, checkpoint_path=args.checkpoint, height=args.height, depth=args.depth, , attention_scaling=attention_scaling)
+        args.model, args.d, attention_type, position_type, tokenizer, max_seq_size, mask_token, pad_token, device, checkpoint_path=args.checkpoint, height=args.height, depth=args.depth, attention_scaling=attention_scaling)
     run_name = t2_run_name_from_params(args)
 
     try: