Add posterior mean/stddev into agent_data

scripts/kafka_consumer_iterate_1LL09_RM.py

@@ -102,6 +102,8 @@
                     'ZnI':'infusion_rate_I2', 
                     'ZnCl':'infusion_rate_Cl'}
 
+new_points_label = ['infusion_rate_CsPb', 'infusion_rate_Br', 'infusion_rate_I2', 'infusion_rate_Cl']
+
 use_good_bad = True
 post_dilute = True
 write_agent_data = True
@@ -417,7 +419,7 @@ def print_message(consumer, doctype, doc,
                             if USE_AGENT_iterate:
 
                                 # print(f"\ntelling agent {agent_data}")
-                                agent = build_agen2(peak_target=peak_target)
+                                agent = build_agen2(peak_target=peak_target, agent_data_path=agent_data_path)
 
                                 if len(agent.table) < 2:
                                     acq_func = "qr"
@@ -530,8 +532,15 @@ def print_message(consumer, doctype, doc,
                 if post_dilute:
                     set_target_list = [0 for i in range(len(pump_list))]
                     # rate_list = new_points['points'].tolist()[0][:-1] + [new_points['points'].sum()]
-                    rate_list = [rr for rr in new_points['points'].tolist()[0] if rr!=0] + [new_points['points'].sum()]
-                    rate_list = np.asarray(rate_list)
+                    # rate_list = [rr for rr in new_points['points'].tolist()[0] if rr!=0] + [new_points['points'].sum()]
+                    # rate_list = np.asarray(rate_list)
+                    rate_list = []
+                    for i in new_points_label:
+                        for key in new_points['points']:
+                            if i == key:
+                                rate_list.append(new_points['points'][key][0])
+                    rate_list.insert(2, sum(rate_list)/10)
+                    rate_list.append(sum(rate_list)*5)
 
                 else:
                     # set_target_list = [0 for i in range(new_points['points'].shape[1])]

scripts/kafka_consumer_iterate_XPD_RM.py

@@ -116,6 +116,7 @@
 USE_AGENT_iterate = False
 peak_target = 515
 if USE_AGENT_iterate:
+    import torch
     from prepare_agent_pdf import build_agen
     agent = build_agen(peak_target=peak_target, agent_data_path=agent_data_path)
 
@@ -549,7 +550,7 @@ def print_message(consumer, doctype, doc,
                             if USE_AGENT_iterate:
 
                                 # print(f"\ntelling agent {agent_data}")
-                                agent = build_agen(peak_target=peak_target)
+                                agent = build_agen(peak_target=peak_target, agent_data_path=agent_data_path)
 
                                 if len(agent.table) < 2:
                                     acq_func = "qr"
@@ -558,6 +559,32 @@ def print_message(consumer, doctype, doc,
 
                                 new_points = agent.ask(acq_func, n=1)
 
+                                ## Get target of agent.ask()
+                                agent_target = agent.objectives.summary['target'].tolist()
+
+                                ## Get mean and standard deviation of agent.ask()
+                                res_values = []
+                                for i in new_points_label:
+                                    if i in new_points['points'].keys():
+                                        res_values.append(new_points['points'][i])
+                                x_tensor = torch.tensor(res_value)
+                                post = agent.posterior(x_tensor)
+                                post_mean = post.mean.tolist()[0]
+                                post_stddev = post.stddev.tolist()[0]
+
+                                ## apply np.exp for log-transform objectives
+                                if_log = agent.objectives.summary['transform']
+                                for j in range(if_log.shape[0]):
+                                    if if_log[j] == 'log':
+                                        post_mean[j] = np.exp(post_mean[j])
+                                        post_stddev[j] = np.exp(post_stddev[j])
+
+                                ## Update target, mean, and standard deviation in agent_data
+                                agent_data.update({'agent_target': agent_target})
+                                agent_data.update({'posterior_mean': post_mean})
+                                agent_data.update({'posterior_stddev': post_stddev})
+
+
                                 # peak_diff = peak_emission - peak_target
                                 peak_diff = False
 

scripts/prepare_agent_pdf.py

@@ -127,14 +127,45 @@ def build_agen(peak_target=660, peak_tolerance=5, size_target=6, agent_data_path
 
 
 '''
- agent.posterior
- import torch
- 
- res = agent.ask('qem', n=1)
- agent.posterior(torch.tensor(res['points'])).mean
-
- x = torch.tensor([[ 24.0426776 , 159.30614932, 101.20516362]])
- agent.posterior(x)
- agent.posterior(x).mean
- agent.plot_acquisition(); plt.show()
+agent.posterior
+import torch
+
+res = agent.ask('qem', n=1)
+agent.posterior(torch.tensor(res['points'])).mean
+
+x = torch.tensor([[ 24.0426776 , 159.30614932, 101.20516362]])
+agent.posterior(x)
+agent.posterior(x).mean
+agent.plot_acquisition(); plt.show()
+
+
+18/2: agent
+18/3: agent.table
+18/4: agent.table.Peak
+18/5: plt.rcParams['font.size'] = 4
+18/6: import matplotlib.pyplot as plt
+18/7: plt.rcParams['font.size'] = 4
+18/8: agent.plot_objectives(); plt.show()
+18/9: agent.objectives
+18/10: agent.ask("qem", n=1)
+18/11: agent.ask("qei", n=1)
+18/12: agent.ask("qei", n=1)
+18/13: import torch
+18/14: x = torch.tensor(res[0])
+18/15: res = agent.ask("qem", n=4)
+18/16: x = torch.tensor(res[0])
+18/17: agent.posterior(x).mean
+18/18: agent.best
+18/19: agent.objectives
+18/20: post = agent.posterior(x)
+18/21: post.mean
+18/22: post.sigma
+18/23: post.stddev
+18/24: agent.objectives
+18/25: agent.plot_acquisition(); plt.show()
+18/26: agent.plot_constraint(); plt.show()
+18/27: agent.dofs
+18/28: agent.objectives
+18/29: agent.best
+
  '''