Enum Layer 

pub enum Layer<B: Backend> {
    Attention(Attention<B>),
    Bias(Bias<B>),
    BatchNorm(BatchNorm<B>),
    Cache(Cache<B>),
    Cat(Ignored<CatLayer>),
    Conv2d(Conv2d<B>),
    CrossEntropy(CrossEntropyLoss<B>),
    Dropout(Dropout),
    Embedding(Embedding<B>),
    Flatten(Ignored<FlattenLayer<Range<isize>>>),
    KQV(KQV<B>),
    LayerNorm(LayerNorm<B>),
    Linear(Linear<B>),
    MaxPool2d(MaxPool2d),
    Mse(MseLoss),
    Relu(Relu),
    Reshape(Ignored<ReshapeLayer<Reshape>>),
    Rotary(RotaryEncoding<B>),
    ScaleShift(ScaleShift<B>),
    Squeeze(Ignored<SqueezeLayer>),
    Stack(Ignored<StackLayer>),
    Sum(Ignored<SumLayer>),
    Tanh(Tanh),
    Unsqueeze(Ignored<UnsqueezeLayer>),
}

enumerates some burn layers

Variants

Attention(Attention<B>)

Bias(Bias<B>)

BatchNorm(BatchNorm<B>)

Cache(Cache<B>)

Cat(Ignored<CatLayer>)

Conv2d(Conv2d<B>)

CrossEntropy(CrossEntropyLoss<B>)

Dropout(Dropout)

Embedding(Embedding<B>)

Flatten(Ignored<FlattenLayer<Range<isize>>>)

KQV(KQV<B>)

LayerNorm(LayerNorm<B>)

Linear(Linear<B>)

MaxPool2d(MaxPool2d)

Mse(MseLoss)

Relu(Relu)

Reshape(Ignored<ReshapeLayer<Reshape>>)

Rotary(RotaryEncoding<B>)

ScaleShift(ScaleShift<B>)

Squeeze(Ignored<SqueezeLayer>)

Stack(Ignored<StackLayer>)

Sum(Ignored<SumLayer>)

Tanh(Tanh)

Unsqueeze(Ignored<UnsqueezeLayer>)

Implementations

impl<B: Backend> Layer<B>

pub fn attention(heads: usize, mask: AttentionMask) -> Self

creates an attention config

pub fn batch_norm(countfeatures: usize, epsilon: f32, momentum: f32) -> Self

creates a batch norm layer

pub fn bias(dim: usize) -> Self

creates a bias config

pub fn cache(limit: usize) -> Self

creates a cache layer

pub fn clear_cache(&mut self) -> bool

clears the cache if the layer has one

pub fn dropout(chance: f32) -> Self

creates a dropout layer

pub fn embedding(input: usize, output: usize, wscale: f32) -> Self

creates a embedding layer

pub fn flatten<R: RangeBounds<isize>>(dims: R) -> Self

creates a flatten layer

pub fn layer_norm(dim: usize) -> Self

creates a layer norm layer

pub fn linear(bias: bool, input: usize, output: usize, wscale: f32) -> Self

creates a linear layer

pub fn max_pool_2d(kernel: [usize; 2], strides: [usize; 2]) -> Self

creates a max pool 2d layer

pub fn relu() -> Self

creates a relu layer

pub fn reshape<R: Into<Reshape>>(args: R) -> Self

creates a reshape layer

pub fn rotary(distance: usize, head: usize) -> Self

creates a rotary layer

pub fn scale_shift() -> Self

creates a scale shift layer

pub fn tanh() -> Self

creates a tanh layer

Trait Implementations

impl<B: Backend> AI<Value<B>, Value<B>> for Layer<B>

fn forward(&self, input: Value<B>) -> Value<B>

applies to the input

fn forward_mut(&mut self, input: Value<B>) -> Value<B>

applies to the input, possibly updating internal caches

impl<B> AutodiffModule<B> for Layer<B>

where B: AutodiffBackend + Backend, <B as AutodiffBackend>::InnerBackend: Backend,

type InnerModule = Layer<<B as AutodiffBackend>::InnerBackend>

Inner module without auto-differentiation.

fn valid(&self) -> Self::InnerModule

Get the same module, but on the inner backend without auto-differentiation.

impl<B: Backend> Clone for Layer<B>

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl<B: Debug + Backend> Debug for Layer<B>

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

Formats the value using the given formatter.

impl<B: Backend> Decompose for Layer<B>

type Decomposition = Layer<B>

the decomposed type

fn compose(decomposition: Self::Decomposition) -> Self

recreates from the decomposition

fn decompose(self) -> Self::Decomposition

owned decomposition

fn decompose_cloned(&self) -> Self::Decomposition

decomposition that copies data

impl<'de, B: Backend> Deserialize<'de> for Layer<B>

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl<B: Backend> Display for Layer<B>

