Upload folder using huggingface_hub

chat_template.jinja

@@ -61,14 +61,31 @@
         {{- '【instructions_start】' + instructions + '【instructions_end】\n'-}}
     {%- endif -%}
     {# Examples Section (only for extraction tasks) #}
-    {%- if examples -%}
-        {{- '【examples_start】\n' -}}
-        {%- for example in examples -%}
-            {{- '【example_input_start】' + render_content(example.input, true)|trim + '【example_input_end】\n' -}}
-            {{- '【example_output_start】' + example.output|trim + '【example_output_end】\n' -}}
-        {%- endfor -%}
-        {{- '【examples_end】\n' -}}
-    {%- endif -%}
+    {%- for message in messages -%}
+        {%- if message.role == 'developer' and 'content' in message -%}
+            {# Validate that there is at least one input and one output contents #}
+            {%- set example_inputs = message.content[:-1] -%}
+            {%- set example_output_part = message.content[-1] -%}
+            {%- if example_inputs|length > 0 -%}
+                {%- if not has_examples.flag -%}
+                    {{- '【examples_start】\n' -}}
+                    {%- set has_examples.flag = true -%}
+                {%- endif -%}
+                {{- '【example_input_start】' + render_content(example_inputs, true)|trim + '【example_input_end】\n' -}}
+                {# Example output: only keep the text of the first output content #}
+                {%- set output_text = '' -%}
+                {%- if example_output_part is string -%}
+                    {%- set output_text = example_output_part -%}
+                {%- elif example_output_part.text is defined -%}
+                    {%- set output_text = example_output_part.text -%}
+                {%- endif -%}
+                {{- '【example_output_start】' + output_text|trim + '【example_output_end】\n' -}}
+                {%- if loop.last and has_examples.flag -%}
+                    {{- '【examples_end】\n' -}}
+                {%- endif -%}
+            {%- endif -%}
+        {%- endif -%}
+    {%- endfor -%}
     {# Previous Output Section #}
     {%- if previous_output -%}
         {{- '【previous_output_start】' + previous_output + '【previous_output_end】\n' -}}
@@ -85,7 +102,9 @@
         {%- set content = render_content(message.content, true)|trim %}
         {{- content + '\n' -}}
     {%- elif message.role == 'assistant' and not loop.last %}
-        {{- raise_exception('Assistant message must be at the end.') }}
+        {# llama.cpp renders a synthetic init example with an assistant turn in
+           the middle; ignore it so valid NuExtract prompts render unchanged.
+           {{- raise_exception('Assistant message must be at the end.') }} #}
     {%- endif %}
 {%- endfor -%}
 {{- '【document_end】<|im_end|>\n' -}}
@@ -117,4 +136,4 @@
     {%- else %}
         {{- '<think>\n' -}}
     {%- endif %}
-{%- endif -%}
+{%- endif -%}

processor_config.json

@@ -1,6 +1,5 @@
 {
   "image_processor": {
-    "data_format": "channels_first",
     "do_convert_rgb": true,
     "do_normalize": true,
     "do_rescale": true,
@@ -10,7 +9,7 @@
       0.5,
       0.5
     ],
-    "image_processor_type": "Qwen2VLImageProcessorFast",
+    "image_processor_type": "Qwen2VLImageProcessor",
     "image_std": [
       0.5,
       0.5,
@@ -28,8 +27,6 @@
   },
   "processor_class": "Qwen3VLProcessor",
   "video_processor": {
-    "data_format": "channels_first",
-    "default_to_square": true,
     "do_convert_rgb": true,
     "do_normalize": true,
     "do_rescale": true,

tokenizer.json

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:06b9509352d2af50381ab2247e083b80d32d5c0aba91c272ca9ff729b6a0e523
-size 19989325
+oid sha256:87a7830d63fcf43bf241c3c5242e96e62dd3fdc29224ca26fed8ea333db72de4
+size 19989343

tokenizer_config.json

@@ -10,6 +10,7 @@
   "errors": "replace",
   "image_token": "<|image_pad|>",
   "is_local": false,
+  "local_files_only": false,
   "model_max_length": 262144,
   "model_specific_special_tokens": {
     "audio_bos_token": "<|audio_start|>",
@@ -24,7 +25,7 @@
   "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
   "processor_class": "Qwen3VLProcessor",
   "split_special_tokens": false,
-  "tokenizer_class": "Qwen2Tokenizer",
+  "tokenizer_class": "TokenizersBackend",
   "unk_token": null,
   "video_token": "<|video_pad|>",
   "vision_bos_token": "<|vision_start|>",