refactoring minimal_vlm_pipeline

Signed-off-by: Peter Staar <taa@zurich.ibm.com>

docs/examples/minimal_vlm_pipeline.py

@@ -11,6 +11,10 @@ from docling.datamodel.pipeline_options import (
     InferenceFramework,
     ResponseFormat,
     VlmPipelineOptions,
+    smoldocling_vlm_mlx_conversion_options,
+    smoldocling_vlm_conversion_options,
+    granite_vision_vlm_conversion_options,
+    granite_vision_vlm_ollama_conversion_options,
 )
 from docling.document_converter import DocumentConverter, PdfFormatOption
 from docling.pipeline.vlm_pipeline import VlmPipeline
@@ -33,7 +37,7 @@ pipeline_options.force_backend_text = False
 # pipeline_options.vlm_options = smoldocling_vlm_conversion_options
 
 ## Pick a VLM model. Fast Apple Silicon friendly implementation for SmolDocling-256M via MLX
-## pipeline_options.vlm_options = smoldocling_vlm_mlx_conversion_options
+pipeline_options.vlm_options = smoldocling_vlm_mlx_conversion_options
 
 ## Alternative VLM models:
 # pipeline_options.vlm_options = granite_vision_vlm_conversion_options
@@ -45,7 +49,7 @@ pixtral_vlm_conversion_options = HuggingFaceVlmOptions(
      response_format=ResponseFormat.MARKDOWN,
      inference_framework=InferenceFramework.TRANSFORMERS_LlavaForConditionalGeneration,
 )
-vlm_conversion_options = pixtral_vlm_conversion_options
+pipeline_options.vlm_options = pixtral_vlm_conversion_options
 """
 
 """
@@ -55,7 +59,7 @@ pixtral_vlm_conversion_options = HuggingFaceVlmOptions(
     response_format=ResponseFormat.MARKDOWN,
     inference_framework=InferenceFramework.TRANSFORMERS_LlavaForConditionalGeneration,
 )
-vlm_conversion_options = pixtral_vlm_conversion_options
+pipeline_options.vlm_options = pixtral_vlm_conversion_options
 """
 
 """
@@ -66,16 +70,19 @@ phi_vlm_conversion_options = HuggingFaceVlmOptions(
     response_format=ResponseFormat.MARKDOWN,
     inference_framework=InferenceFramework.TRANSFORMERS_AutoModelForCausalLM,
 )
-vlm_conversion_options = phi_vlm_conversion_options
+pipeline_options.vlm_options = phi_vlm_conversion_options
 """
 
+"""
 pixtral_vlm_conversion_options = HuggingFaceVlmOptions(
     repo_id="mlx-community/pixtral-12b-bf16",
     prompt="Convert this page to markdown. Do not miss any text and only output the bare MarkDown!",
     response_format=ResponseFormat.MARKDOWN,
     inference_framework=InferenceFramework.MLX,
+    scale=1.0,
 )
-vlm_conversion_options = pixtral_vlm_conversion_options
+pipeline_options.vlm_options = pixtral_vlm_conversion_options
+"""
 
 """
 qwen_vlm_conversion_options = HuggingFaceVlmOptions(
@@ -84,11 +91,9 @@ qwen_vlm_conversion_options = HuggingFaceVlmOptions(
     response_format=ResponseFormat.MARKDOWN,
     inference_framework=InferenceFramework.MLX,
 )
-vlm_conversion_options = qwen_vlm_conversion_options
+pipeline_options.vlm_options = qwen_vlm_conversion_options
 """
 
-pipeline_options.vlm_options = vlm_conversion_options
-
 ## Set up pipeline for PDF or image inputs
 converter = DocumentConverter(
     format_options={
@@ -116,19 +121,16 @@ for source in sources:
     res = converter.convert(source)
 
     print("")
-    print(res.document.export_to_markdown())
+    #print(res.document.export_to_markdown())
 
-    for page in res.pages:
+    for i,page in enumerate(res.pages):
         print("")
-        print(f"Predicted page in {pipeline_options.vlm_options.response_format}:")
+        print(f" ---------- Predicted page {i} in {pipeline_options.vlm_options.response_format}:")
         print(page.predictions.vlm_response.text)
+        print(f" ---------- ")
 
-    res.document.save_as_html(
-        filename=Path(f"{out_path}/{res.input.file.stem}.html"),
-        image_mode=ImageRefMode.REFERENCED,
-        labels=[*DEFAULT_EXPORT_LABELS, DocItemLabel.FOOTNOTE],
-    )
-
+    print("===== Final output of the converted document =======")
+    
     with (out_path / f"{res.input.file.stem}.json").open("w") as fp:
         fp.write(json.dumps(res.document.export_to_dict()))
 
@@ -136,19 +138,27 @@ for source in sources:
         out_path / f"{res.input.file.stem}.json",
         image_mode=ImageRefMode.PLACEHOLDER,
     )
-
+    print(f" => produced {out_path / res.input.file.stem}.json")
+    
     res.document.save_as_markdown(
         out_path / f"{res.input.file.stem}.md",
         image_mode=ImageRefMode.PLACEHOLDER,
     )
-
+    print(f" => produced {out_path / res.input.file.stem}.md")
+    
+    res.document.save_as_html(
+        out_path / f"{res.input.file.stem}.html",
+        image_mode=ImageRefMode.EMBEDDED,
+        labels=[*DEFAULT_EXPORT_LABELS, DocItemLabel.FOOTNOTE],
+        # split_page_view=True,
+    )
+    print(f" => produced {out_path / res.input.file.stem}.html")
+    
     pg_num = res.document.num_pages()
     print("")
     inference_time = time.time() - start_time
     print(
         f"Total document prediction time: {inference_time:.2f} seconds, pages: {pg_num}"
     )
-
-print("================================================")
-print("done!")
-print("================================================")
+    print("====================================================")
+