-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathmain_cms_predict_vae.py
285 lines (217 loc) · 12.5 KB
/
main_cms_predict_vae.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
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
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
import os
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
#import setGPU
import tensorflow as tf
import numpy as np
tf.debugging.enable_check_numerics()
import case_paths.util.event_sample as es
import vae.losses as losses
from vae.vae_particle import VAEparticle
import case_paths.util.sample_factory as sf
import case_paths.path_constants.sample_dict_file_parts_input as sdi
import case_paths.path_constants.sample_dict_file_parts_reco as sdr
import case_readers.data_reader as dare
import case_paths.phase_space.cut_constants as cuts
import case_paths.util.experiment as expe
import training as train
import sys
# ********************************************************
# runtime params
# ********************************************************
test_samples = ['QstarToQW_M_2000_mW_170',
'QstarToQW_M_2000_mW_25',
'QstarToQW_M_2000_mW_400',
'QstarToQW_M_2000_mW_80',
'QstarToQW_M_3000_mW_170',
'QstarToQW_M_3000_mW_25',
'QstarToQW_M_3000_mW_400',
'QstarToQW_M_3000_mW_80',
'QstarToQW_M_5000_mW_170',
'QstarToQW_M_5000_mW_25',
'QstarToQW_M_5000_mW_400',
'QstarToQW_M_5000_mW_80',
'RSGravitonToGluonGluon_kMpl01_M_1000',
'RSGravitonToGluonGluon_kMpl01_M_2000',
'RSGravitonToGluonGluon_kMpl01_M_3000',
'RSGravitonToGluonGluon_kMpl01_M_5000',
'WkkToWRadionToWWW_M2000_Mr170',
'WkkToWRadionToWWW_M2000_Mr400',
'WkkToWRadionToWWW_M3000_Mr170',
'WkkToWRadionToWWW_M3000_Mr400',
'WkkToWRadionToWWW_M5000_Mr170',
'WkkToWRadionToWWW_M5000_Mr400',
'WpToBpT_Wp2000_Bp170_Top170_Zbt',
'WpToBpT_Wp2000_Bp25_Top170_Zbt',
'WpToBpT_Wp2000_Bp400_Top170_Zbt',
'WpToBpT_Wp2000_Bp80_Top170_Zbt',
'WpToBpT_Wp3000_Bp170_Top170_Zbt',
'WpToBpT_Wp3000_Bp25_Top170_Zbt',
'WpToBpT_Wp3000_Bp400_Top170_Zbt',
'WpToBpT_Wp3000_Bp80_Top170_Zbt',
'WpToBpT_Wp5000_Bp170_Top170_Zbt',
'WpToBpT_Wp5000_Bp25_Top170_Zbt',
'WpToBpT_Wp5000_Bp400_Top170_Zbt',
'WpToBpT_Wp5000_Bp80_Top170_Zbt',
'qcdSigTest',
'qcdSigMCOrig'
]
#test_samples = ['qcdSigQRTrain','qcdSigQRTest']
#test_samples = ['qcdSigMCOrig']
test_samples = [
'XToYYprimeTo4Q_MX2000_MY170_MYprime400_narrow',
'XToYYprimeTo4Q_MX2000_MY25_MYprime25_narrow',
'XToYYprimeTo4Q_MX2000_MY25_MYprime80_narrow',
'XToYYprimeTo4Q_MX2000_MY400_MYprime170_narrow',
'XToYYprimeTo4Q_MX2000_MY400_MYprime400_narrow',
'XToYYprimeTo4Q_MX2000_MY400_MYprime80_narrow',
'XToYYprimeTo4Q_MX2000_MY80_MYprime170_narrow',
'XToYYprimeTo4Q_MX2000_MY80_MYprime400_narrow',
'XToYYprimeTo4Q_MX2000_MY80_MYprime80_narrow',
'XToYYprimeTo4Q_MX3000_MY170_MYprime170_narrow',
