atropos/environments/community/pytorch_optimizer_coding/FOB/baselines/translation.yaml
2025-05-26 14:35:16 +10:00

28 lines
877 B
YAML

task:
name: translation
optimizer:
- name: adamw_baseline
learning_rate: [3.16e-3, 1.e-3, 3.16e-4]
# learning_rate: [3.16e-3, 1.77e-3, 1.e-3, 5.16e-3, 3.16e-4] # finer grid
weight_decay: [1.e-0, 1.e-1, 1.e-2]
- name: adamcpr_fast
learning_rate: [3.16e-3, 1.e-3, 3.16e-4]
# learning_rate: [3.16e-3, 1.77e-3, 1.e-3, 5.16e-3, 3.16e-4] # finer grid
kappa_init_param: [0.5, 1, 2]
engine:
seed: [1, 2, 3]
# data_dir: ./data
# output_dir: ./experiments
plot: false
silent: true
sbatch_script_template: baselines/sbatch_template.sh # adapt the template to your needs
run_scheduler: slurm_array
sbatch_time_factor: 2.0 # increase this for slower machine
sbatch_args:
partition: single # adapt to your cluster
evaluation:
output_types: [pdf]
plot:
x_axis:
- optimizer.kappa_init_param
- optimizer.weight_decay