ml-illustrated / ddbg

DatasetDebugger

Debugger for ML Datasets

Features

from ddbg import DatasetDebugger
ddbg_project = DatasetDebugger.from_recipe('mnist')
ddbg_results = ddbg_project.calc_train_dataset__mislabel_scores()

import matplotlib.pyplot as plt
from ddbg import DdbgVisualize
ddbg_viz = DdbgVisualize( ddbg_project )
# ddbg_results = ddbg_project.load_ddbg_project_results()

ddbg_viz.visualize_top_self_influence( ddbg_results, end_idx=16 )

ddbg_viz.visualize_top_mislabel_score_items( ddbg_results, end_idx=16 )

ddbg_viz.visualize_dataset_embeddings( ddbg_results )

• TODO

Credits