Make assistant response EOS token

config.json

@@ -11,9 +11,8 @@
   "attention_dropout": 0.0,
   "pad_token_id": 151329,
   "eos_token_id": [
-    151329,
-    151336,
-    151338
+    151334,
+    151329
   ],
   "head_dim": 128,
   "hidden_act": "silu",
@@ -45,4 +44,4 @@
   "use_cache": true,
   "use_qk_norm": false,
   "vocab_size": 151552
-}
+}

generation_config.json

@@ -1,8 +1,8 @@
 {
   "_from_model_config": true,
   "eos_token_id": [
-    151329,
-    151334
+    151334,
+    151329
   ],
   "pad_token_id": 151329,
   "transformers_version": "4.54.0"

tokenizer_config.json

@@ -315,7 +315,7 @@
   ],
   "clean_up_tokenization_spaces": false,
   "do_lower_case": false,
-  "eos_token": "<|endoftext|>",
+  "eos_token": "<|im_end|>",
   "extra_special_tokens": {},
   "model_max_length": 128000,
   "pad_token": "<|endoftext|>",