| { | |
| "adafactor": true, | |
| "batch_size": 8, | |
| "compile": false, | |
| "depth": 5, | |
| "dev": [], | |
| "encoder": "google/mt5-large", | |
| "epochs": 60, | |
| "exp": "/lnet/work/people/nassajian/coref_project/experiments/system2_best", | |
| "label_smoothing": 0.2, | |
| "learning_rate": 2e-05, | |
| "learning_rate_decay": true, | |
| "load": [ | |
| "/lnet/work/people/nassajian/hf_cache/hub/models--ufal--corpipe25-corefud1.3-large-251101/snapshots/cf37530e02e2dc2b5f09d5dffa8da427e772ee14" | |
| ], | |
| "logdir": "/lnet/work/people/nassajian/coref_project/experiments/system2_best", | |
| "right": 50, | |
| "sampling_exponent": 0.7, | |
| "sampling_mode": "sentences", | |
| "seed": 42, | |
| "segment": 512, | |
| "steps_per_epoch": 200, | |
| "test": null, | |
| "threads": 2, | |
| "train": true, | |
| "treebanks": [ | |
| "/lnet/work/people/nassajian/coref_project/data/mehr_splits/fa_mehr-train.conllu", | |
| "/lnet/work/people/nassajian/coref_project/data/fold1/fa_littleprince-train.conllu" | |
| ], | |
| "warmup": 0.1 | |
| } |