--- a
+++ b/PhenPred/vae/Hypers.py
@@ -0,0 +1,75 @@
+import json
+from PhenPred.vae import data_folder, plot_folder
+from PhenPred.vae.Losses import CLinesLosses
+
+
+class Hypers:
+    @classmethod
+    def read_json(cls, json_file):
+        with open(json_file, "r") as f:
+            hypers = json.load(f)
+        return hypers
+
+    @classmethod
+    def read_hyperparameters(
+        cls, hypers_json=None, parse_torch_functions=True, timestamp=None
+    ):
+        if timestamp is not None:
+            hypers_json = f"{plot_folder}/files/{timestamp}_hyperparameters.json"
+        elif hypers_json is None:
+            hypers_json = f"{plot_folder}/files/hyperparameters.json"
+
+        hypers = cls.read_json(hypers_json)
+
+        if timestamp is not None:
+            hypers["load_run"] = timestamp
+
+        if "model" not in hypers:
+            hypers["model"] = "MOSA"
+
+        if "standardize" not in hypers:
+            hypers["standardize"] = False
+
+        if "w_rec" not in hypers:
+            hypers["w_rec"] = 1
+
+        if "w_gauss" not in hypers:
+            hypers["w_gauss"] = 0.01
+
+        if "w_cat" not in hypers:
+            hypers["w_cat"] = 0.01
+
+        if hypers["view_loss_weights"] is None:
+            hypers["view_loss_weights"] = [1.0] * len(hypers["views"])
+
+        if hypers["use_conditionals"] is None:
+            hypers["use_conditionals"] = True
+
+        if timestamp is None:  # full path is already stored in previous json config
+            hypers["datasets"] = {
+                k: f"{data_folder}/{v}" for k, v in hypers["datasets"].items()
+            }
+        print(f"# ---- Hyperparameters")
+        print(json.dumps(hypers, indent=4, sort_keys=True))
+
+        if parse_torch_functions:
+            hypers = cls.parse_torch_functions(hypers)
+
+        return hypers
+
+    @classmethod
+    def parse_torch_functions(cls, hypers):
+        if type(hypers["activation_function"]) == str:
+            hypers["activation_function"] = CLinesLosses.activation_function(
+                hypers["activation_function"]
+            )
+
+        if type(hypers["reconstruction_loss"]) == str:
+            hypers["reconstruction_loss"] = CLinesLosses.reconstruction_loss_method(
+                hypers["reconstruction_loss"]
+            )
+
+        if type(hypers["hidden_dims"]) == str:
+            hypers["hidden_dims"] = [float(l) for l in hypers["hidden_dims"].split(",")]
+
+        return hypers