Pre-train mT5 on Custom Datastet

[sumanthd17]

Is it possible to pre-train mT5 on a custom dataset. If yes,

1. Should the data be part of TFDS or other file formats supported?
2. Should the data be pre-processed in the format \t before calling the below code? Where task will be the custom task

python -m t5.models.mesh_transformer_main \
  --tpu="${TPU}" \
  --gcp_project="${PROJECT}" \
  --tpu_zone="${ZONE}" \
  --model_dir="${MODEL_DIR}" \
  --gin_file="models/t5.1.1.large.gin" \
  --gin_param="MIXTURE_NAME = '${TASK}'" \
  --gin_param="utils.run.sequence_length = {'inputs': 1024, 'targets': 256}" \
  --gin_param="utils.run.batch_size = ('tokens_per_batch', 1048576)" \
  --gin_param="utils.run.learning_rate_schedule=@learning_rate_schedules.rsqrt_no_ramp_down" \
  --gin_param="run.train_steps = 1000000" \
  --gin_param="utils.tpu_mesh_shape.model_parallelism = 1" \
  --gin_param="utils.tpu_mesh_shape.tpu_topology = 'v3-256'" \
  --eval_mode="perplexity_eval" \
  --eval_gin_param="mesh_eval_dataset_fn.num_eval_examples = 10000" \
  --t5_tfds_data_dir="${BUCKET}/t5-tfds" \
  --module_import="multilingual_t5.tasks"