Struct RnnImageInferenceLayer 

pub struct RnnImageInferenceLayer { /* private fields */ }

Wraps MPSRNNImageInferenceLayer.

Implementations

impl RnnImageInferenceLayer

pub const fn as_ptr(&self) -> *mut c_void

Returns the retained Objective-C pointer backing this wrapper.

impl RnnImageInferenceLayer

pub fn new(device: &MetalDevice, descriptor: &RnnDescriptor) -> Option<Self>

Wraps a constructor on MPSRNNImageInferenceLayer.

Examples found in repository

examples/09_rnn_image_inference.rs (line 13)

fn main() {
    let device = MetalDevice::system_default().expect("no Metal device available");
    let queue = device.new_command_queue().expect("command queue");

    let single_gate = RnnSingleGateDescriptor::new(1, 1).expect("single gate descriptor");
    single_gate.set_use_layer_input_unit_transform_mode(true);
    let descriptor = single_gate.as_descriptor().expect("base descriptor");
    let layer = RnnImageInferenceLayer::new(&device, &descriptor).expect("rnn layer");

    let image_descriptor = ImageDescriptor::new(1, 1, 1, feature_channel_format::FLOAT32);
    let src0 = Image::new(&device, image_descriptor).expect("src0");
    let src1 = Image::new(&device, image_descriptor).expect("src1");
    let dst0 = Image::new(&device, image_descriptor).expect("dst0");
    let dst1 = Image::new(&device, image_descriptor).expect("dst1");
    src0.write_f32(&[0.25]).expect("write src0");
    src1.write_f32(&[0.75]).expect("write src1");

    let command_buffer = queue.new_command_buffer().expect("command buffer");
    let recurrent_state = layer
        .encode_sequence(&command_buffer, &[&src0, &src1], &[&dst0, &dst1], None)
        .expect("recurrent state");
    command_buffer.commit();
    command_buffer.wait_until_completed();

    let recurrent_output = recurrent_state
        .recurrent_output_image_for_layer_index(0)
        .expect("recurrent output image");
    println!(
        "{:?} {}x{}x{}",
        dst1.read_f32().expect("dst1 output"),
        recurrent_output.width(),
        recurrent_output.height(),
        recurrent_output.feature_channels()
    );
}

pub fn new_stack( device: &MetalDevice, descriptors: &[&RnnDescriptor], ) -> Option<Self>

Wraps a constructor on MPSRNNImageInferenceLayer.

pub fn input_feature_channels(&self) -> usize

Wraps the corresponding MPSRNNImageInferenceLayer method.

pub fn output_feature_channels(&self) -> usize

Wraps the corresponding MPSRNNImageInferenceLayer method.

pub fn number_of_layers(&self) -> usize

Wraps the corresponding MPSRNNImageInferenceLayer method.

pub fn recurrent_output_is_temporary(&self) -> bool

Wraps the corresponding MPSRNNImageInferenceLayer method.

pub fn set_recurrent_output_is_temporary(&self, value: bool)

Wraps the corresponding MPSRNNImageInferenceLayer setter.

pub fn store_all_intermediate_states(&self) -> bool

Wraps the corresponding MPSRNNImageInferenceLayer method.

pub fn set_store_all_intermediate_states(&self, value: bool)

Wraps the corresponding MPSRNNImageInferenceLayer setter.

pub fn bidirectional_combine_mode(&self) -> usize

Wraps the corresponding MPSRNNImageInferenceLayer method.

pub fn set_bidirectional_combine_mode(&self, value: usize)

Wraps the corresponding MPSRNNImageInferenceLayer setter.

pub fn encode_sequence( &self, command_buffer: &CommandBuffer, source_images: &[&Image], destination_images: &[&Image], recurrent_input_state: Option<&RnnRecurrentImageState>, ) -> Option<RnnRecurrentImageState>

Wraps the corresponding MPSRNNImageInferenceLayer encode entry point.

Examples found in repository

examples/09_rnn_image_inference.rs (line 25)

fn main() {
    let device = MetalDevice::system_default().expect("no Metal device available");
    let queue = device.new_command_queue().expect("command queue");

    let single_gate = RnnSingleGateDescriptor::new(1, 1).expect("single gate descriptor");
    single_gate.set_use_layer_input_unit_transform_mode(true);
    let descriptor = single_gate.as_descriptor().expect("base descriptor");
    let layer = RnnImageInferenceLayer::new(&device, &descriptor).expect("rnn layer");

    let image_descriptor = ImageDescriptor::new(1, 1, 1, feature_channel_format::FLOAT32);
    let src0 = Image::new(&device, image_descriptor).expect("src0");
    let src1 = Image::new(&device, image_descriptor).expect("src1");
    let dst0 = Image::new(&device, image_descriptor).expect("dst0");
    let dst1 = Image::new(&device, image_descriptor).expect("dst1");
    src0.write_f32(&[0.25]).expect("write src0");
    src1.write_f32(&[0.75]).expect("write src1");

    let command_buffer = queue.new_command_buffer().expect("command buffer");
    let recurrent_state = layer
        .encode_sequence(&command_buffer, &[&src0, &src1], &[&dst0, &dst1], None)
        .expect("recurrent state");
    command_buffer.commit();
    command_buffer.wait_until_completed();

    let recurrent_output = recurrent_state
        .recurrent_output_image_for_layer_index(0)
        .expect("recurrent output image");
    println!(
        "{:?} {}x{}x{}",
        dst1.read_f32().expect("dst1 output"),
        recurrent_output.width(),
        recurrent_output.height(),
        recurrent_output.feature_channels()
    );
}

Trait Implementations

impl Drop for RnnImageInferenceLayer

fn drop(&mut self)

Executes the destructor for this type.

fn pin_drop(self: Pin<&mut Self>)

🔬This is a nightly-only experimental API. (pin_ergonomics)

Execute the destructor for this type, but different to Drop::drop, it requires self to be pinned.

impl Send for RnnImageInferenceLayer

impl Sync for RnnImageInferenceLayer