classy-szfast 0.0.14__py3-none-any.whl → 0.0.16__py3-none-any.whl

classy_szfast/restore_nn.py

@@ -87,26 +87,82 @@ class Restore_NN(tf.keras.Model):
             print(multiline_str)
 
 
-    # restore attributes
-    def restore(self, 
-                filename
-                ):
+
+    # from https://github.com/HTJense/cosmopower/blob/packaging-paper/cosmopower/cosmopower_NN.py
+    def restore(self, filename: str, allow_pickle: bool = False) -> None:
         r"""
-        Load pre-trained model
+        Load pre-trained model.
+        The default file format is compressed numpy files (.npz). The
+        Module will attempt to use this as a file extension and restore
+        from there (i.e. look for `filename.npz`). If this file does
+        not exist, and `allow_pickle` is set to True, then the file
+        `filename.pkl` will be attempted to be read by `restore_pickle`.
+
+        The function will trim the file extension from `filename`, so
+        `restore("filename")` and `restore("filename.npz")` are identical.
 
         Parameters:
-            filename (str):
-                filename tag (without suffix) where model was saved
+        :param filename: filename (without suffix) where model was saved.
+        :param allow_pickle: whether or not to permit passing this filename
+                             to the `restore_pickle` function.
         """
-        # load attributes
-        with open(filename + ".pkl", 'rb') as f:
-            self.W_, self.b_, self.alphas_, self.betas_, \
-            self.parameters_mean_, self.parameters_std_, \
-            self.features_mean_, self.features_std_, \
-            self.n_parameters, self.parameters, \
-            self.n_modes, self.modes, \
-            self.n_hidden, self.n_layers, self.architecture = pickle.load(f)
-
+        # Check if npz file exists.
+        filename_npz = filename + ".npz"
+        if not os.path.exists(filename_npz):
+            # Can we load this file as a pickle file?
+            filename_pkl = filename + ".pkl"
+            if allow_pickle and os.path.exists(filename_pkl):
+                self.restore_pickle(filename_pkl)
+                return
+
+            raise IOError(f"Failed to restore network from {filename}: "
+                          + (" is a pickle file, try setting 'allow_pickle = \
+                              True'" if os.path.exists(filename_pkl) else
+                             " does not exist."))
+
+        with open(filename_npz, "rb") as fp:
+            fpz = np.load(fp, allow_pickle=True)["arr_0"].flatten()[0]
+
+            # print('filename_npz:', filename_npz)
+            # print("Keys in the .npz file:", list(fpz.keys()))
+
+            self.architecture = fpz["architecture"]
+            self.n_layers = fpz["n_layers"]
+            self.n_hidden = fpz["n_hidden"]
+            self.n_parameters = fpz["n_parameters"]
+            self.n_modes = fpz["n_modes"]
+
+            self.parameters = list(fpz["parameters"])
+            self.modes = fpz["modes"]
+
+            # self.parameters_mean_ = fpz["parameters_mean"]
+            # self.parameters_std_ = fpz["parameters_std"]
+            # self.features_mean_ = fpz["features_mean"]
+            # self.features_std_ = fpz["features_std"]
+
+            # Attempt to load 'parameters_mean' or fall back to 'param_train_mean'
+            self.parameters_mean_ = fpz.get("parameters_mean", fpz.get("param_train_mean"))
+            self.parameters_std_ = fpz.get("parameters_std", fpz.get("param_train_std"))
+            self.features_mean_ = fpz.get("features_mean", fpz.get("feature_train_mean"))
+            self.features_std_ = fpz.get("features_std", fpz.get("feature_train_std"))
+
+
+            # Fallback to 'weights_' if individual 'W_i' are not found
+            if "weights_" in fpz:
+                # Assign the list of weight arrays from 'weights_' directly
+                self.W_ = fpz["weights_"]
+            else:
+                # Use individual weight arrays if available
+                self.W_ = [fpz[f"W_{i}"] for i in range(self.n_layers)]
+
+            # Fallback to 'biases_' if individual 'b_i' are not found
+            if "biases_" in fpz:
+                self.b_ = fpz["biases_"]
+            else:
+                self.b_ = [fpz[f"b_{i}"] for i in range(self.n_layers)]
+
+            self.alphas_ = fpz[f"alphas_"]
+            self.betas_ = fpz[f"betas_"]
 
 
     # auxiliary function to sort input parameters
@@ -274,29 +330,110 @@ class Restore_PCAplusNN(tf.keras.Model):
 
 
 
-    # restore attributes
-    def restore(self, 
-                filename,
-                ):
+    # from https://github.com/HTJense/cosmopower/blob/packaging-paper/cosmopower/cosmopower_PCAplusNN.py
+    def restore(self, filename: str, allow_pickle: bool = False) -> None:
         r"""
-        Load pre-trained model
+        Load pre-trained model.
+        The default file format is compressed numpy files (.npz). The
+        Module will attempt to use this as a file extension and restore
+        from there (i.e. look for `filename.npz`). If this file does
+        not exist, and `allow_pickle` is set to True, then the file
+        `filename.pkl` will be attempted to be read by `restore_pickle`.
+
+        The function will trim the file extension from `filename`, so
+        `restore("filename")` and `restore("filename.npz")` are identical.
 
