task: full_length_llps
max_sequence_length: 2048
optimizer: AdamW
feature_reconstruction_required: true
training_protocol: data/training_plan/