From a3085dba073fe8bdcfb5120729a84560f5d024c3 Mon Sep 17 00:00:00 2001
From: oobabooga <112222186+oobabooga@users.noreply.github.com>
Date: Sun, 9 Apr 2023 21:19:39 -0300
Subject: [PATCH] Fix LlamaTokenizer eos_token (attempt)

---
 modules/models.py | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/modules/models.py b/modules/models.py
index 75dff911..37e8e78e 100644
--- a/modules/models.py
+++ b/modules/models.py
@@ -174,6 +174,9 @@ def load_model(model_name):
         tokenizer = AutoTokenizer.from_pretrained(Path(f"{shared.args.model_dir}/gpt-j-6B/"))
     elif type(model) is transformers.LlamaForCausalLM:
         tokenizer = LlamaTokenizer.from_pretrained(Path(f"{shared.args.model_dir}/{shared.model_name}/"), clean_up_tokenization_spaces=True)
+        tokenizer.eos_token_id = 2
+        tokenizer.bos_token_id = 1
+        tokenizer.pad_token_id = 0
     else:
         tokenizer = AutoTokenizer.from_pretrained(Path(f"{shared.args.model_dir}/{shared.model_name}/"))
     tokenizer.truncation_side = 'left'