adding dropouts to the mlp discrepancy and providing functions to int…

…errogate the lf model plus discrepancy

linfa/eval_model_from_chkpt.py

@@ -0,0 +1,61 @@
+import os,torch
+from linfa.discrepancy import Discrepancy
+from linfa.maf import MAF, RealNVP
+
+def eval_discrepancy(file_path,test_data):
+
+ # Read in data
+ exp_name = os.path.basename(file_path)
+ dir_name = os.path.dirname(file_path)
+
+ # Create new discrepancy
+ dicr = Discrepancy(model_name = exp_name, 
+ model_folder = dir_name,
+ lf_model = None,
+ input_size = None,
+ output_size = None,
+ var_grid_in = None,
+ var_grid_out = None)
+ dicr.surrogate_load()
+
+ # Evaluate discrepancy over test grid
+ return dicr.forward(test_data) 
+
+
+def eval_model(exp,discr_chkpt_file,nf_chkpt_file,test_data):
+
+ # Evaluate discrepancy at tp data "test_data"
+ res_discr = eval_discrepancy(discr_chkpt_file,test_data)
+ # Sample from normalzing flow
+
+ # Create NF model from experiment
+ if exp.flow_type == 'maf':
+ nf = MAF(exp.n_blocks, exp.input_size, exp.hidden_size, exp.n_hidden, None,
+ exp.activation_fn, exp.input_order, batch_norm=exp.batch_norm_order)
+ elif exp.flow_type == 'realnvp': # Under construction
+ nf = RealNVP(exp.n_blocks, exp.input_size, exp.hidden_size, exp.n_hidden, None,
+ batch_norm=exp.batch_norm_order)
+
+ # Read state dictionary 
+ nf.state_dict(torch.load(nf_chkpt_file)) 
+
+ # Sample calibration parameter realizations
+ x00 = nf.base_dist.sample([num_calib_samples])
+ xkk, _ = nf(x00)
+
+ # Solve models 
+ res_lf = exp.model.solve_t(exp.transform.forward(xkk))
+
+ # return 
+ return res_lf + res_discr
+
+
+
+
+
+
+ res_model = eval_discrepancy(file_path,test_data_discr)
+
+
+
+