-
Notifications
You must be signed in to change notification settings - Fork 0
/
run_panguwx2.py
186 lines (176 loc) · 7.75 KB
/
run_panguwx2.py
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
from netCDF4 import Dataset
import numpy as np
import sys, os, shutil
import onnx
import onnxruntime as ort
import dateutils
from datetime import datetime
datapath=sys.argv[1]
datapatho=sys.argv[2]
analdate=sys.argv[3]
nanal=int(sys.argv[4])
grav=9.8066
valid_date=dateutils.dateshift(analdate,6)
lons1d = np.linspace(0,359.75,1440)
lats1d = np.linspace(-90,90,721)[::-1]
nlevs=13
pfull_arr = [50,100,150,200,250,300,400,500,600,700,850,925,1000]
phalf_arr = [25,75,125,175,225,275,325,450,550,650,750,900,950,1050]
charnanal='mem%03i' % nanal
print('ens member ',nanal)
nc = Dataset(os.path.join(datapath,'sanl_%s_fhr06_mem%03i') % (analdate,nanal))
data_upper = np.empty((5,13,721,1440),np.float32)
data_surface = np.empty((4,721,1440),np.float32)
data_upper[0] = nc['z'][0,::-1,...]*grav
data_upper[1] = nc['spfh'][0,::-1,...]
data_upper[2] = nc['tmp'][0,::-1,...]
data_upper[3] = nc['ugrd'][0,::-1,...]
data_upper[4] = nc['vgrd'][0,::-1,...]
data_surface[0] = nc['mslp'][:].squeeze()
nc.close()
nc = Dataset(os.path.join(datapath,'banl_%s_fhr06_mem%03i') % (analdate,nanal))
data_surface[1] = nc['ugrd10m'][:].squeeze()
data_surface[2] = nc['vgrd10m'][:].squeeze()
data_surface[3] = nc['tmp2m'][:].squeeze()
nc.close()
# Set the behavier of onnxruntime
options = ort.SessionOptions()
options.enable_cpu_mem_arena=False
options.enable_mem_pattern = False
options.enable_mem_reuse = False
# Increase the number for faster inference and more memory consumption
omp_num_threads = os.getenv('OMP_NUM_THREADS')
if omp_num_threads is not None:
omp_num_threads = int(omp_num_threads)
else:
omp_num_threads = 1
options.intra_op_num_threads = omp_num_threads
# Set the behavier of cuda provider
cuda_provider_options = {'arena_extend_strategy':'kSameAsRequested',}
# Initialize onnxruntime session for Pangu-Weather Models
ort_session3 = ort.InferenceSession('/work/noaa/gsienkf/whitaker/python/Pangu-Weather/pangu_weather_3.onnx', sess_options=options, providers=['CPUExecutionProvider'])
ort_session6 = ort.InferenceSession('/work/noaa/gsienkf/whitaker/python/Pangu-Weather/pangu_weather_6.onnx', sess_options=options, providers=['CPUExecutionProvider'])
def write_history(output,fhr):
# save to netcdf
# write GFS history file
nc = Dataset(os.path.join(datapatho,'sfg_%s_fhr%02i_mem%03i') % (valid_date,fhr,nanal),'w')
x = nc.createDimension('grid_xt',len(lons1d))
y = nc.createDimension('grid_yt',len(lats1d))
z = nc.createDimension('pfull',nlevs)
zi = nc.createDimension('phalf',nlevs+1)
t = nc.createDimension('time',1)
nchar = nc.createDimension('nchars',20)
pfull = nc.createVariable('pfull',np.float32,'pfull')
phalf = nc.createVariable('phalf',np.float32,'phalf')
phalf[:] = phalf_arr
phalf.units = 'mb'
phalf.units = 'mb'
pfull[:] = pfull_arr
pfull.units = 'mb'
grid_xt = nc.createVariable('grid_xt',np.float64,'grid_xt')
lon = nc.createVariable('lon',np.float64,('grid_yt','grid_xt'))
grid_yt = nc.createVariable('grid_yt',np.float32,'grid_yt')
lat = nc.createVariable('lat',np.float64,('grid_yt','grid_xt'))
time = nc.createVariable('time',np.float64,'time')
time_iso = nc.createVariable('time_iso','S1',('time','nchars'))
time_iso._Encoding = 'ascii'
yyyy,mm,dd,hh = dateutils.splitdate(analdate)
time.units = 'hours since %04i-%02i-%02i %02i:00:00' % (yyyy,mm,dd,hh)
time[0] = fhr
valid_date2=dateutils.dateshift(analdate,fhr)
valid_time = datetime(*dateutils.splitdate(valid_date2))
time_iso[0] = valid_time.isoformat()+'Z'
grid_xt[:] = lons1d
grid_xt.units = 'degrees_E'
lon.units = 'degrees_E'
grid_yt[:] = lats1d
grid_yt.units = 'degrees_N'
lons,lats = np.meshgrid(lons1d,lats1d)
lon[:] = lons; lat[:] = lats
lat.units = 'degrees_N'
gh_var = nc.createVariable('z',np.float32,('time','pfull','grid_yt','grid_xt'),fill_value=9.9e20)
gh_var.cell_methods = "time: point"
gh_var[0,...] = output[0,::-1,...]/grav
spfh_var = nc.createVariable('spfh',np.float32,('time','pfull','grid_yt','grid_xt'),fill_value=9.9e20)
spfh_var.cell_methods = "time: point"
spfh_var[0,...] = output[1,::-1,...]
tmp_var = nc.createVariable('tmp',np.float32,('time','pfull','grid_yt','grid_xt'),fill_value=9.9e20)
tmp_var.cell_methods = "time: point"
tmp_var[0,...] = output[2,::-1,...]
