Update model arc

chat_template.jinja

@@ -0,0 +1,5 @@
+{%set seps=['
+
+','<｜end▁of▁sentence｜>']%}{%set i=0%}{%for message in messages%}{%if message['role']|lower=='user'%}<|User|>: {%elif message['role']|lower=='assistant'%}<|Assistant|>:{%if not (loop.last and not add_generation_prompt and message['content'][0]['type']=='text' and message['content'][0]['text']=='')%} {%endif%}{%else%}{{message['role'].capitalize()}}: {%endif%}{%for content in message['content']%}{%if content['type']=='image'%}{%if not loop.first%}{{'
+'}}{%endif%}<image_placeholder>{%if not loop.last%}{{'
+'}}{%endif%}{%elif content['type']=='text'%}{%set text=content['text']%}{%if loop.first%}{%set text=text.lstrip()%}{%endif%}{%if loop.last%}{%set text=text.rstrip()%}{%endif%}{%if not loop.first and message['content'][loop.index0-1]['type']=='text'%}{{' '+text}}{%else%}{{text}}{%endif%}{%endif%}{%endfor%}{%if not loop.last or add_generation_prompt%}{%if message['role']|lower=='user'%}{{seps[0]}}{%else%}{{seps[1]}}{%endif%}{%endif%}{%endfor%}{%if add_generation_prompt%}<|Assistant|>:{%endif%}

config.json

@@ -5,27 +5,77 @@
   "image_token_index": 100581,
   "model_type": "janus",
   "text_config": {
+    "attention_bias": false,
+    "attention_dropout": 0.0,
     "bos_token_id": 100000,
     "eos_token_id": 100001,
+    "head_dim": 128,
+    "hidden_act": "silu",
     "hidden_size": 2048,
+    "initializer_range": 0.02,
     "intermediate_size": 5632,
     "max_position_embeddings": 16384,
+    "mlp_bias": false,
     "model_type": "llama",
     "num_attention_heads": 16,
     "num_hidden_layers": 24,
     "num_key_value_heads": 16,
     "pad_token_id": 100002,
+    "pretraining_tp": 1,
+    "rms_norm_eps": 1e-06,
+    "rope_scaling": null,
+    "rope_theta": 10000.0,
     "torch_dtype": "bfloat16",
+    "use_cache": true,
     "vocab_size": 102400
   },
   "torch_dtype": "bfloat16",
-  "transformers_version": "4.50.0.dev0",
+  "transformers_version": "4.52.0.dev0",
   "vision_config": {
-    "model_type": "janus_vision_model"
+    "attention_bias": true,
+    "attention_dropout": 0.0,
+    "depth": 2,
+    "hidden_act": "gelu",
+    "hidden_dropout_rate": 0.0,
+    "hidden_size": 1024,
+    "image_size": 384,
+    "initializer_range": 0.02,
+    "layer_norm_eps": 1e-06,
+    "mlp_ratio": 4.0,
+    "model_type": "janus_vision_model",
+    "num_attention_heads": 16,
+    "num_channels": 3,
+    "num_hidden_layers": 24,
+    "num_image_tokens": 576,
+    "patch_size": 16,
+    "projection_dim": 2048,
+    "projection_dropout": 0.0,
+    "use_qk_norm": false
   },
   "vq_config": {
+    "base_channels": 128,
+    "channel_multiplier": [
+      1,
+      1,
+      2,
+      2,
+      4
+    ],
     "depth": 2,
+    "double_latent": false,
+    "dropout": 0.0,
+    "embed_dim": 8,
+    "hidden_act": "gelu",
+    "image_token_embed_dim": 2048,
+    "in_channels": 3,
+    "initializer_range": 0.02,
+    "latent_channels": 256,
     "model_type": "janus_vqgan",
-    "num_patches": 24
+    "num_embeddings": 16384,
+    "num_hidden_layers": 2,
+    "num_patches": 24,
+    "num_res_blocks": 2,
+    "out_channels": 3,
+    "projection_dim": 2048
   }
 }

generation_config.json

@@ -7,5 +7,5 @@
   },
   "guidance_scale": 5,
   "pad_token_id": 100002,
-  "transformers_version": "4.50.0.dev0"
+  "transformers_version": "4.52.0.dev0"
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:bd57e5fe8a43ce8f39c624bf0f0561326d5f05fda4e0b852501a642cce6839c3
-size 4178592926
+oid sha256:9d1a416f95fb58d6e02858623c9c676003d66006d51fb5d5cc93348ba78cb942
+size 4153396574