Initial commit

skills/hyperparameter-tuner/assets/hyperparameter_space_template.json

@@ -0,0 +1,81 @@
+{
+  "_comment": "Template for defining the hyperparameter search space.  This file should be used as a guide for creating your own hyperparameter configuration.",
+  "algorithm": {
+    "_comment": "The search algorithm to use.  Options: 'grid', 'random', 'bayesian'.",
+    "type": "string",
+    "default": "random",
+    "enum": ["grid", "random", "bayesian"]
+  },
+  "objective": {
+    "_comment": "The metric to optimize.  The plugin will attempt to maximize this metric.",
+    "type": "string",
+    "default": "val_loss"
+  },
+  "max_trials": {
+    "_comment": "The maximum number of trials to run.  Each trial will explore a different set of hyperparameters.",
+    "type": "integer",
+    "default": 10
+  },
+  "hyperparameters": {
+    "_comment": "A dictionary of hyperparameters to search.  Each key is the name of the hyperparameter, and the value is a dictionary defining the search space for that hyperparameter.",
+    "type": "object",
+    "properties": {
+      "learning_rate": {
+        "_comment": "Example: Learning rate for a neural network.",
+        "type": "number",
+        "distribution": "loguniform",
+        "min": 0.0001,
+        "max": 0.1
+      },
+      "num_layers": {
+        "_comment": "Example: Number of layers in a neural network.",
+        "type": "integer",
+        "distribution": "uniform",
+        "min": 2,
+        "max": 6
+      },
+      "dropout_rate": {
+        "_comment": "Example: Dropout rate for regularization.",
+        "type": "number",
+        "distribution": "uniform",
+        "min": 0.0,
+        "max": 0.5
+      },
+      "batch_size": {
+        "_comment": "Example: Batch size for training.",
+        "type": "integer",
+        "distribution": "categorical",
+        "values": [32, 64, 128, 256]
+      },
+       "optimizer": {
+        "_comment": "Example: Optimization algorithm to use",
+        "type": "string",
+        "distribution": "categorical",
+        "values": ["adam", "sgd", "rmsprop"]
+      }
+    },
+    "required": ["learning_rate", "num_layers"]
+  },
+  "early_stopping": {
+    "_comment": "Parameters for early stopping. If enabled, the tuning process will stop if the objective metric does not improve for a specified number of epochs.",
+    "type": "object",
+    "properties": {
+      "monitor": {
+        "_comment": "The metric to monitor for early stopping.",
+        "type": "string",
+        "default": "val_loss"
+      },
+      "patience": {
+        "_comment": "The number of epochs with no improvement after which training will be stopped.",
+        "type": "integer",
+        "default": 3
+      },
+      "enabled": {
+        "_comment": "Whether early stopping is enabled.",
+        "type": "boolean",
+        "default": true
+      }
+    },
+    "required": ["monitor", "patience", "enabled"]
+  }
+}