-
Notifications
You must be signed in to change notification settings - Fork 0
/
coord_effect_study.py
215 lines (162 loc) · 8.32 KB
/
coord_effect_study.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
import os
import torch
import numpy as np
from random import shuffle
import argparse
from model.networks import MyModel
import matplotlib.pyplot as plt
import matplotlib.gridspec as gridspec
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
# ploting_dict = {0.5:(-2.0, 0.5), 0.7:(-2.0,0.5), 0.9:(-2.0,0.5)}
yb = (-2.2, 0.5)
def Loading_model(args, device=device, model_name='MyModel'):
if model_name == 'MyModel':
model = MyModel(args.num_feature).to(device)
file_name = f'{args.num_feature}_{args.epochs}epochs_{args.k}_model.pth'
# file_name = '6_500epochs_51_model_lr6e-4maxmax.pth'
file_path = os.path.join(args.checkpoint_path, file_name)
if os.path.isfile(file_path):
Checkpoint = torch.load(file_path)
model.load_state_dict(Checkpoint['model_state_dict'])
print('Loading model successfully!')
return model
else:
print(f'No {file_name} file!')
os._exit(0)
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('--epochs', type=int, default=1001, help='Numbers of Epoch to train')
parser.add_argument('--k',type=int, default=5, help='k-fold')
parser.add_argument('--checkpoint_path', type=str, default='checkpoint/')
parser.add_argument('--save_path', type=str, default='result/coord_effect')
parser.add_argument('--num_feature', type=int, default=6)
args = parser.parse_args()
model = Loading_model(args)
# elements_dict maps atomic number of an element (Ru, Rh, Ir, Pt or Pd)
# to the first 3 features used in the neural network model (period number,
# group number, and electronegativity). Note that these features are
# referenced to Ru (a simplified normalization)
# elements_dict = {44: [0,0,0], 45: [0,1,0], 46: [0,2,3], 77: [1,1,1.5], 78: [1,2,4.5],}
elements_dict = [[5.0,1.338,0,0,2.20,101.07],
[5.0,1.345,0,0,2.28,102.906],
[5.0,1.375,0,0,2.20,106.42],
[6.0,1.357,0,0,2.20,192.20],
[6.0,1.387,0,0,2.28,195.08]]
# Radius/Å Pauling electronegativity VEC
# 44 1.338 2.20 8
# 45 1.345 2.28 9
# 46 1.375 2.20 10
# 77 1.357 2.20 9
# 78 1.387 2.28 10 相对分子量 摩尔质量
coord_nums_dict = {}; gen_coord_nums_dict = {}
# 'coord_nums.csv' file contains the coordination environment information
# of different active sites on different crystal surfaces, as discussed
# in the paperK
with open('data/coord_nums.csv') as f:
for l in f.readlines():
items = l.split(',')
label = items[0]
coord_nums_dict[label] = list(map(int, items[1:]))
gen_coord_nums_dict[label] = sum(coord_nums_dict[label][2:]) / 12.
# normalized by 12, which is the maximal coord. num. in fcc structure
gen_coord_nums = np.array(list(gen_coord_nums_dict.values()))
reorder_idx = gen_coord_nums.argsort()
elements = elements_dict * 5
results = {}
results_averages = {}
for envir in coord_nums_dict:
coord_nums = coord_nums_dict[envir]
new_input = []
for i in range(10000):
shuffle(elements)
new_result = []
#= coordination numbers setting
for j in range(len(coord_nums)):
temp = elements[j].copy()
new_result += [temp]
new_result[j][2] = coord_nums[j]
#= acitvate site setting
if j == 0 or j == 1:
new_result[j][3] = 1
for j in range(len(coord_nums), len(elements)):
new_result += [[5.0,1.338,0,0,2.20,101.07]] # [elements[j] + [11, 0, ]]
new_input += [new_result]
#! new_results.shape: (10000, 25, 5)
new_input = np.array(new_input).astype('float32')
new_input = (new_input - np.mean(new_input, axis=1, keepdims=True)) / np.mean(new_input, axis=1, keepdims=True)
new_input = torch.tensor(new_input, dtype=torch.float32).to(device) # array([1]) used for placeholder
model.eval()
new_results = model(new_input)