#for k in range(nlevs):
# print(k,ak[k],bk[k],tmp_var[0,k,...].min(),tmp_var[0,k,...].max())
ugrd_var = nc.createVariable('ugrd',np.float32,('time','pfull','grid_yt','grid_xt'),fill_value=9.9e20)
ugrd_var.cell_methods = "time: point"
ugrd_var[0,...] = output[3,::-1,...]
vgrd_var = nc.createVariable('vgrd',np.float32,('time','pfull','grid_yt','grid_xt'),fill_value=9.9e20)
vgrd_var.cell_methods = "time: point"
vgrd_var[0,...] = output[4,::-1,...]
hgtsfc_var = nc.createVariable('hgtsfc',np.float32,('time','grid_yt','grid_xt'),fill_value=9.9e20)
hgtsfc_var.cell_methods = "time: point"
hgtsfc_var[0,...] = np.zeros((721,1440))
pressfc_var = nc.createVariable('mslp',np.float32,('time','grid_yt','grid_xt'),fill_value=9.9e20)
pressfc_var.cell_methods = "time: point"
pressfc_var[0,...] = output_surface[0]
nc.grid='gaussian'
nc.grid_id=1
#nc.ak = ak[:]
#nc.bk = bk[:]
nc.ncnsto = 4
nc.close()
nc = Dataset(os.path.join(datapatho,'bfg_%s_fhr%02i_mem%03i') % (valid_date,fhr,nanal),'w')
x = nc.createDimension('grid_xt',len(lons1d))
y = nc.createDimension('grid_yt',len(lats1d))
z = nc.createDimension('pfull',nlevs)
zi = nc.createDimension('phalf',nlevs+1)
t = nc.createDimension('time',1)
nchar = nc.createDimension('nchars',20)
pfull = nc.createVariable('pfull',np.float32,'pfull')
phalf = nc.createVariable('phalf',np.float32,'phalf')
phalf[:] = phalf_arr
phalf.units = 'mb'
phalf.units = 'mb'
pfull[:] = pfull_arr
pfull.units = 'mb'
grid_xt = nc.createVariable('grid_xt',np.float64,'grid_xt')
lon = nc.createVariable('lon',np.float64,('grid_yt','grid_xt'))
grid_yt = nc.createVariable('grid_yt',np.float32,'grid_yt')
lat = nc.createVariable('lat',np.float64,('grid_yt','grid_xt'))
time = nc.createVariable('time',np.float64,'time')
time_iso = nc.createVariable('time_iso','S1',('time','nchars'))
time_iso._Encoding = 'ascii'
time.units = 'hours since %04i-%02i-%02i %02i:00:00' % (yyyy,mm,dd,hh)
time[0] = fhr
time_iso[0] = valid_time.isoformat()+'Z'
grid_xt[:] = lons1d
grid_xt.units = 'degrees_E'
lon.units = 'degrees_E'
grid_yt[:] = lats1d[::-1]
grid_yt.units = 'degrees_N'
lons,lats = np.meshgrid(lons1d,lats1d)
lon[:] = lons; lat[:] = lats[::-1]
lat.units = 'degrees_N'
nc.grid='gaussian'
nc.grid_id=1
nc.fhzero=3
tmp2m_var = nc.createVariable('tmp2m',np.float32,('time','grid_yt','grid_xt'),fill_value=9.9e20)
tmp2m_var.cell_methods = "time: point"
tmp2m_var[0,...] = output_surface[3]
u10m_var = nc.createVariable('ugrd10m',np.float32,('time','grid_yt','grid_xt'),fill_value=9.9e20)
u10m_var.cell_methods = "time: point"
u10m_var[0,...] = output_surface[1]
v10m_var = nc.createVariable('vgrd10m',np.float32,('time','grid_yt','grid_xt'),fill_value=9.9e20)
v10m_var.cell_methods = "time: point"
v10m_var[0,...] = output_surface[2]
nc.close()
# Run the inference sessions, write output for 3,6 and 9 hour forecasts.
input, input_surface = data_upper, data_surface
fhr=3
output, output_surface = ort_session3.run(None, {'input':input, 'input_surface':input_surface})
write_history(output,fhr)
fhr=6
output, output_surface = ort_session6.run(None, {'input':input, 'input_surface':input_surface})
write_history(output,fhr)
input = output; input_surface = output_surface
fhr=9
output, output_surface = ort_session3.run(None, {'input':input, 'input_surface':input_surface})
write_history(output,fhr)