ITrainable<'Smpl, 'T>

Namespace: Models
Parent Module: Train

Interface for a trainable model.

Instance members

Instance member	Description
`InitModel(seed)` Signature: seed:int -> unit Modifiers: abstract	Initializes the model using the given random seed.
`InitOptState()` Signature: unit -> unit Modifiers: abstract	Initialize optimizer state.
`LoadModel(hdf prefix)` Signature: hdf:HDF5 -> prefix:string -> unit Modifiers: abstract	Load model parameters from specified file.
`LoadOptState(hdf prefix)` Signature: hdf:HDF5 -> prefix:string -> unit Modifiers: abstract	Load optimizer state from specified file.
`Losses(sample)` Signature: sample:'Smpl -> float list Modifiers: abstract	Loss of given sample.
`ModelParameters()` Signature: unit -> Tensor<'T> Modifiers: abstract	Model parameter values (i.e. weights). CompiledName: `set_ModelParameters`
`ModelParameters()` Signature: unit -> unit Modifiers: abstract	Model parameter values (i.e. weights). CompiledName: `get_ModelParameters`
`Optimize(learningRate sample)` Signature: learningRate:float -> sample:'Smpl -> Lazy<float list> Modifiers: abstract	Perform an optimization step with the given learning rate and sample and return the loss.
`PrintInfo()` Signature: unit -> unit Modifiers: abstract	Prints information about the model.
`ResetModelState()` Signature: unit -> unit Modifiers: abstract	Resets the internal model state. (for example the latent state of an RNN)
`SaveModel(hdf prefix)` Signature: hdf:HDF5 -> prefix:string -> unit Modifiers: abstract	Save model parameters to specified file.
`SaveOptState(hdf prefix)` Signature: hdf:HDF5 -> prefix:string -> unit Modifiers: abstract	Save optimizer state to specified file.

Fork me on GitHub