mirror of
https://github.com/docling-project/docling-eval.git
synced 2026-05-17 13:10:47 +00:00
Nikos Livathinos
9d04a56b93
* feat: Extend evaluate_dpbench_on_external_predictions.sh to include visualisations of the evaluations Signed-off-by: Nikos Livathinos <nli@zurich.ibm.com> * fix: Improve error checking in main.py:visualize() Signed-off-by: Nikos Livathinos <nli@zurich.ibm.com> * fix: Improve logging Signed-off-by: Nikos Livathinos <nli@zurich.ibm.com> * feat: Parallelize the computation of PixelLayoutEvaluator at the level of page Signed-off-by: Nikos Livathinos <nli@zurich.ibm.com> * fix: Make DatasetPixelLayoutEvaluation a subclass of DatasetEvaluation Signed-off-by: Nikos Livathinos <nli@zurich.ibm.com> * feat: Parallelize the MarkdownTextEvaluator Signed-off-by: Nikos Livathinos <nli@zurich.ibm.com> * chore: Improve logging Signed-off-by: Nikos Livathinos <nli@zurich.ibm.com> * feat: Parallelize the table evaluation Signed-off-by: Nikos Livathinos <nli@zurich.ibm.com> * fix: Enclose evaluate_tables() in try catch Signed-off-by: Nikos Livathinos <nli@zurich.ibm.com> * fix: Improve logging in TableEvaluator Signed-off-by: Nikos Livathinos <nli@zurich.ibm.com> * fix: Fix loop counter in TableEvaluator and initialize with the correct concurrency Signed-off-by: Nikos Livathinos <nli@zurich.ibm.com> * feat: Refactor ExternalDoclingDocumentLoader to enable caching of loaded documents. Refactor main to initialize a single object of the external loader and refactor all evaluators to receive the loader instead of the raw Path with external predictions. Signed-off-by: Nikos Livathinos <nli@zurich.ibm.com> * fix: Improve logging in ExternalDoclingDocumentLoader Signed-off-by: Nikos Livathinos <nli@zurich.ibm.com> * feat: Extend the evaluate() and the CLI to support multiple modalities and multiple evaluation results: - Refactor the codebase to support the new signature of evaluate() - All modalities share a common ExternalDoclingDocumentLoader object with caching. - Support multiple evaluation results to keep results for all evaluation metrics Signed-off-by: Nikos Livathinos <nli@zurich.ibm.com> * chore: Improve logging in evaluators Signed-off-by: Nikos Livathinos <nli@zurich.ibm.com> * fix: Fix the log level for some messages Signed-off-by: Nikos Livathinos <nli@zurich.ibm.com> * fix: Fix setting the max_new_tokens parameter when initializing GraniteDocling Signed-off-by: Nikos Livathinos <nli@zurich.ibm.com> * fix: Improve API. Code styling Signed-off-by: Nikos Livathinos <nli@zurich.ibm.com> --------- Signed-off-by: Nikos Livathinos <nli@zurich.ibm.com>