config_bsc.json

{
    "lr": 0.0001,
    "batch_size": 128,
    "microbatch": 64,
    "learning_steps": 80000,
    "log_interval": 50,
    "save_interval": 5000,
    "eval_interval": 500,
    "ema_rate": "0.9999",
    "resume_checkpoint": "none",
    "schedule_sampler": "lossaware",
    "diffusion_steps": 2000,
    "noise_schedule": "sqrt",
    "timestep_respacing": "",
    "vocab": "bert",
    "use_plm_init": "no",
    "vocab_size": 0,
    "config_name": "bert-base-chinese",
    "gpt_config_name": "benjamin/gpt2-wechsel-chinese",
    "notes": "folder-notes",
    "data_dir": "processed_data_bsc",
    "dataset": "dataset-name",
    "checkpoint_path": "checkpoint-path/test-run",
    "seq_len": 68,
    "hidden_t_dim": 68,
    "hidden_dim": 256,
    "dropout": 0.1,
    "use_fp16": false,
    "fp16_scale_growth": 0.001,
    "seed": 102,
    "gradient_clipping": -1.0,
    "weight_decay": 0.0,
    "learn_sigma": false,
    "use_kl": false,
    "predict_xstart": true,
    "rescale_timesteps": true,
    "rescale_learned_sigmas": false,
    "sigma_small": false,
    "emb_scale_factor": 1.0,
    "num_transformer_layers": 12,
    "num_transformer_heads": 8,
    "one_noise_step": true,
    "mask_padding": false,
    "celer_only_L1": true,
    "data_split_criterion": "scanpath",
    "corpus": "bsc",
    "inference": "none",
    "n_folds": 5,
    "ablation_type": "none",
    "nll_in_loss": false,
    "load_from_checkpoint": false,
    "load_train_data": "-"
  }