File size: 5,265 Bytes
c649472 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 |
work_dir: ./results
ckpt_dir: ./results/checkpoints/
log_dir: ./results/logs/
print_config: true
ignore_warnings: true
test_after_training: false
save_config_to_wandb: true
verbose: true
seed: 11
torch_matmul_precision: high
name: FV3GFS-Ipol6h
name_suffix: null
suffix: null
trainer:
profiler: {}
_target_: pytorch_lightning.Trainer
accelerator: gpu
devices: 4
strategy: ddp_find_unused_parameters_false
min_epochs: 1
max_epochs: 60
gradient_clip_val: 0.5
accumulate_grad_batches: 3
precision: 16
num_sanity_val_steps: 1
val_check_interval: 1.0
deterministic: true
log_every_n_steps: 50
model:
_target_: src.models.sfno.sfnonet.SphericalFourierNeuralOperatorNet
embed_dim: 256
spectral_transform: sht
filter_type: linear
operator_type: dhconv
num_layers: 8
use_mlp: true
mlp_ratio: 2.0
scale_factor: 1
separable: false
spectral_layers: 3
activation_function: gelu
pos_embed: true
dropout_filter: 0
dropout_mlp: 0.1
pos_emb_dropout: 0.0
drop_path_rate: 0.1
num_blocks: 8
sparsity_threshold: 0.0
normalization_layer: instance_norm
hard_thresholding_fraction: 1.0
checkpointing: 0
time_scale_shift_before_filter: true
loss_function:
_target_: src.losses.losses.LpLoss
p: 2
relative: true
name: SFNO
verbose: true
num_conditional_channels: 0
with_time_emb: true
datamodule:
data_dir: /data
batch_size: 6
eval_batch_size: 2
num_workers: 8
pin_memory: false
persistent_workers: true
drop_last: false
shuffle_train_data: true
verbose: true
window: 1
horizon: 6
max_train_samples: null
max_val_samples: 80
forcing_names:
- DSWRFtoa
in_names:
- HGTsfc
- PRESsfc
- surface_temperature
- air_temperature_0
- air_temperature_1
- air_temperature_2
- air_temperature_3
- air_temperature_4
- air_temperature_5
- air_temperature_6
- air_temperature_7
- specific_total_water_0
- specific_total_water_1
- specific_total_water_2
- specific_total_water_3
- specific_total_water_4
- specific_total_water_5
- specific_total_water_6
- specific_total_water_7
- eastward_wind_0
- eastward_wind_1
- eastward_wind_2
- eastward_wind_3
- eastward_wind_4
- eastward_wind_5
- eastward_wind_6
- eastward_wind_7
- northward_wind_0
- northward_wind_1
- northward_wind_2
- northward_wind_3
- northward_wind_4
- northward_wind_5
- northward_wind_6
- northward_wind_7
out_names:
- PRESsfc
- surface_temperature
- air_temperature_0
- air_temperature_1
- air_temperature_2
- air_temperature_3
- air_temperature_4
- air_temperature_5
- air_temperature_6
- air_temperature_7
- specific_total_water_0
- specific_total_water_1
- specific_total_water_2
- specific_total_water_3
- specific_total_water_4
- specific_total_water_5
- specific_total_water_6
- specific_total_water_7
- eastward_wind_0
- eastward_wind_1
- eastward_wind_2
- eastward_wind_3
- eastward_wind_4
- eastward_wind_5
- eastward_wind_6
- eastward_wind_7
- northward_wind_0
- northward_wind_1
- northward_wind_2
- northward_wind_3
- northward_wind_4
- northward_wind_5
- northward_wind_6
- northward_wind_7
auxiliary_names: null
prediction_horizon: null
_target_: src.datamodules.fv3gfs_ensemble.FV3GFSEnsembleDataModule
data_dir_stats: /data/stats
training_sub_paths: null
prescriber:
_target_: fme.core.prescriber.Prescriber
prescribed_name: surface_temperature
mask_name: ocean_fraction
mask_value: 1
module:
optimizer:
name: adamw
lr: 0.0004
weight_decay: 1.0e-05
eps: 1.0e-08
betas:
- 0.9
- 0.99
scheduler:
_target_: torch.optim.lr_scheduler.CosineAnnealingLR
T_max: 60
monitor: val/avg/crps
mode: min
name: ''
use_ema: true
ema_decay: 0.9999
enable_inference_dropout: true
num_predictions: 16
prediction_inputs_noise: 0.0
logging_infix: ''
log_every_step_up_to: 1000
verbose: true
seed: 11
work_dir: ./results
_target_: src.experiment_types.interpolation.InterpolationExperiment
stack_window_to_channel_dim: true
callbacks:
model_checkpoint:
_target_: pytorch_lightning.callbacks.ModelCheckpoint
monitor: val/avg/crps
mode: min
save_top_k: 1
save_last: true
verbose: true
dirpath: ./results/checkpoints/01HER6GMAMHXPA62240XSV4TD4
filename: FV3GFS-Ipol6h_None_epoch{epoch:03d}_seed11
auto_insert_metric_name: false
watch_model:
_target_: src.utilities.wandb_callbacks.WatchModel
log: all
log_freq: 500
summarize_best_val_metric:
_target_: src.utilities.wandb_callbacks.SummarizeBestValMetric
learning_rate_logging:
_target_: pytorch_lightning.callbacks.LearningRateMonitor
logger:
wandb:
_target_: src.utilities.wandb_callbacks.MyWandbLogger
entity: ai2cm
name: FV3GFS-Ipol6h_SFNO_EMA_256x8h_L2R_44lr_10mlpDr_10dpr_15wd_cos_11seed
tags:
- fv3gfs
- interpolation
notes: '...'
project: FME-v2-salva
group: FV3GFS-Ipol6h_SFNO_EMA_256x8h_L2R_44lr_10mlpDr_10dpr_15wd_cos
resume: allow
reinit: true
mode: online
save_dir: ./results/
offline: false
id: 01HER6GMAMHXPA62240XSV4TD4
log_model: false
prefix: ''
n_gpus: 4
effective_batch_size: 72
|