'XToYYprimeTo4Q_MX3000_MY170_MYprime25_narrow',
'XToYYprimeTo4Q_MX3000_MY25_MYprime25_narrow',
# 'XToYYprimeTo4Q_MX3000_MY25_MYprime400_narrow',
'XToYYprimeTo4Q_MX3000_MY25_MYprime80_narrow',
'XToYYprimeTo4Q_MX3000_MY400_MYprime170_narrow',
'XToYYprimeTo4Q_MX3000_MY400_MYprime25_narrow',
'XToYYprimeTo4Q_MX3000_MY400_MYprime400_narrow',
'XToYYprimeTo4Q_MX3000_MY80_MYprime170_narrow',
'XToYYprimeTo4Q_MX3000_MY80_MYprime25_narrow',
'XToYYprimeTo4Q_MX3000_MY80_MYprime400_narrow',
'XToYYprimeTo4Q_MX5000_MY170_MYprime170_narrow',
'XToYYprimeTo4Q_MX5000_MY170_MYprime80_narrow',
'XToYYprimeTo4Q_MX5000_MY25_MYprime170_narrow',
'XToYYprimeTo4Q_MX5000_MY25_MYprime25_narrow',
'XToYYprimeTo4Q_MX5000_MY25_MYprime400_narrow',
'XToYYprimeTo4Q_MX5000_MY25_MYprime80_narrow',
'XToYYprimeTo4Q_MX5000_MY400_MYprime170_narrow',
'XToYYprimeTo4Q_MX5000_MY400_MYprime25_narrow',
'XToYYprimeTo4Q_MX5000_MY400_MYprime400_narrow',
'XToYYprimeTo4Q_MX5000_MY80_MYprime25_narrow',
'XToYYprimeTo4Q_MX5000_MY80_MYprime400_narrow'
]
#test_samples = ['qcdSig','qcdSideExt']
#test_samples = ['qcdSig', 'GtoWW35na']
#test_samples = ['qcdSideExt']
#test_samples = ['gravitonSig']
run_n = int(sys.argv[1])
cuts = cuts.sideband_cuts if 'qcdSideExt' in test_samples else cuts.signalregion_cuts #{}
experiment = expe.Experiment(run_n=run_n).setup(model_dir=True)
batch_n = 2048
print(os.path.join(experiment.model_dir, 'best_so_far'))
# ********************************************
# load model
# ********************************************
vae = VAEparticle.from_saved_model(path=os.path.join(experiment.model_dir, 'best_so_far'))
print('beta factor: ', vae.beta)
loss_fn = losses.threeD_loss
print(sdi.path_dict)
print(sdi.path_dict['sample_names'])
input_paths = sf.SamplePathDirFactory(sdi.path_dict)
print(input_paths)
result_paths = sf.SamplePathDirFactory(sdr.path_dict).update_base_path({'$run$': experiment.run_dir})
for sample_id in test_samples:
# ********************************************
# read test data (events)
# ********************************************
list_ds = tf.data.Dataset.list_files(input_paths.sample_dir_path(sample_id)+'/*')
print(list_ds)
n_testsamples = 200
#if 'Side' not in sample_id and 'qcd' in sample_id:
# n_testsamples = 8
#if 'Side' in sample_id:
# n_testsamples = 1
for file_path in list_ds.take(n_testsamples):
#print("XXX")
file_name = file_path.numpy().decode('utf-8').split(os.sep)[-1]
#if 'bkg' not in file_name:
# continue
#if 'batch19' not in file_name:
# continue
print(file_name)
test_sample = es.CaseEventSample.from_input_file(sample_id, file_path.numpy().decode('utf-8'), **cuts)
test_evts_j1, test_evts_j2 = test_sample.get_particles()
print('{}: {} j1 evts, {} j2 evts'.format(file_path.numpy().decode('utf-8'), len(test_evts_j1), len(test_evts_j2)))
