Update tiny models for QianfanOCRVisionModel

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+tokenizer.json filter=lfs diff=lfs merge=lfs -text

chat_template.jinja

@@ -0,0 +1,77 @@
+{%- if messages[0]['role'] == 'system' %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0]['content'] is string %}
+        {{- messages[0]['content'] }}
+    {%- else %}
+        {%- for item in messages[0]['content'] %}
+            {%- if item['type'] == 'text' %}
+                {{- item['text'] }}
+            {%- endif %}
+        {%- endfor %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {{- '<|im_start|>system\n你是Qianfan-VL，由百度智能云研发的多模态大语言模型。<|im_end|>\n' }}
+{%- endif %}
+{%- set ns = namespace(found_last_user=false, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if not ns.found_last_user and message['role'] == 'user' %}
+        {%- set ns.found_last_user = true %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if messages[0]['role'] != 'system' or not loop.first %}
+        {%- if message['role'] == 'user' or (message['role'] == 'system' and not loop.first) %}
+            {%- set append_think = (enable_thinking is defined and enable_thinking and message['role'] == 'user' and loop.index0 == ns.last_query_index) %}
+                {{- '<|im_start|>' + message['role'] + '\n' }}
+            {%- if message['content'] is string %}
+                {{- message['content'] }}
+            {%- else %}
+            {%- for item in message['content'] %}
+                {%- if item['type'] == 'image' %}
+                    {{- '<image>\n' }}
+                {%- elif item['type'] == 'video' %}
+                    {{- '<video>\n' }}
+                {%- elif item['type'] == 'text' %}
+                    {{- item['text'] }}
+                {%- endif %}
+            {%- endfor %}
+        {%- endif %}
+        {%- if append_think %}
+            {{- '<think>' }}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+        {%- elif message['role'] == 'assistant' %}
+            {%- if message['content'] is string %}
+                {%- set raw_content = message['content'] %}
+            {%- else %}
+                {%- set content_ns = namespace(raw='') %}
+                {%- for item in message['content'] %}
+                    {%- if item['type'] == 'text' %}
+                        {%- set content_ns.raw = content_ns.raw + item['text'] %}
+                    {%- endif %}
+                {%- endfor %}
+                {%- set raw_content = content_ns.raw %}
+            {%- endif %}
+            {%- set content = raw_content %}
+            {%- set reasoning_content = '' %}
+            {%- if 'reasoning_content' in message and message['reasoning_content'] is not none %}
+                {%- set reasoning_content = message['reasoning_content'] %}
+            {%- elif '</think>' in raw_content %}
+                {%- set content = raw_content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = raw_content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+            {%- if loop.index0 > ns.last_query_index and reasoning_content %}
+                {{- '<|im_start|>' + message['role'] + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message['role'] + '\n' + content }}
+            {%- endif %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+{%- endif %}

config.json

@@ -0,0 +1,35 @@
+{
+  "architectures": [
+    "QianfanOCRVisionModel"
+  ],
+  "attention_bias": true,
+  "attention_dropout": 0.0,
+  "drop_path_rate": 0.0,
+  "dtype": "float32",
+  "hidden_act": "quick_gelu",
+  "hidden_dropout_prob": 0.0,
+  "hidden_size": 32,
+  "image_size": [
+    32,
+    32
+  ],
+  "initializer_range": 0.02,
+  "intermediate_size": 128,
+  "layer_norm_eps": 1e-06,
+  "layer_scale_init_value": 0.1,
+  "model_type": "qianfan_ocr_vision",
+  "norm_type": "layer_norm",
+  "num_attention_heads": 4,
+  "num_channels": 3,
+  "num_hidden_layers": 2,
+  "patch_size": [
+    4,
+    4
+  ],
+  "projection_dropout": 0.0,
+  "transformers_version": "5.16.0.dev0",
+  "use_absolute_position_embeddings": true,
+  "use_mask_token": false,
+  "use_mean_pooling": true,
+  "use_qk_norm": false
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:09fa96a6206b49d1bde9066a1e12301b488164866823ded620d61b12bc9c1c5e
+size 120560

preprocessor_config.json

@@ -0,0 +1,42 @@
+{
+  "crop_size": {
+    "height": [
+      32,
+      32
+    ],
+    "width": [
+      32,
+      32
+    ]
+  },
+  "crop_to_patches": false,
+  "do_convert_rgb": true,
+  "do_normalize": true,
+  "do_rescale": true,
+  "do_resize": true,
+  "image_mean": [
+    0.48145466,
+    0.4578275,
+    0.40821073
+  ],
+  "image_processor_type": "GotOcr2ImageProcessor",
+  "image_std": [
+    0.26862954,
+    0.26130258,
+    0.27577711
+  ],
+  "max_patches": 12,
+  "min_patches": 1,
+  "resample": 3,
+  "rescale_factor": 0.00392156862745098,
+  "size": {
+    "height": [
+      32,
+      32
+    ],
+    "width": [
+      32,
+      32
+    ]
+  }
+}

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:91c686341cbb6ff3bfd9ddea3a2e2477418ddccadcd7ad7d5ddca9741342c8d7
+size 11615167

tokenizer_config.json

@@ -0,0 +1,23 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "context_image_token": "<IMG_CONTEXT>",
+  "end_image_token": "</img>",
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "is_local": true,
+  "local_files_only": false,
+  "model_max_length": 1010000,
+  "model_specific_special_tokens": {
+    "context_image_token": "<IMG_CONTEXT>",
+    "end_image_token": "</img>",
+    "start_image_token": "<img>"
+  },
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "start_image_token": "<img>",
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

video_preprocessor_config.json

@@ -0,0 +1,42 @@
+{
+  "crop_size": {
+    "height": [
+      32,
+      32
+    ],
+    "width": [
+      32,
+      32
+    ]
+  },
+  "do_convert_rgb": true,
+  "do_normalize": true,
+  "do_rescale": true,
+  "do_resize": true,
+  "do_sample_frames": false,
+  "image_mean": [
+    0.48145466,
+    0.4578275,
+    0.40821073
+  ],
+  "image_std": [
+    0.26862954,
+    0.26130258,
+    0.27577711
+  ],
+  "initial_shift": true,
+  "resample": 3,
+  "rescale_factor": 0.00392156862745098,
+  "return_metadata": false,
+  "size": {
+    "height": [
+      32,
+      32
+    ],
+    "width": [
+      32,
+      32
+    ]
+  },
+  "video_processor_type": "InternVLVideoProcessor"
+}