Change model module names.giti

examples/gaussian_diagcov.py

@@ -164,7 +164,7 @@ def run_example(
     # ======================================================================
     # Train hyper-spherical model
     # ======================================================================
-    model = hm.model.HyperSphere(ndim, domains_sphere)
+    model = hm.model_legacy.HyperSphere(ndim, domains_sphere)
     model.set_R(max_r_prob)
     model.fitted = True
 

examples/gaussian_nondiagcov.py

@@ -146,7 +146,7 @@ def run_example(
         hm.logs.debug_log("r scale = {}".format(r_scale))
         domains = [r_scale * np.array([1e-1, 1e0])]
         hm.logs.debug_log("Domain = {}".format(domains))
-        model = hm.model.HyperSphere(ndim, domains)
+        model = hm.model_legacy.HyperSphere(ndim, domains)
         fit_success, objective = model.fit(
             chains_train.samples, chains_train.ln_posterior
         )

examples/gaussian_nondiagcov_nvp.py

@@ -13,7 +13,7 @@
 import utils
 
 sys.path.append("harmonic")
-import model_nf
+import model as md
 
 
 def ln_analytic_evidence(ndim, cov):
@@ -144,9 +144,7 @@ def run_example(
         # Fit model
         # =======================================================================
         hm.logs.info_log("Fit model for {} epochs...".format(epochs_num))
-        model = model_nf.RealNVPModel(
-            ndim, standardize=standardize, temperature=temperature
-        )
+        model = md.RealNVPModel(ndim, standardize=standardize, temperature=temperature)
         model.fit(chains_train.samples, epochs=epochs_num, verbose=verbose)
 
         # Use chains and model to compute inverse evidence.

examples/gaussian_nondiagcov_splines.py

@@ -13,7 +13,7 @@
 import utils
 
 sys.path.append("harmonic")
-from harmonic import model_nf
+from harmonic import model as md
 import jax
 import jax.numpy as jnp
 
@@ -150,7 +150,7 @@ def run_example(ndim=2, nchains=100, samples_per_chain=1000, plot_corner=False):
         # Fit model
         # =======================================================================
         hm.logs.info_log("Fit model for {} epochs...".format(epochs_num))
-        model = model_nf.RQSplineModel(
+        model = md.RQSplineModel(
             ndim,
             n_layers=n_layers,
             n_bins=n_bins,

examples/normal_gamma.py

@@ -279,7 +279,7 @@ def run_example(
             domains_MGMM,
             hyper_parameters_MGMM,
             nfold=nfold,
-            modelClass=hm.model.ModifiedGaussianMixtureModel,
+            modelClass=hm.model_legacy.ModifiedGaussianMixtureModel,
             seed=0,
         )
         hm.logs.debug_log(
@@ -293,7 +293,7 @@ def run_example(
             domains_sphere,
             hyper_parameters_sphere,
             nfold=nfold,
-            modelClass=hm.model.HyperSphere,
+            modelClass=hm.model_legacy.HyperSphere,
             seed=0,
         )
         hm.logs.debug_log(
@@ -316,12 +316,12 @@ def run_example(
             hm.logs.debug_log(
                 "Using MGMM with hyper_parameters = {}".format(best_hyper_param_MGMM)
             )
-            model = hm.model.ModifiedGaussianMixtureModel(
+            model = hm.model_legacy.ModifiedGaussianMixtureModel(
                 ndim, domains_MGMM, hyper_parameters=best_hyper_param_MGMM
             )
         else:
             hm.logs.debug_log("Using HyperSphere")
-            model = hm.model.HyperSphere(
+            model = hm.model_legacy.HyperSphere(
                 ndim, domains_sphere, hyper_parameters=best_hyper_param_sphere
             )
         fit_success = model.fit(chains_train.samples, chains_train.ln_posterior)

examples/normal_gamma_nvp.py

@@ -13,7 +13,7 @@
 import utils
 
 sys.path.append("harmonic")
-import model_nf
+import model as md
 
 
 def ln_likelihood(x_mean, x_std, x_n, mu, tau):
@@ -268,9 +268,7 @@ def run_example(
         # Fit model
         # =======================================================================
         hm.logs.info_log("Fit model for {} epochs...".format(epochs_num))
-        model = model_nf.RealNVPModel(
-            ndim, standardize=standardize, temperature=temperature
-        )
+        model = md.RealNVPModel(ndim, standardize=standardize, temperature=temperature)
         model.fit(chains_train.samples, epochs=epochs_num)
 
