Error while passing Multiple Datasets during Training/Fine-Tuning

Hi there. I was able to generate Dataset (or you can say, run Dataset Preparation steps) however, I am receiving the following error, when I'm trying to pass multiple datasets in the training command, as suggested by you, i.e.

python train.py \
   --config configs/generation/slat_flow_img_dit_L_64l8p2_fp16.json \
   --output_dir outputs/slat_flow_img_dit_L_64l8p2_fp16_1node \
   --data_dir datasets/ABO, datasets/HSSD \
   --profile

It is giving me the following error:

Error: BasicTrainer.run_step() missing 1 required positional argument: 'data_list'

Please confirm if it supports multiple dataset or not, because otherwise it's useless to keep separate datasets and one has to combine everything during data preparation. I tried some changes, but the issue may require deep digging and changes in the whole codebase (responsible for data loading and training).

Start with train.py and BasicTrainer.run_step, then trace how the --data_dir value is parsed and passed as data_list. Reproduce the command using the listed config and dataset directories; done means multiple datasets are accepted without the missing data_list error and training starts.