Hugging Face's logo

OneScience-Group
/
FourCastNet_v2

English
fourcastnet_v2
OneScience
Earth Science
Weather Forecast
Short-to-Medium-Range Weather Forecast
ERA5
FourCastNet
SFNO
Model card Files
xet
 Community
FourCastNet_v2
File size: 3,219 Bytes
eca4864
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
project:
  name: FourCastNet_v2
  seed: 42
  output_dir: ./result
  checkpoint_dir: ./data/checkpoint

data:
  dataset_dir: ./data/era5_fake
  train_years: [2014,2015]
  val_years: [2016]
  test_years: [2018]
  official_splits:
    train_years: [2014, 2015]
    val_years: [2016, 2017]
    test_years: [2018]
  input_steps: 1
  output_steps: 1
  time_step_hours: 6
  grid_shape: [721, 1440]
  normalize: true
  variables:
    - u10m
    - v10m
    - u100m
    - v100m
    - t2m
    - sp
    - msl
    - tcwv
    - u50
    - u100
    - u150
    - u200
    - u250
    - u300
    - u400
    - u500
    - u600
    - u700
    - u850
    - u925
    - u1000
    - v50
    - v100
    - v150
    - v200
    - v250
    - v300
    - v400
    - v500
    - v600
    - v700
    - v850
    - v925
    - v1000
    - z50
    - z100
    - z150
    - z200
    - z250
    - z300
    - z400
    - z500
    - z600
    - z700
    - z850
    - z925
    - z1000
    - t50
    - t100
    - t150
    - t200
    - t250
    - t300
    - t400
    - t500
    - t600
    - t700
    - t850
    - t925
    - t1000
    - r50
    - r100
    - r150
    - r200
    - r250
    - r300
    - r400
    - r500
    - r600
    - r700
    - r850
    - r925
    - r1000

fake_data:
  time_steps_per_year: 20
  chunk_time_steps: 1
  fill_value: 0.0
  materialize_pattern: true

model:
  profile: smoke
  profiles:
    full_resolution:
      img_size: [721, 1440]
      in_channels: 73
      out_channels: 73
      spectral_transform: sht
      filter_type: non-linear
      scale_factor: 6
      embed_dim: 256
      num_layers: 12
      num_blocks: 8
      normalization_layer: instance_norm
      mlp_mode: distributed
      spectral_layers: 3
      complex_activation: real
      hard_thresholding_fraction: 1.0
      big_skip: true
    smoke:
      img_size: [16, 32]
      in_channels: 73
      out_channels: 73
      spectral_transform: sht
      filter_type: linear
      scale_factor: 4
      embed_dim: 8
      num_layers: 2
      num_blocks: 1
      normalization_layer: instance_norm
      mlp_mode: serial
      spectral_layers: 1
      complex_activation: real
      hard_thresholding_fraction: 0.5
      big_skip: true
      # The smoke profile keeps the same equations but reduces the grid/model.

checkpoint:
  initialize_from: scratch
  prefix: model_bak
  finetune_from: ./data/checkpoint/one_step/model_bak.pt
  strict: true

training:
  stage: one_step
  epochs: 3
  batch_size: 1
  num_workers: 0
  learning_rate: 0.0006
  weight_decay: 0.0
  optimizer_betas: [0.9, 0.95]
  max_grad_norm: 32.0
  scheduler: cosine
  amp: false
  max_train_batches: null
  max_val_batches: null
  finetune:
    autoregressive_steps: 2
    epochs: 3
    learning_rate: 0.0001

distributed:
  backend: nccl
  master_addr: 127.0.0.1
  master_port: 29500

inference:
  checkpoint_path: ./data/checkpoint/finetune/model_bak.pt
  rollout_steps: 1
  max_samples: 1
  save_normalized: false
  output_dir: ./result/output

visualization:
  variable: t2m
  sample_index: 0
  output_dir: ./result/figures
  cmap: coolwarm

slurm:
  job_name: fcnv2_train
  nodes: 1
  gpus_per_node: 8
  cpus_per_task: 8
  time: "24:00:00"
  partition: null
  conda_env: fourcastnetv2_develop