         Parameters:
-            filename (str):
-                filename tag (without suffix) where model was saved
+        :param filename: filename (without suffix) where model was saved.
+        :param allow_pickle: whether or not to permit passing this filename to
+                             the `restore_pickle` function.
         """
-        # load attributes
-        f = open(filename + ".pkl", 'rb')
-        self.W_, self.b_, self.alphas_, self.betas_, \
-        self.parameters_mean_, self.parameters_std_, \
-        self.pca_mean_, self.pca_std_, \
-        self.features_mean_, self.features_std_, \
-        self.parameters, self.n_parameters, \
-        self.modes, self.n_modes, \
-        self.n_pcas, self.pca_transform_matrix_, \
-        self.n_hidden, self.n_layers, self.architecture = pickle.load(f)
-        f.close()
-
+        # Check if npz file exists.
+        filename_npz = filename + ".npz"
+        if not os.path.exists(filename_npz):
+            # Can we load this file as a pickle file?
+            filename_pkl = filename + ".pkl"
+            if allow_pickle and os.path.exists(filename_pkl):
+                self.restore_pickle(filename_pkl)
+                return
+
+            raise IOError(f"Failed to restore network from {filename}: "
+                          + (" is a pickle file, try setting 'allow_pickle = \
+                              True'" if os.path.exists(filename_pkl) else
+                             " does not exist."))
+
+        with open(filename_npz, "rb") as fp:
+            fpz = np.load(fp, allow_pickle=True)["arr_0"].flatten()[0]
+
+            # print('filename_npz:', filename_npz)
+            # print("Keys in the .npz file:", list(fpz.keys()))
+
+
+            self.architecture = fpz["architecture"]
+            self.n_layers = fpz["n_layers"]
+            self.n_hidden = fpz["n_hidden"]
+            self.n_parameters = fpz["n_parameters"]
+            self.n_modes = fpz["n_modes"]
+
+            self.parameters = fpz["parameters"]
+            self.modes = fpz["modes"]
+
+            # Attempt to load 'parameters_mean' or fall back to 'param_train_mean'
+            self.parameters_mean_ = fpz.get("parameters_mean", fpz.get("param_train_mean"))
+            self.parameters_std_ = fpz.get("parameters_std", fpz.get("param_train_std"))
+            self.features_mean_ = fpz.get("features_mean", fpz.get("feature_train_mean"))
+            self.features_std_ = fpz.get("features_std", fpz.get("feature_train_std"))
+
+            # Handle PCA-related keys
+            # self.pca_mean_ = fpz["pca_mean"]
+            # self.pca_std_ = fpz["pca_std"]
+            # self.n_pcas = fpz["n_pcas"]
+
+            self.pca_mean_ = fpz["pca_mean"]
+            self.pca_std_ = fpz["pca_std"]
+            self.n_pcas = fpz["n_pcas"]
+            self.pca_transform_matrix_ = fpz["pca_transform_matrix"]
+
+            # print('n_pcas:', self.n_pcas)
+
+            # filename  = "/Users/boris/Work/CLASS-SZ/SO-SZ/cosmopower-organization/lcdm/TTTEEE/TE_v1.pkl"
+            # f = open(filename, 'rb')
+            # self.W_, self.b_, self.alphas_, self.betas_, \
+            # self.parameters_mean_, self.parameters_std_, \
+            # self.pca_mean_, self.pca_std_, \
+            # self.features_mean_, self.features_std_, \
+            # self.parameters, self.n_parameters, \
+            # self.modes, self.n_modes, \
+            # self.n_pcas, self.pca_transform_matrix_, \
+            # self.n_hidden, self.n_layers, self.architecture = pickle.load(f)
+
+            # print('self.n_pcas:', self.n_pcas)
+            # print('self.pca_transform_matrix_:', self.pca_transform_matrix_)
+            # print('PCA mean:', self.pca_mean_)
+            # print('PCA std:', self.pca_std_)
+            # f.close()
+            # import sys
+            # sys.exit(0)
+
+            # self.pca_transform_matrix_ = fpz["pca_transform_matrix"]        
+
+            # Fallback to 'weights_' if individual 'W_i' are not found
+            if "weights_" in fpz:
+                # Assign the list of weight arrays from 'weights_' directly
+                self.W_ = fpz["weights_"]
+            else:
+                # Use individual weight arrays if available
+                self.W_ = [fpz[f"W_{i}"] for i in range(self.n_layers)]
+
+            # Fallback to 'biases_' if individual 'b_i' are not found
+            if "biases_" in fpz:
+                self.b_ = fpz["biases_"]
+            else:
+                self.b_ = [fpz[f"b_{i}"] for i in range(self.n_layers)]
+
+            # Handle alphas and betas
+            self.alphas_ = fpz.get("alphas_", [fpz.get(f"alphas_{i}") for i in range(self.n_layers - 1)])
+            self.betas_ = fpz.get("betas_", [fpz.get(f"betas_{i}") for i in range(self.n_layers - 1)])
 
