Initial commit

skills/deep-learning-optimizer/assets/README.md

@@ -0,0 +1,7 @@
+# Assets
+
+Bundled resources for deep-learning-optimizer skill
+
+- [ ] optimization_config.json: Template for configuring optimization parameters.
+- [ ] example_models/: Directory containing example deep learning models for testing and demonstration.
+- [ ] visualization_templates/: Directory containing templates for visualizing model performance and optimization results.

skills/deep-learning-optimizer/assets/optimization_config.json

@@ -0,0 +1,47 @@
+{
+  "_comment": "Optimization configuration template for deep learning models.",
+  "optimizer_name": "Adam",
+  "_comment": "Name of the optimization algorithm to use. Options: Adam, SGD, RMSprop, AdamW, etc.",
+  "learning_rate": 0.001,
+  "_comment": "Learning rate for the optimizer. A smaller value might be needed for complex models.",
+  "weight_decay": 0.0001,
+  "_comment": "L2 regularization strength. Helps prevent overfitting.",
+  "beta1": 0.9,
+  "_comment": "Beta1 parameter for Adam optimizer (exponential decay rate for the 1st moment estimates).",
+  "beta2": 0.999,
+  "_comment": "Beta2 parameter for Adam optimizer (exponential decay rate for the 2nd moment estimates).",
+  "epsilon": 1e-08,
+  "_comment": "Epsilon parameter for Adam optimizer (term added to the denominator to improve numerical stability).",
+  "momentum": 0.0,
+  "_comment": "Momentum factor for SGD optimizer. Typically a value between 0 and 1.",
+  "nesterov": false,
+  "_comment": "Whether to use Nesterov momentum for SGD optimizer.",
+  "learning_rate_scheduler": {
+    "enabled": true,
+    "_comment": "Enable or disable learning rate scheduling.",
+    "scheduler_type": "ReduceLROnPlateau",
+    "_comment": "Type of learning rate scheduler. Options: StepLR, MultiStepLR, ExponentialLR, ReduceLROnPlateau, CosineAnnealingLR, CyclicLR, etc.",
+    "factor": 0.1,
+    "_comment": "Factor by which the learning rate will be reduced.",
+    "patience": 10,
+    "_comment": "Number of epochs with no improvement after which learning rate will be reduced.",
+    "threshold": 0.0001,
+    "_comment": "Threshold for measuring the new optimum, to only focus on significant changes.",
+    "threshold_mode": "rel",
+    "_comment": "One of rel, abs. In rel mode, dynamic_threshold = best * ( 1 + threshold ) in 'max' mode or best * ( 1 - threshold ) in min mode. In abs mode, dynamic_threshold = best + threshold in max mode or best - threshold in min mode.",
+    "cooldown": 0,
+    "_comment": "Number of epochs to wait before resuming normal operation after lr has been reduced.",
+    "min_lr": 0,
+    "_comment": "A scalar or a list of scalars. A lower bound on the learning rate of all param groups or each group respectively.",
+    "verbose": true
+     "_comment": "If True, prints a message to stdout for each update."
+  },
+  "gradient_clipping": {
+    "enabled": true,
+    "_comment": "Enable or disable gradient clipping.",
+    "clip_value": 1.0,
+    "_comment": "The clipping threshold. Gradients will be clipped to this value.",
+    "clip_norm_type": 2.0,
+    "_comment": "The type of the norm used for clipping. Can be 2.0 (L2 norm), inf (infinity norm), etc."
+  }
+}