{
  "architectures": [
    "GridTSTForTimeSeriesPrediction"
  ],
  "attention_dropout": 0.2,
  "attention_strategy": "channel_first",
  "d_model": 256,
  "dropout": 0.2,
  "ffn_dim": 256,
  "head_dropout": 0.0,
  "init_std": 0.2,
  "label_len": 96,
  "model_type": "gridtst",
  "norm_type": "batchnorm",
  "num_channels": 321,
  "num_heads": 16,
  "num_layers": 3,
  "num_patches": 8,
  "patch_len": 96,
  "qkv_bias": true,
  "revin_affine": false,
  "seq_len": 720,
  "stride": 96,
  "torch_dtype": "float32",
  "transformers_version": "4.35.2"
}