config-ort.yaml

data_path: 'D:/data/cnn_cln'

trainer:
    max_train_steps_per_epoch : null # Maximum train steps per epoch.
    max_val_steps_per_epoch : null # Maximum validation steps per epoch.
    train_batch_size: 32 # Training global batch size.
    val_batch_size: 32 # Validation batch size per GPU. 
    epochs: 3 # Total epochs to run.
    gpu_batch_size_limit : 4 # Max limit for GPU batch size during training.
    disable_tqdm : False
    writers: ["stdout", "aml", "tensorboard"]
    backend: 'ddp-amp-ort'
module:
    max_length_encoder : 1024
    max_length_decoder : 128
wrt:
    tb_log_dir : 'logs'
stat:
    log_steps : 50
chkp:
    checkpoint : True
    delete_existing_checkpoints: False
    save_dir: 'outputs' #aml output path. does not require mounting
    load_dir: null
    load_filename: null

# add more from BartForConditionalGeneration.generate?
generate:
    max_length: 128
    do_sample : False
    num_beams : 5
# support everything in a yaml. ignore (print warning) everything that's not present.
# Do not add the requirement to define anything in the parser other than yamls