 
 

{classy_szfast-0.0.14.dist-info → classy_szfast-0.0.16.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: classy_szfast
-Version: 0.0.14
+Version: 0.0.16
 Summary: The accelerator of the class_sz code from https://github.com/CLASS-SZ
 Maintainer-email: Boris Bolliet <bb667@cam.ac.uk>
 Project-URL: Homepage, https://github.com/CLASS-SZ

{classy_szfast-0.0.14.dist-info → classy_szfast-0.0.16.dist-info}/RECORD

@@ -5,14 +5,14 @@ classy_szfast/config.py,sha256=4CvejtLcFOQR30bJ8tlEeBHhu3Rr7LakeLO6dbFgPSU,210
 classy_szfast/cosmopower.py,sha256=eym72TFAcSJSTUlrwD-sAg8_9e2GdZq0m3lLPQ7uvPU,9858
 classy_szfast/cosmosis_classy_szfast_interface.py,sha256=zAnxvFtn73a5yS7jgs59zpWFEYKCIQyraYPs5hQ4Le8,11483
 classy_szfast/pks_and_sigmas.py,sha256=drtuujE1HhlrYY1hY92DyY5lXlYS1uE15MSuVI4uo6k,6625
-classy_szfast/restore_nn.py,sha256=OyxaRRk9D4hOJTvUSY3c5wAWTPCZJRMxBtin4kq_xd0,14149
+classy_szfast/restore_nn.py,sha256=xjwhGvd1W7fVhKvaeZNaQ9KOA4CyKuC2ornBzZmkxEo,21088
 classy_szfast/suppress_warnings.py,sha256=6wIBml2Sj9DyRGZlZWhuA9hqvpxqrNyYjuz6BPK_a6E,202
 classy_szfast/utils.py,sha256=VdaRsJK2ttHI9zkyxVhergxHPC6t99usrlycblyqcP8,1464
 classy_szfast/custom_bias/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 classy_szfast/custom_bias/custom_bias.py,sha256=aR2t5RTIwv7P0m2bsEU0Eq6BTkj4pG10AebH6QpG4qM,486
 classy_szfast/custom_profiles/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 classy_szfast/custom_profiles/custom_profiles.py,sha256=4LZwb2XoqwCyWNmW2s24Z7AJdmgVdaRG7yYaBYe-d9Q,1188
-classy_szfast-0.0.14.dist-info/METADATA,sha256=Y_uk64BUO3RTDgeBFbhe2m7ALSfkC5uAyNrhLbzPY80,472
-classy_szfast-0.0.14.dist-info/WHEEL,sha256=Wyh-_nZ0DJYolHNn1_hMa4lM7uDedD_RGVwbmTjyItk,91
-classy_szfast-0.0.14.dist-info/top_level.txt,sha256=hRgqpilUck4lx2KkaWI2y9aCDKqF6pFfGHfNaoPFxv0,14
-classy_szfast-0.0.14.dist-info/RECORD,,
+classy_szfast-0.0.16.dist-info/METADATA,sha256=a5hUhLGckjGstj0NwojhKp5Ld_L-5Dlf4kLDYD4a4og,472
+classy_szfast-0.0.16.dist-info/WHEEL,sha256=Wyh-_nZ0DJYolHNn1_hMa4lM7uDedD_RGVwbmTjyItk,91
+classy_szfast-0.0.16.dist-info/top_level.txt,sha256=hRgqpilUck4lx2KkaWI2y9aCDKqF6pFfGHfNaoPFxv0,14
+classy_szfast-0.0.16.dist-info/RECORD,,