| batch_size: 128 | |
| contexual_dim: 512 | |
| data_path: /mnt/lustre/work/eickhoff/esx400/qb/metrics/data/collections | |
| dataset_path: /mnt/lustre/work/eickhoff/esx400/qb/metrics/data/collections | |
| decay_rate: 0.005 | |
| eval_steps: 1000 | |
| gamma: 0.1 | |
| learning_rate: 1e-4 | |
| margin: 1 | |
| max_samples: 1000000 | |
| mixed_data: /mnt/lustre/work/eickhoff/esx400/qb/metrics/data/mixed/nli_para | |
| model_name: gpt2 | |
| num_epochs: 100 | |
| output_path: /mnt/lustre/work/eickhoff/esx400/metrics/curriculum/gpt2s | |
| seed: 42 | |
| step_size: 10 | |
| tokenizer_name: gpt2 | |