|
{ |
|
"dim": 768, |
|
"head": 4, |
|
"layers": 3, |
|
"epochs": 5, |
|
"split": 256, |
|
"max_len": 30, |
|
"do_train": true, |
|
"do_test": false, |
|
"simple": false, |
|
"complex": false, |
|
"fp16": false, |
|
"fp16_opt_level": "O1", |
|
"lr_default": 5e-06, |
|
"load_from": "", |
|
"weight_decay": 0.0, |
|
"gradient_accumulation_steps": 1, |
|
"max_grad_norm": 1.0, |
|
"batch_size": 16, |
|
"model": "bert-base-multilingual-uncased", |
|
"output_dir": "models/gnn_fp16_no_numeric", |
|
"encoding": "gnn", |
|
"max_length": 512, |
|
"max_batch_size": 12, |
|
"id": 1, |
|
"no_numeric": true, |
|
"attr": false, |
|
"attention": "cross" |
|
} |