@reicek/neataptic-ts 0.1.25 → 0.1.26

package/src/architecture/network/runtime/README.md

@@ -10,216 +10,216 @@ schedule policies callers can tune directly.
 
 ## architecture/network/runtime/network.runtime.controls.utils.ts
 
-### configurePruning
+### clearStochasticDepthSchedule
 
 ```ts
-configurePruning(
-  configuration: PruningConfiguration,
-): void
+clearStochasticDepthSchedule(): void
 ```
 
-Configure scheduled pruning during training.
-
-This stores the pruning window and target policy on the network so the
-training loop can opportunistically apply structured sparsification later.
+Clear the stochastic-depth schedule function.
 
 Parameters:
 - `this` - Target network instance.
-- `configuration` - Pruning schedule and ranking configuration.
 
 Returns: Nothing.
 
-### enableWeightNoise
+### clearWeightNoiseSchedule
 
 ```ts
-enableWeightNoise(
-  configuration: WeightNoiseConfiguration,
-): void
+clearWeightNoiseSchedule(): void
 ```
 
-Enable weight noise using either one global standard deviation or per-hidden-layer values.
-
-A single global value is useful for quick experiments, while the per-hidden
-schedule keeps layered models explicit about which hidden stage receives how
-much perturbation.
+Clear the dynamic global weight-noise schedule.
 
 Parameters:
 - `this` - Target network instance.
-- `configuration` - Global standard deviation or one value per hidden layer.
 
 Returns: Nothing.
 
-### disableWeightNoise
+### configurePruning
 
 ```ts
-disableWeightNoise(): void
+configurePruning(
+  configuration: PruningConfiguration,
+): void
 ```
 
-Disable all configured weight noise.
+Configure scheduled pruning during training.
+
+This stores the pruning window and target policy on the network so the
+training loop can opportunistically apply structured sparsification later.
 
 Parameters:
 - `this` - Target network instance.
+- `configuration` - Pruning schedule and ranking configuration.
 
 Returns: Nothing.
 
-### setWeightNoiseSchedule
+### disableStochasticDepth
 
 ```ts
-setWeightNoiseSchedule(
-  schedule: (step: number) => number,
-): void
+disableStochasticDepth(): void
 ```
 
-Set a dynamic scheduler for global weight noise.
+Disable stochastic depth.
 
 Parameters:
 - `this` - Target network instance.
-- `schedule` - Function mapping the current training step to a standard deviation.
 
 Returns: Nothing.
 
-### clearWeightNoiseSchedule
+### disableWeightNoise
 
 ```ts
-clearWeightNoiseSchedule(): void
+disableWeightNoise(): void
 ```
 
-Clear the dynamic global weight-noise schedule.
+Disable all configured weight noise.
 
 Parameters:
 - `this` - Target network instance.
 
 Returns: Nothing.
 
-### setRandom
+### enableWeightNoise
 
 ```ts
-setRandom(
-  randomFunction: () => number,
+enableWeightNoise(
+  configuration: WeightNoiseConfiguration,
 ): void
 ```
 
-Replace the network random number generator.
+Enable weight noise using either one global standard deviation or per-hidden-layer values.
 
-This lets advanced callers share one deterministic source across mutation,
-stochastic depth, DropConnect, and other runtime randomness.
+A single global value is useful for quick experiments, while the per-hidden
+schedule keeps layered models explicit about which hidden stage receives how
+much perturbation.
 
 Parameters:
 - `this` - Target network instance.
-- `randomFunction` - RNG function returning values in $[0,1)$.
+- `configuration` - Global standard deviation or one value per hidden layer.
 
 Returns: Nothing.
 
-### testForceOverflow
+### getLastSkippedLayers
 
 ```ts
-testForceOverflow(): void
+getLastSkippedLayers(): number[]
 ```
 
-Force the next mixed-precision overflow path.
-
-This is a test-oriented hook used to exercise loss-scale recovery logic
-without waiting for a real floating-point overflow.
+Read the last hidden-layer indices skipped by stochastic depth.
 
 Parameters:
 - `this` - Target network instance.
 
-Returns: Nothing.
+Returns: Snapshot of the last skipped hidden-layer indices.
 
-### getTrainingStep
+### getRuntimeRegularizationStats
 
 ```ts
-getTrainingStep(): number
+getRuntimeRegularizationStats(): Record<string, unknown> | null
 ```
 
-Read the current training-step counter.
+Read regularization statistics collected during training.
 
 Parameters:
 - `this` - Target network instance.
 
-Returns: Current training step.
+Returns: Last regularization stats payload or `null` when none exists yet.
 
-### getLastSkippedLayers
+### getTrainingStep
 
 ```ts
-getLastSkippedLayers(): number[]
+getTrainingStep(): number
 ```
 
-Read the last hidden-layer indices skipped by stochastic depth.
+Read the current training-step counter.
 
 Parameters:
 - `this` - Target network instance.
 
-Returns: Snapshot of the last skipped hidden-layer indices.
+Returns: Current training step.
 
-### setStochasticDepthSchedule
+### setRandom
 
 ```ts
-setStochasticDepthSchedule(
-  schedule: StochasticDepthSchedule,
+setRandom(
+  randomFunction: () => number,
 ): void
 ```
 
-Set the stochastic-depth schedule function.
+Replace the network random number generator.
+
+This lets advanced callers share one deterministic source across mutation,
+stochastic depth, DropConnect, and other runtime randomness.
 