         # ===================================================================

examples/normal_gamma_splines.py

@@ -13,7 +13,7 @@
 import utils
 
 sys.path.append("harmonic")
-import model_nf
+import model as md
 
 
 def ln_likelihood(x_mean, x_std, x_n, mu, tau):
@@ -268,9 +268,7 @@ def run_example(
         # Fit model
         # =======================================================================
         hm.logs.info_log("Fit model for {} epochs...".format(epochs_num))
-        model = model_nf.RQSplineModel(
-            ndim, standardize=standardize, temperature=temperature
-        )
+        model = md.RQSplineModel(ndim, standardize=standardize, temperature=temperature)
         model.fit(chains_train.samples, epochs=epochs_num)
 
         # ===================================================================

examples/pima_indian.py

@@ -260,7 +260,7 @@ def run_example(
         domains_MGMM,
         hyper_parameters_MGMM,
         nfold=nfold,
-        modelClass=hm.model.ModifiedGaussianMixtureModel,
+        modelClass=hm.model_legacy.ModifiedGaussianMixtureModel,
         seed=0,
     )
     hm.logs.debug_log(
@@ -275,7 +275,7 @@ def run_example(
         domains_sphere,
         hyper_parameters_sphere,
         nfold=nfold,
-        modelClass=hm.model.HyperSphere,
+        modelClass=hm.model_legacy.HyperSphere,
         seed=0,
     )
     hm.logs.debug_log(
@@ -298,12 +298,12 @@ def run_example(
         hm.logs.info_log(
             "Using MGMM with hyper_parameters = {}".format(best_hyper_param_MGMM)
         )
-        model = hm.model.ModifiedGaussianMixtureModel(
+        model = hm.model_legacy.ModifiedGaussianMixtureModel(
             ndim, domains_MGMM, hyper_parameters=best_hyper_param_MGMM
         )
     else:
         hm.logs.info_log("Using HyperSphere")
-        model = hm.model.HyperSphere(
+        model = hm.model_legacy.HyperSphere(
             ndim, domains_sphere, hyper_parameters=best_hyper_param_sphere
         )
 

examples/pima_indian_nvp.py

@@ -13,7 +13,7 @@
 import utils
 
 sys.path.append("harmonic")
-import model_nf
+import model as md
 import flows
 
 
@@ -255,7 +255,7 @@ def run_example(
     minimises the validation variances.
     """
 
-    model = model_nf.RealNVPModel(
+    model = md.RealNVPModel(
         ndim,
         n_scaled_layers=n_scaled,
         n_unscaled_layers=n_unscaled,

examples/radiata_pine.py

@@ -440,7 +440,7 @@ def run_example(
     """
 
     hm.logs.info_log("Using HyperSphere")
-    model = hm.model.HyperSphere(ndim, domains_sphere, hyper_parameters=None)
+    model = hm.model_legacy.HyperSphere(ndim, domains_sphere, hyper_parameters=None)
 
     fit_success = model.fit(chains_train.samples, chains_train.ln_posterior)
     hm.logs.debug_log("fit_success = {}".format(fit_success))

examples/radiata_pine_nvp.py

@@ -13,7 +13,7 @@
 import utils
 
 sys.path.append("harmonic")
-import model_nf
+import model as md
 import flows
 
 
@@ -436,15 +436,15 @@ def run_example(
     Fit model by selecing the configuration of hyper-parameters which 
     minimises the validation variances.
     """
-    model = model_nf.RealNVPModel(
+    model = md.RealNVPModel(
         ndim,
         n_scaled_layers=n_scaled,
         n_unscaled_layers=n_unscaled,
         learning_rate=learning_rate,
         standardize=standardize,
         temperature=temperature,
     )
-    # model = model_nf.RQSplineFlow(ndim)
+    # model = md.RQSplineFlow(ndim)
     model.fit(chains_train.samples, epochs=epochs_num)
 
     # ===========================================================================

examples/radiata_pine_splines.py

@@ -13,7 +13,7 @@
 import utils
 
 sys.path.append("harmonic")
-import model_nf
+import model as md
 import flows
 
 
@@ -433,9 +433,7 @@ def run_example(
     Fit model by selecing the configuration of hyper-parameters which 
     minimises the validation variances.
     """
-    model = model_nf.RQSplineModel(
-        ndim, standardize=standardize, temperature=temperature
-    )
+    model = md.RQSplineModel(ndim, standardize=standardize, temperature=temperature)
     model.fit(chains_train.samples, epochs=epochs_num)
 
     # ===========================================================================

examples/rastrigin.py

@@ -198,7 +198,7 @@ def run_example(
             domain,
             hyper_parameters,
             nfold=nfold,
-            modelClass=hm.model.KernelDensityEstimate,
+            modelClass=hm.model_legacy.KernelDensityEstimate,
             seed=0,
         )
 
