| { | |
| "architectures": [ | |
| "SimCSEModel" | |
| ], | |
| "base_model": "klue/roberta-large", | |
| "batch_size": 64, | |
| "dataset_dir": "/opt/ml/input/data/training/", | |
| "debug": "False", | |
| "device": "cuda:0", | |
| "eval_steps": 50, | |
| "local_rank": 0, | |
| "lr": 3e-05, | |
| "lr_scheduler_type": "linear", | |
| "max_seq_len": 32, | |
| "num_epochs": 3, | |
| "num_warmup_steps": 0, | |
| "output_dir": "/opt/ml/model/", | |
| "rank": 0, | |
| "save_path": "../model", | |
| "seed": 42, | |
| "temperature": 0.05, | |
| "torch_dtype": "float32", | |
| "transformers_version": "4.33.3", | |
| "use_fp16": "True", | |
| "version": 1.0, | |
| "world_size": 4 | |
| } | |