Two new options: truncation length and ban eos token

modules/models.py

@@ -189,7 +189,6 @@ def load_model(model_name):
             pass
     else:
         tokenizer = AutoTokenizer.from_pretrained(Path(f"{shared.args.model_dir}/{shared.model_name}/"))
-    tokenizer.truncation_side = 'left'
 
     print(f"Loaded the model in {(time.time()-t0):.2f} seconds.")
     return model, tokenizer