Update tiny models for JanusForConditionalGeneration

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

@@ -0,0 +1,106 @@
+{
+  "architectures": [
+    "JanusForConditionalGeneration"
+  ],
+  "dtype": "float32",
+  "image_token_id": 0,
+  "model_type": "janus",
+  "text_config": {
+    "attention_bias": false,
+    "attention_dropout": 0.0,
+    "attention_probs_dropout_prob": 0.1,
+    "bos_token_id": 100000,
+    "eos_token_id": 100001,
+    "head_dim": 8,
+    "hidden_act": "gelu",
+    "hidden_dropout_prob": 0.1,
+    "hidden_size": 32,
+    "id2label": {
+      "0": "LABEL_0",
+      "1": "LABEL_1",
+      "2": "LABEL_2"
+    },
+    "initializer_range": 0.02,
+    "intermediate_size": 37,
+    "is_training": true,
+    "label2id": {
+      "LABEL_0": 0,
+      "LABEL_1": 1,
+      "LABEL_2": 2
+    },
+    "max_position_embeddings": 512,
+    "mlp_bias": false,
+    "model_type": "llama",
+    "num_attention_heads": 4,
+    "num_choices": 4,
+    "num_hidden_layers": 2,
+    "num_key_value_heads": 4,
+    "pad_token_id": 100002,
+    "pretraining_tp": 1,
+    "rms_norm_eps": 1e-06,
+    "rope_parameters": {
+      "rope_theta": 10000.0,
+      "rope_type": "default"
+    },
+    "seq_length": 7,
+    "tie_word_embeddings": false,
+    "type_sequence_label_size": 2,
+    "type_vocab_size": 16,
+    "use_cache": true,
+    "use_input_mask": true,
+    "use_labels": true,
+    "use_token_type_ids": false,
+    "vocab_size": 100590
+  },
+  "tie_word_embeddings": true,
+  "transformers_version": "5.16.0.dev0",
+  "vision_config": {
+    "attention_bias": true,
+    "attention_dropout": 0.1,
+    "depth": 2,
+    "dropout": 0.1,
+    "hidden_act": "gelu",
+    "hidden_dropout_rate": 0.0,
+    "hidden_size": 32,
+    "image_size": 20,
+    "initializer_range": 0.02,
+    "is_training": true,
+    "layer_norm_eps": 1e-06,
+    "mlp_ratio": 2,
+    "model_type": "janus_vision_model",
+    "num_attention_heads": 4,
+    "num_channels": 3,
+    "num_hidden_layers": 2,
+    "num_image_tokens": 16,
+    "num_key_value_heads": 1,
+    "patch_size": 5,
+    "projection_dim": 32,
+    "projection_dropout": 0.0,
+    "use_labels": true,
+    "use_qk_norm": false,
+    "vision_feature_layer": -1,
+    "vision_feature_select_strategy": "default"
+  },
+  "vq_config": {
+    "base_channels": 32,
+    "channel_multiplier": [
+      1,
+      1
+    ],
+    "double_latent": false,
+    "dropout": 0.0,
+    "embed_dim": 12,
+    "hidden_act": "gelu",
+    "image_token_embed_dim": 32,
+    "in_channels": 3,
+    "initializer_range": 0.02,
+    "latent_channels": 12,
+    "model_type": "janus_vqgan",
+    "num_embeddings": 12,
+    "num_hidden_layers": 2,
+    "num_patches": 4,
+    "num_res_blocks": 2,
+    "out_channels": 3,
+    "projection_dim": 10
+  }
+}

generation_config.json

@@ -0,0 +1,10 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 100000,
+  "eos_token_id": 100001,
+  "output_attentions": false,
+  "output_hidden_states": false,
+  "pad_token_id": 100002,
+  "transformers_version": "5.16.0.dev0",
+  "use_cache": true
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6e472eb5c93375ff58e20710722694eed4e5a2632892cae055233f233b9de0a3
+size 14335060

preprocessor_config.json

@@ -0,0 +1,33 @@
+{
+  "background_color": [
+    127,
+    127,
+    127
+  ],
+  "crop_size": {
+    "height": 20,
+    "width": 20
+  },
+  "do_normalize": true,
+  "do_pad": true,
+  "do_rescale": true,
+  "do_resize": true,
+  "image_mean": [
+    0.48145466,
+    0.4578275,
+    0.40821073
+  ],
+  "image_processor_type": "JanusImageProcessor",
+  "image_std": [
+    0.26862954,
+    0.26130258,
+    0.27577711
+  ],
+  "min_size": 14,
+  "resample": 3,
+  "rescale_factor": 0.00392156862745098,
+  "size": {
+    "height": 20,
+    "width": 20
+  }
+}

tokenizer_config.json

@@ -0,0 +1,25 @@
+{
+  "add_prefix_space": null,
+  "backend": "tokenizers",
+  "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｜>",
+  "image_token": "<image_placeholder>",
+  "is_local": true,
+  "legacy": true,
+  "local_files_only": false,
+  "model_max_length": 16384,
+  "model_specific_special_tokens": {
+    "boi_token": "<begin_of_image>",
+    "eoi_token": "<end_of_image>",
+    "image_token": "<image_placeholder>"
+  },
+  "pad_token": "<｜▁pad▁｜>",
+  "padding_side": "left",
+  "processor_class": "JanusProcessor",
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "use_default_system_prompt": true
+}