Struct CModule

pub struct CModule { /* private fields */ }

A jit PyTorch module.

These modules can be created via the TorchScript python api.

Implementations

impl CModule

pub fn load<T: AsRef<Path>>(path: T) -> Result<CModule, TchError>

Loads a PyTorch saved JIT model from a file.

pub fn load_on_device<T: AsRef<Path>>( path: T, device: Device, ) -> Result<CModule, TchError>

Loads a PyTorch saved JIT model from a file onto the given device.

This function loads the model directly on the specified device, which means it also allows loading a GPU model on the CPU without having a CUDA enabled GPU.

pub fn load_data<T: Read>(f: &mut T) -> Result<CModule, TchError>

Loads a PyTorch saved JIT model from a read instance.

pub fn load_data_on_device<T: Read>( f: &mut T, device: Device, ) -> Result<CModule, TchError>

Loads a PyTorch saved JIT model from a read instance.

This function loads the model directly on the specified device, which means it also allows loading a GPU model on the CPU without having a CUDA enabled GPU.

pub fn forward_ts<T: Borrow<Tensor>>( &self, ts: &[T], ) -> Result<Tensor, TchError>

Performs the forward pass for a model on some specified tensor inputs. This is equivalent to calling method_ts with the ‘forward’ method name, and returns a single tensor.

pub fn forward_is<T: Borrow<IValue>>( &self, ts: &[T], ) -> Result<IValue, TchError>

Performs the forward pass for a model on some specified ivalue inputs. This is equivalent to calling method_is with the ‘forward’ method name, and returns an arbitrary ivalue.

pub fn method_ts<T: Borrow<Tensor>>( &self, method_name: &str, ts: &[T], ) -> Result<Tensor, TchError>

Runs a specified entry point for a model on some given tensor inputs.

pub fn method_is<T: Borrow<IValue>>( &self, method_name: &str, ts: &[T], ) -> Result<IValue, TchError>

Runs a specified entry point for a model on some given ivalue inputs.

pub fn create_class_is<T: Borrow<IValue>>( &self, clz_name: &str, ts: &[T], ) -> Result<IValue, TchError>

Create a specified custom JIT class object with the given class name, eg: __torch__.foo.Bar

pub fn f_set_eval(&mut self) -> Result<(), TchError>

Switches the module to evaluation mode.

pub fn set_eval(&mut self)

Switches the module to evaluation mode.

pub fn f_set_train(&mut self) -> Result<(), TchError>

Switches the module to training mode.

pub fn set_train(&mut self)

Switches the module to training mode.

pub fn to(&mut self, device: Device, kind: Kind, non_blocking: bool)

Moves the module to a different device and converts the kind.

pub fn save<T: AsRef<Path>>(&self, path: T) -> Result<(), TchError>

Saves a module to a given path.

pub fn named_parameters(&self) -> Result<Vec<(String, Tensor)>, TchError>

Loads some named tensors from a module

pub fn create_by_tracing<F>( modl_name: &str, fn_name: &str, inputs: &[Tensor], closure: &mut F, ) -> Result<CModule, TchError>

where F: FnMut(&[Tensor]) -> Vec<Tensor>,

Create a new module by tracing the application of the specified function on the given inputs.

Trait Implementations

impl Debug for CModule

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Drop for CModule

fn drop(&mut self)

Executes the destructor for this type.

impl Module for CModule

fn forward(&self, xs: &Tensor) -> Tensor

impl Send for CModule

impl Sync for CModule