test_j1_ds = tf.data.Dataset.from_tensor_slices(test_evts_j1).batch(batch_n)
test_j2_ds = tf.data.Dataset.from_tensor_slices(test_evts_j2).batch(batch_n)
# *******************************************************
# forward pass test data -> reco and losses
# *******************************************************
print("HEREEE")
print(sdi.path_dict['sample_names'])
print(sample_id)
print('predicting {}'.format(sdi.path_dict['sample_names'][sample_id]))
reco_j1, loss_j1_reco, loss_j1_kl, orig_j1 = train.predict(vae.model, loss_fn, test_j1_ds)
reco_j2, loss_j2_reco, loss_j2_kl, orig_j2 = train.predict(vae.model, loss_fn, test_j2_ds)
#print("AAAAAAAAAAAAA test_j1_ds shape")
#print(orig_j1.shape)
z_mean1, z_log_var1, zs1 = train.predict_with_latent(vae.encoder, loss_fn, test_j1_ds)
z_mean2, z_log_var2, zs2 = train.predict_with_latent(vae.encoder, loss_fn, test_j2_ds)
#print("AAA")
#print(z_mean1[2])
#print("BBB")
#print(z_log_var1[2])
#print("CCC")
#print(zs1[2])
reco_j1_ptetaphi = np.array(reco_j1)
reco_j2_ptetaphi = np.array(reco_j2)
reco_j1_ptetaphi[...,0] = np.sqrt(reco_j1[:,:,0]**2+reco_j1[:,:,1]**2)
reco_j2_ptetaphi[...,0] = np.sqrt(reco_j2[:,:,0]**2+reco_j2[:,:,1]**2)
pt1 = np.array(reco_j1_ptetaphi[...,0])
pt2 = np.array(reco_j2_ptetaphi[...,0])
reco_j1_ptetaphi[:,:,1] = np.arcsinh(np.divide(np.array(reco_j1[:,:,2]),pt1, out=np.zeros_like(pt1), where=pt1!=0)) # eta = arcsinh(pz/pt)
reco_j2_ptetaphi[:,:,1] = np.arcsinh(np.divide(np.array(reco_j2[:,:,2]),pt2, out=np.zeros_like(pt2), where=pt2!=0)) # eta = arcsinh(pz/pt)
reco_j1_ptetaphi[:,:,2] = np.arccos(np.divide(np.array(reco_j1[:,:,1]),pt1, out=np.zeros_like(pt1), where=pt1!=0)) # eta = arcsinh(pz/pt)
reco_j2_ptetaphi[:,:,2] = np.arccos(np.divide(np.array(reco_j2[:,:,1]),pt2, out=np.zeros_like(pt2), where=pt2!=0)) # eta = arcsinh(pz/pt)
x_j1 = np.argsort(np.asarray(reco_j1_ptetaphi)[...,0]*(-1), axis=1)
reco_j1 = np.take_along_axis(np.asarray(reco_j1_ptetaphi), x_j1[...,None], axis=1)
x_j2 = np.argsort(np.asarray(reco_j2_ptetaphi)[...,0]*(-1), axis=1)
reco_j2 = np.take_along_axis(np.asarray(reco_j2_ptetaphi), x_j2[...,None], axis=1)
#print("YYYYYYYYYYYY")
#print(orig_j1[1])
orig_j1_ptetaphi = np.array(orig_j1)
orig_j2_ptetaphi = np.array(orig_j2)
orig_j1_ptetaphi[...,0] = np.sqrt(orig_j1[:,:,0]**2+orig_j1[:,:,1]**2)
orig_j2_ptetaphi[...,0] = np.sqrt(orig_j2[:,:,0]**2+orig_j2[:,:,1]**2)
pt1 = np.array(orig_j1_ptetaphi[...,0])
pt2 = np.array(orig_j2_ptetaphi[...,0])
orig_j1_ptetaphi[:,:,1] = np.arcsinh(np.divide(np.array(orig_j1[:,:,2]),pt1, out=np.zeros_like(pt1), where=pt1!=0)) # eta = arcsinh(pz/pt)
orig_j2_ptetaphi[:,:,1] = np.arcsinh(np.divide(np.array(orig_j2[:,:,2]),pt2, out=np.zeros_like(pt2), where=pt2!=0)) # eta = arcsinh(pz/pt)
