|
| 1 | +# Pipeline configuration |
| 2 | +build: "ocr-evaluation-pipeline-20254003" |
| 3 | +run_name: "ocr_evaluation_run" |
| 4 | + |
| 5 | +# Enable flags |
| 6 | +enable_artifact_metadata: True |
| 7 | +enable_artifact_visualization: True |
| 8 | +enable_cache: False |
| 9 | +enable_step_logs: True |
| 10 | + |
| 11 | +# Pipeline parameters |
| 12 | +parameters: |
| 13 | + mode: "evaluation" # "batch" or "evaluation" |
| 14 | + input_image_folder: "./assets/samples" |
| 15 | + input_image_paths: [] |
| 16 | + selected_models: ["gpt4o", "granite", "llava34b", "pixtral", "gemma3", "llava-phi3"] |
| 17 | + |
| 18 | +# Pipeline settings |
| 19 | +settings: |
| 20 | + docker: |
| 21 | + requirements: |
| 22 | + - "litellm>=1.10.0" |
| 23 | + - "openai>=1.10.0" |
| 24 | + - "pillow>=10.0.0" |
| 25 | + - "ollama>=0.1.0" |
| 26 | + - "jiwer>=3.0.0" |
| 27 | + - "textdistance>=4.6.0" |
| 28 | + - "polars>=0.19.19" |
| 29 | + |
| 30 | +# Component-specific configurations |
| 31 | +steps: |
| 32 | + run_ocr: |
| 33 | + parameters: |
| 34 | + models: ["gpt4o", "granite", "llava34b"] |
| 35 | + image_folder: "./assets/samples" |
| 36 | + image_paths: [] |
| 37 | + custom_prompt: null |
| 38 | + enable_cache: False |
| 39 | + |
| 40 | + save_ocr_results: |
| 41 | + parameters: |
| 42 | + save_locally: True |
| 43 | + output_dir: "ocr_results" |
| 44 | + |
| 45 | + evaluate_models: |
| 46 | + parameters: |
| 47 | + custom_prompt: null |
| 48 | + ground_truth_folder: "ground_truth_texts" |
| 49 | + ground_truth_files: [] |
| 50 | + result_files: [] |
| 51 | + results_dir: "ocr_results" |
| 52 | + visualization_output_dir: "visualizations" |
| 53 | + enable_cache: False |
| 54 | + |
| 55 | + save_visualization: |
| 56 | + parameters: |
| 57 | + save_locally: True |
| 58 | + output_dir: "visualizations" |
| 59 | + |
| 60 | +# Models registry - maintains metadata about available models |
| 61 | +models_registry: |
| 62 | + - name: "mistral/pixtral-12b-2409" |
| 63 | + shorthand: "pixtral" |
| 64 | + ocr_processor: "litellm" |
| 65 | + provider: "mistral" |
| 66 | + |
| 67 | + - name: "gpt-4o-mini" |
| 68 | + shorthand: "gpt4o" |
| 69 | + ocr_processor: "openai" |
| 70 | + |
| 71 | + - name: "gemma3:27b" |
| 72 | + shorthand: "gemma3" |
| 73 | + ocr_processor: "ollama" |
| 74 | + |
| 75 | + - name: "llava:34b" |
| 76 | + shorthand: "llava34b" |
| 77 | + ocr_processor: "ollama" |
| 78 | + |
| 79 | + - name: "llava-phi3" |
| 80 | + shorthand: "llava-phi3" |
| 81 | + ocr_processor: "ollama" |
| 82 | + |
| 83 | + - name: "granite3.2-vision" |
| 84 | + shorthand: "granite" |
| 85 | + ocr_processor: "ollama" |
0 commit comments