@@ -215,7 +215,7 @@ def run_example(
         Fit model by selecing the configuration of hyper-parameters which 
         minimises the validation variances.
         """
-        model = hm.model.KernelDensityEstimate(
+        model = hm.model_legacy.KernelDensityEstimate(
             ndim, domain, hyper_parameters=best_hyper_param
         )
         fit_success = model.fit(chains_train.samples, chains_train.ln_posterior)

examples/rastrigin_splines.py

@@ -12,7 +12,7 @@
 import utils
 
 sys.path.append("harmonic")
-import model_nf
+import model as md
 import flows
 
 
@@ -200,7 +200,7 @@ def run_example(
         """
         Fit model.
         """
-        model = model_nf.RQSplineModel(
+        model = md.RQSplineModel(
             ndim,
             n_layers=n_layers,
             n_bins=n_bins,

examples/rosenbrock.py

@@ -237,7 +237,7 @@ def run_example(
             domain,
             hyper_parameters,
             nfold=nfold,
-            modelClass=hm.model.KernelDensityEstimate,
+            modelClass=hm.model_legacy.KernelDensityEstimate,
             seed=0,
         )
         hm.logs.debug_log("validation_variances = {}".format(validation_variances))
@@ -253,7 +253,7 @@ def run_example(
         Fit model by selecing the configuration of hyper-parameters which 
         minimises the validation variances.
         """
-        model = hm.model.KernelDensityEstimate(
+        model = hm.model_legacy.KernelDensityEstimate(
             ndim, domain, hyper_parameters=best_hyper_param
         )
         fit_success = model.fit(chains_train.samples, chains_train.ln_posterior)

examples/rosenbrock_nvp.py

@@ -10,7 +10,7 @@
 
 sys.path.append("examples")
 import utils
-from harmonic import model_nf
+from harmonic import model as md
 
 
 def ln_prior_uniform(x, xmin=-10.0, xmax=10.0, ymin=-5.0, ymax=15.0):
@@ -220,9 +220,7 @@ def run_example(
         # Fit model
         # =======================================================================
         hm.logs.info_log("Fit model for {} epochs...".format(epochs_num))
-        model = model_nf.RealNVPModel(
-            ndim, standardize=standardize, temperature=temperature
-        )
+        model = md.RealNVPModel(ndim, standardize=standardize, temperature=temperature)
         model.fit(chains_train.samples, epochs=epochs_num)
 
         # =======================================================================

examples/rosenbrock_splines.py

@@ -10,7 +10,7 @@
 
 sys.path.append("examples")
 import utils
-from harmonic import model_nf
+from harmonic import model as md
 
 
 def ln_prior_uniform(x, xmin=-10.0, xmax=10.0, ymin=-5.0, ymax=15.0):
@@ -220,9 +220,7 @@ def run_example(
         # Fit model
         # =======================================================================
         hm.logs.info_log("Fit model for {} epochs...".format(epochs_num))
-        model = model_nf.RQSplineModel(
-            ndim, standardize=standardize, temperature=temperature
-        )
+        model = md.RQSplineModel(ndim, standardize=standardize, temperature=temperature)
         model.fit(chains_train.samples, epochs=epochs_num)
 
         # =======================================================================

examples/temperature_diagram.py

@@ -10,7 +10,7 @@
 tfp = tfp.substrates.jax
 tfd = tfp.distributions
 tfb = tfp.bijectors
-from harmonic import model_nf
+from harmonic import model as md
 from harmonic import flows
 
 
@@ -101,7 +101,7 @@ def get_batch(seed):
 epochs_num = 300
 
 posterior_samples = np.array(get_moon(0.05).sample(batch_size, seed=seed))
-model = model_nf.RealNVPModel(
+model = md.RealNVPModel(
     ndim,
     flow=flows.RealNVP(ndim, n_scaled_layers=n_scaled, n_unscaled_layers=n_unscaled),
 )

harmonic/__init__.py

@@ -1,6 +1,7 @@
 from .chains import Chains
 from . import evidence
 from .evidence import Evidence, Shifting
-from . import model
+
+# from . import model
 from . import utils
 from . import logs

harmonic/model_nf.py → harmonic/model.py

@@ -1,5 +1,5 @@
 from typing import Sequence, Union
-from harmonic import model as md
+from harmonic import model_abstract as mda
 from harmonic import flows
 import jax
 import jax.numpy as jnp
@@ -94,7 +94,7 @@ def train_flow(
     return train_flow, train_epoch, train_step
 
 
-class FlowModel(md.Model):
+class FlowModel(mda.Model):
     """Normalizing flow model to approximate the log_e posterior by a normalizing flow."""
 
     def __init__(