freeze_lm: true
freeze_input_proj: false
freeze_output_proj: true
prompt: 'generate a caption' # the prompting information for the enc-side alignment.
train:
    warmup_rate: 0.1
    epochs: 1
    max_length: 512
    max_shard_size: 10GB
    dataset_name_list: ['cc3m_enc', 'webvid_enc', 'audiocap_enc']