orig_j1_ptetaphi[:,:,2] = np.arccos(np.divide(np.array(orig_j1[:,:,1]),pt1, out=np.zeros_like(pt1), where=pt1!=0)) # eta = arcsinh(pz/pt)
orig_j2_ptetaphi[:,:,2] = np.arccos(np.divide(np.array(orig_j2[:,:,1]),pt2, out=np.zeros_like(pt2), where=pt2!=0)) # eta = arcsinh(pz/pt)
orig_x_j1 = np.argsort(np.asarray(orig_j1_ptetaphi)[...,0]*(-1), axis=1)
orig_j1 = np.take_along_axis(np.asarray(orig_j1_ptetaphi), orig_x_j1[...,None], axis=1)
orig_x_j2 = np.argsort(np.asarray(orig_j2_ptetaphi)[...,0]*(-1), axis=1)
orig_j2 = np.take_along_axis(np.asarray(orig_j2_ptetaphi), orig_x_j2[...,None], axis=1)
#print(reco_j1[1])
#print("XXXXXXXXXXX")
#print(orig_j1[1])
#losses_j1 = [losses.total_loss(loss_j1_reco, loss_j1_kl, vae.beta), loss_j1_reco, loss_j1_kl]
#losses_j2 = [losses.total_loss(loss_j2_reco, loss_j2_kl, vae.beta), loss_j2_reco, loss_j2_kl]
losses_j1 = [losses.total_loss(loss_j1_reco, loss_j1_kl, 0.5), loss_j1_reco, loss_j1_kl]
losses_j2 = [losses.total_loss(loss_j2_reco, loss_j2_kl, 0.5), loss_j2_reco, loss_j2_kl]
z_j1 = [z_mean1,z_log_var1,zs1]
z_j2 = [z_mean2,z_log_var2,zs2]
# *******************************************************
# add losses to DataSample and save
# *******************************************************
#reco_sample = es.CaseEventSample(sample_id + 'Reco', particles=[reco_j1, reco_j2], jet_features=test_sample.get_event_features(), particle_feature_names=test_sample.particle_feature_names)
reco_sample = es.CaseEventSample(sample_id + 'Reco', particles=[reco_j1, reco_j2], jet_features=test_sample.get_event_features(), particle_feature_names=test_sample.particle_feature_names, orig_particles = [orig_j1, orig_j2])
for loss, label in zip( losses_j1, ['j1TotalLoss', 'j1RecoLoss', 'j1KlLoss']):
# import ipdb; ipdb.set_trace()
reco_sample.add_event_feature(label, loss)
for loss, label in zip( losses_j2, ['j2TotalLoss', 'j2RecoLoss', 'j2KlLoss']):
reco_sample.add_event_feature(label, loss)
#for e, label in zip( [z_mean1], ['z_mean1_0']):
# print(len(np.array(e)))
# print(np.array(e).shape)
# reco_sample.add_event_feature(label, np.array(e)[:,0])
#for e, label in zip( z_j2, ['z_mean2','z_log_var2','zs2']):
# reco_sample.add_event_feature(label, e)
# *******************************************************
# write predicted data
# *******************************************************
print(sdr.path_dict['sample_names'])
print('writing results for {} to {}'.format(sdr.path_dict['sample_names'][reco_sample.name], os.path.join(result_paths.sample_dir_path(reco_sample.name), file_name)))
#reco_sample.dump(os.path.join(result_paths.sample_dir_path(reco_sample.name, mkdir=True), file_name))
reco_sample.dump_with_orig(os.path.join(result_paths.sample_dir_path(reco_sample.name, mkdir=True), file_name))