cvd new data auto train-val

--- dataset_info: features: - name: image dtype: image - name: label dtype: image splits: - name: train num_bytes: 1623749.0 num_examples: 41 download_size: 1563056 datas