Changes to config

config.json

@@ -21,7 +21,7 @@
   "torch_dtype": "bfloat16",
   "transformers_version": "4.50.0.dev0",
   "vision_config": {
-    "model_type": "siglip_vision_model"
+    "model_type": "janus_vision_model"
   },
   "vq_config": {
     "depth": 2,

generation_config.json

@@ -2,6 +2,9 @@
   "_from_model_config": true,
   "bos_token_id": 100000,
   "eos_token_id": 100001,
+  "generation_kwargs": {
+    "boi_token_id": 100003
+  },
   "guidance_scale": 5,
   "pad_token_id": 100002,
   "transformers_version": "4.50.0.dev0"

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b899f1e818e24d78515dd15b4553852808af772e54992ed13f387b7276997a83
-size 4178574806
+oid sha256:bd57e5fe8a43ce8f39c624bf0f0561326d5f05fda4e0b852501a642cce6839c3
+size 4178592926

special_tokens_map.json

@@ -10,6 +10,7 @@
     "<|User|>",
     "<|Assistant|>"
   ],
+  "boi_token": "<begin_of_image>",
   "bos_token": {
     "content": "<｜begin▁of▁sentence｜>",
     "lstrip": false,
@@ -17,6 +18,7 @@
     "rstrip": false,
     "single_word": false
   },
+  "eoi_token": "<end_of_image>",
   "eos_token": {
     "content": "<｜end▁of▁sentence｜>",
     "lstrip": false,
@@ -24,6 +26,7 @@
     "rstrip": false,
     "single_word": false
   },
+  "image_token": "<image_placeholder>",
   "pad_token": {
     "content": "<｜▁pad▁｜>",
     "lstrip": false,

tokenizer_config.json

@@ -4735,10 +4735,17 @@
     "<|User|>",
     "<|Assistant|>"
   ],
+  "boi_token": "<begin_of_image>",
   "bos_token": "<｜begin▁of▁sentence｜>",
   "clean_up_tokenization_spaces": false,
+  "eoi_token": "<end_of_image>",
   "eos_token": "<｜end▁of▁sentence｜>",
-  "extra_special_tokens": {},
+  "extra_special_tokens": {
+    "boi_token": "<begin_of_image>",
+    "eoi_token": "<end_of_image>",
+    "image_token": "<image_placeholder>"
+  },
+  "image_token": "<image_placeholder>",
   "legacy": true,
   "model_max_length": 16384,
   "pad_token": "<｜▁pad▁｜>",