 Parameters:
 - `this` - Target network instance.
-- `schedule` - Function mapping the current step and schedule to a new schedule.
+- `randomFunction` - RNG function returning values in $[0,1)$.
 
 Returns: Nothing.
 
-### clearStochasticDepthSchedule
+### setStochasticDepth
 
 ```ts
-clearStochasticDepthSchedule(): void
+setStochasticDepth(
+  survivalProbabilities: number[],
+): void
 ```
 
-Clear the stochastic-depth schedule function.
+Configure stochastic depth with one survival probability per hidden layer.
+
+Matching survival values to hidden layers keeps the runtime contract explicit
+and avoids silently applying one layer's policy to another.
 
 Parameters:
 - `this` - Target network instance.
+- `survivalProbabilities` - Survival probabilities for each hidden layer.
 
 Returns: Nothing.
 
-### getRuntimeRegularizationStats
+### setStochasticDepthSchedule
 
 ```ts
-getRuntimeRegularizationStats(): Record<string, unknown> | null
+setStochasticDepthSchedule(
+  schedule: StochasticDepthSchedule,
+): void
 ```
 
-Read regularization statistics collected during training.
+Set the stochastic-depth schedule function.
 
 Parameters:
 - `this` - Target network instance.
+- `schedule` - Function mapping the current step and schedule to a new schedule.
 
-Returns: Last regularization stats payload or `null` when none exists yet.
+Returns: Nothing.
 
-### setStochasticDepth
+### setWeightNoiseSchedule
 
 ```ts
-setStochasticDepth(
-  survivalProbabilities: number[],
+setWeightNoiseSchedule(
+  schedule: (step: number) => number,
 ): void
 ```
 
-Configure stochastic depth with one survival probability per hidden layer.
-
-Matching survival values to hidden layers keeps the runtime contract explicit
-and avoids silently applying one layer's policy to another.
+Set a dynamic scheduler for global weight noise.
 
 Parameters:
 - `this` - Target network instance.
-- `survivalProbabilities` - Survival probabilities for each hidden layer.
+- `schedule` - Function mapping the current training step to a standard deviation.
 
 Returns: Nothing.
 
-### disableStochasticDepth
+### testForceOverflow
 
 ```ts
-disableStochasticDepth(): void
+testForceOverflow(): void
 ```
 
-Disable stochastic depth.
+Force the next mixed-precision overflow path.
+
+This is a test-oriented hook used to exercise loss-scale recovery logic
+without waiting for a real floating-point overflow.
 
 Parameters:
 - `this` - Target network instance.
@@ -234,22 +234,6 @@ This chapter owns the public readers and small runtime controls that expose
 training-health state, DropConnect policy, and dropout-mask cleanup without
 changing the network topology itself.
 
-### enableDropConnect
-
-```ts
-enableDropConnect(
-  probability: number,
-): void
-```
-
-Enable DropConnect with a probability in $[0,1)$.
-
-Parameters:
-- `this` - Target network instance.
-- `probability` - DropConnect probability.
-
-Returns: Nothing.
-
 ### disableDropConnect
 
 ```ts
@@ -263,34 +247,34 @@ Parameters:
 
 Returns: Nothing.
 
-### resetDropoutMasks
+### enableDropConnect
 
 ```ts
-resetDropoutMasks(): void
+enableDropConnect(
+  probability: number,
+): void
 ```
 
-Reset every dropout mask to `1`.
-
-This is useful after training so later inference does not inherit transient
-node-level dropout state from a previous activation pass.
+Enable DropConnect with a probability in $[0,1)$.
 
 Parameters:
 - `this` - Target network instance.
+- `probability` - DropConnect probability.
 
 Returns: Nothing.
 
-### getRawGradientNorm
+### getLastGradClipGroupCount
 
 ```ts
-getRawGradientNorm(): number
+getLastGradClipGroupCount(): number
 ```
 
-Read the last recorded raw gradient norm.
+Read the last recorded gradient-clipping group count.
 
 Parameters:
 - `this` - Target network instance.
 
-Returns: Last raw gradient norm.
+Returns: Last gradient-clipping group count.
 
 ### getLossScale
 
@@ -305,18 +289,18 @@ Parameters:
 
 Returns: Current loss scale.
 
-### getLastGradClipGroupCount
+### getRawGradientNorm
 
 ```ts
-getLastGradClipGroupCount(): number
+getRawGradientNorm(): number
 ```
 
-Read the last recorded gradient-clipping group count.
+Read the last recorded raw gradient norm.
 
 Parameters:
 - `this` - Target network instance.
 
-Returns: Last gradient-clipping group count.
+Returns: Last raw gradient norm.
 
 ### getTrainingStats
 
@@ -330,3 +314,19 @@ Parameters:
 - `this` - Target network instance.
 
 Returns: Training statistics snapshot.
+
+### resetDropoutMasks
+
+```ts
+resetDropoutMasks(): void
+```
+
+Reset every dropout mask to `1`.
+
+This is useful after training so later inference does not inherit transient
+node-level dropout state from a previous activation pass.
+
+Parameters:
+- `this` - Target network instance.
+
+Returns: Nothing.