new_results = new_results.detach().cpu().numpy().reshape(-1).tolist()
discretized_results = {}
for i in range(int(min(new_results)*100-1), int(max(new_results)*100+1)):
discretized_results[i] = 0
for i in new_results:
discretized_results[int(i*100-1)] += 1
#= ？？
results[envir] = []
for i in discretized_results:
results[envir].append([i/100, discretized_results[i]])
results[envir] = np.array(results[envir])
results_averages[envir] = np.array(new_results).mean()
print('%s %f' % (envir, np.array(new_results).std()))
#================ ploting ====================
#plt.figure(figsize = (1, len(coord_nums_dict)))
plt.figure(figsize=(16, len(coord_nums_dict)))
gs = gridspec.GridSpec(1, len(coord_nums_dict)+4, width_ratios=[0.75,]*len(results) + [0.2, 6, 0.2, 2])
gs.update(wspace=0.0, hspace=0.1)
for i in range(len(coord_nums_dict)):
ax = plt.subplot(gs[i], xlim=(-0.3, 3), ylim= yb,)
# ax = plt.subplot(gs[i])
envir = list(coord_nums_dict.keys())[reorder_idx[i]]
plt.fill(results[envir][:,1]/125, results[envir][:,0], color=[0.5,0.5,1])
plt.scatter([0., ], [results_averages[envir],], color='black', marker='x', zorder=3)
plt.text(1.0, 0.35, envir[:-6]+'\n'+envir[-6:], ha='center', va='top', fontsize=10)
if i == 0:
ax.set_ylabel('Neural network-predicted \n' +
r'$\Delta E_{\mathrm{OH}}-\Delta E_{\mathrm{OH, Pt(111)}}$ (eV)')
ax.set_xticklabels([])
ax.tick_params(bottom=False)
if i > 0:
ax.set_yticklabels([])
ax.spines['left'].set_color('white')
ax.tick_params(left=False)
if i < len(results) - 1:
ax.spines['right'].set_color('white')
if i == int(len(results)/2):
ax.set_xlabel('\nRelative frequency')
ax.tick_params(direction='in', )
ax.tick_params(right=True)
ax.legend(['Frequency distribution', 'Mean of distribution'], \
fancybox=False, edgecolor='black', loc='lower right', fontsize=12)
ax = plt.subplot(gs[-3], ylim=yb)
ax.scatter(list(gen_coord_nums_dict.values()), list(results_averages.values()), zorder=3, color='black', marker='x',)
ax.set_xlabel('Total CN of nearest neighbours')
ax.tick_params(direction='in', right=True, top=True)
ax.set_yticklabels([])
from numpy import polyfit
results_averages = np.array(list(results_averages.values()))
a, b = polyfit(list(gen_coord_nums_dict.values()), results_averages, deg=1)
plt.plot([gen_coord_nums[reorder_idx[0]], gen_coord_nums[reorder_idx[-1]]], \
[a*gen_coord_nums[reorder_idx[0]]+b, a*gen_coord_nums[reorder_idx[-1]]+b],
color='blue', zorder=1)
R_2 = 1 - sum((gen_coord_nums*a+b-results_averages)**2) / sum((results_averages-results_averages.mean())**2)
MAE = abs(gen_coord_nums*a+b-results_averages).mean()
RMSE = (((gen_coord_nums*a+b-results_averages)**2).mean())**0.5
ax.text(10, -1.5, '$R^2$: %.2f\nMAE: %.2f eV\nRMSE: %.2f eV\n'%(R_2, MAE, RMSE),ha='right', va='bottom', fontsize=10)
ax.legend(['Linear fit', 'Mean of distribution'], fancybox=False, edgecolor='black', loc='lower right', fontsize=12)
ax = plt.subplot(gs[-1], xlim=(-2., 0.5), ylim=yb)
# ax.plot([-2.7+0.8, -1.0+0.8, 1.72-0.5-0.8], [-2.7, -1.0, 0.5], color='blue',)
ax.plot([yb[0]+0.8, -0.5+0.8, 0.5-yb[1]-0.8], [yb[0], -0.5, yb[1]], color='blue',)
ax.scatter(results_averages+0.8, results_averages, zorder=3, color='black', marker='x',)
ax.set_xlabel('Limiting\npotential (V)')
ax.tick_params(direction='in', right=True, top=True)
ax.set_yticklabels([])
print('='*30+f"\na:{a}\tb:{b}")
#? save image
if not os.path.exists(args.save_path):
os.mkdir(args.save_path)
# plt.savefig(os.path.join(args.save_path, f'{args.epochs}epochs.pdf'), format='pdf',bbox_inches = 'tight',dpi=700)
plt.savefig(os.path.join(args.save_path, f'6_{args.epochs}_{args.k}_coord_effect.svg'), bbox_inches='tight', dpi=1200)
plt.show()
# a:0.24662662415136724 b:-3.169708219866001