SupraWeather1.5-Small / training_config.json
AxionLab-official's picture
Upload 4 files
31462fd verified
Raw
History Blame Contribute Delete
368 Bytes
{
"model_name": "SupraWeather-Nano-1.5",
"architecture": "FT-Transformer",
"d_token": 256,
"n_blocks": 4,
"n_heads": 8,
"loss": "FocalLoss(gamma=2.0, alpha=class_balanced)",
"optimizer": "AdamW",
"lr": 0.001,
"weight_decay": 0.0001,
"scheduler": "cosine",
"fp16": true,
"train_samples": 240000,
"val_samples": 30000,
"test_samples": 30000
}