Closed
Milestone
started on May 10, 2023
DAN-P2: Clean code and remove unused code
Unstarted Issues (open and unassigned)
0
Ongoing Issues (open and assigned)
0
Completed Issues (closed)
38
- Disable MLFlow by default
- Rename models file to encoder
- Merge zero_optimizers and zero_optimizer into a single function
- Remove unused args and methods from GenericTrainingManager
- Remove unused utils functions
- Remove unused image transformations
- Remove `num_op` from PostProcessingModule
- Remove keep_only_ner_tokens and edit_wer_from_string from metrics
- Remove training_info from dataset
- Remove token_line and token_pg from GlobalHTADecoder
- Remove `unchanged_label` from sample
- Remove synthetic data generation doc from README
- Investigate mlflow logging during training test
- Update readme to add development setup/infos
- Remove force_gpu from training parameters
- Keep only seq2seq mode, remove ctc
- keep_only_tokens should be renamed to keep_only_ner_tokens
- Clean the samples passed to the training function
- Save evaluation results in yml file
- Remove code for synthetic data generation from Manager/OCR file
- Remove get_valid_fonts function
- Remove nb_cols from dataset file
- Rename teacher_forcing_scheduler to label_noise_scheduler
- Remove teacher_forcing_error_rate
- Rename apply_teacher_forcing to add_label_noise in training code
- Remove syn_max_lines and syn_prob_lines from training.py
- Remove cer_by_nb_cols metric from training code
- Remove hardware, software, date from saved models
- Remove code for progressively growing models
- Remove coverage_vector from training.py
- Remove the dan/ocr/line/ folder
- Remove new_tokens from OCR code
- Remove edit_and_num_items_for_ged_from_str_simara function
- Remove syn_max_lines, syn_prob_lines metrics
- Remove coverage_vector from predict.py
- Remove loer, precision, map_cer_per_class, layout_precision_per_class_per_threshold
- Remove raw_line_seg_label, paragraphs_label, pages_label
- Remove linear_scheduler
Loading
Loading
Loading