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

Formats the value using the given formatter.

impl<B: Backend> From<Attention<B>> for Layer<B>

fn from(value: Attention<B>) -> Self

Converts to this type from the input type.

impl<B: Backend> From<BatchNorm<B>> for Layer<B>

fn from(value: BatchNorm<B>) -> Self

Converts to this type from the input type.

impl<B: Backend> From<Cache<B>> for Layer<B>

fn from(value: Cache<B>) -> Self

Converts to this type from the input type.

impl<B: Backend> From<CatLayer> for Layer<B>

fn from(value: CatLayer) -> Self

Converts to this type from the input type.

impl<B: Backend> From<CrossEntropyLoss<B>> for Layer<B>

fn from(value: CrossEntropyLoss<B>) -> Self

Converts to this type from the input type.

impl<B: Backend> From<Dropout> for Layer<B>

fn from(value: Dropout) -> Self

Converts to this type from the input type.

impl<B: Backend> From<Embedding<B>> for Layer<B>

fn from(value: Embedding<B>) -> Self

Converts to this type from the input type.

impl<B: Backend> From<FlattenLayer<Range<isize>>> for Layer<B>

fn from(value: FlattenLayer<Range<isize>>) -> Self

Converts to this type from the input type.

impl<B: Backend> From<LayerNorm<B>> for Layer<B>

fn from(value: LayerNorm<B>) -> Self

Converts to this type from the input type.

impl<B: Backend> From<Linear<B>> for Layer<B>

fn from(value: Linear<B>) -> Self

Converts to this type from the input type.

impl<B: Backend> From<MaxPool2d> for Layer<B>

fn from(value: MaxPool2d) -> Self

Converts to this type from the input type.

impl<B: Backend> From<MseLoss> for Layer<B>

fn from(value: MseLoss) -> Self

Converts to this type from the input type.

impl<B: Backend> From<Relu> for Layer<B>

fn from(value: Relu) -> Self

Converts to this type from the input type.

impl<B: Backend> From<ReshapeLayer<Reshape>> for Layer<B>

fn from(value: ReshapeLayer<Reshape>) -> Self

Converts to this type from the input type.

impl<B: Backend> From<RotaryEncoding<B>> for Layer<B>

fn from(value: RotaryEncoding<B>) -> Self

Converts to this type from the input type.

impl<B: Backend> From<SqueezeLayer> for Layer<B>

fn from(value: SqueezeLayer) -> Self

Converts to this type from the input type.

impl<B: Backend> From<StackLayer> for Layer<B>

fn from(value: StackLayer) -> Self

Converts to this type from the input type.

impl<B: Backend> From<SumLayer> for Layer<B>

fn from(value: SumLayer) -> Self

Converts to this type from the input type.

impl<B: Backend> From<Tanh> for Layer<B>

fn from(value: Tanh) -> Self

Converts to this type from the input type.

impl<B: Backend> From<UnsqueezeLayer> for Layer<B>

fn from(value: UnsqueezeLayer) -> Self

Converts to this type from the input type.

impl<B: Backend, M: AI<M::Output, M::Output> + Op> IntoSequence<M> for Layer<B>

where Layer<B>: Into<M>,

fn into_sequence(self) -> Sequential<Vec<M>>

converts into a sequential module list

impl<B: Backend> Module<B> for Layer<B>

type Record = LayerRecord<B>

Type to save and load the module.

fn load_record(self, record: Self::Record) -> Self

Load the module state from a record.

fn into_record(self) -> Self::Record

Convert the module into a record containing the state.

fn num_params(&self) -> usize

Get the number of parameters the module has, including all of its sub-modules.

fn visit<Visitor: ModuleVisitor<B>>(&self, visitor: &mut Visitor)

Visit each tensor parameter in the module with a visitor.

fn map<Mapper: ModuleMapper<B>>(self, mapper: &mut Mapper) -> Self

Map each tensor parameter in the module with a mapper.

fn collect_devices(&self, devices: Devices<B>) -> Devices<B>

Return all the devices found in the underneath module tree added to the given vector without duplicates.

fn to_device(self, device: &B::Device) -> Self

Move the module and all of its sub-modules to the given device.

fn fork(self, device: &B::Device) -> Self

Fork the module and all of its sub-modules to the given device.

fn devices(&self) -> Vec<<B as Backend>::Device>

Return all the devices found in the underneath module tree without duplicates.

fn no_grad(self) -> Self

Each tensor in the module tree will not require grad.

fn save_file<FR, PB>( self, file_path: PB, recorder: &FR, ) -> Result<(), RecorderError>

where FR: FileRecorder<B>, PB: Into<PathBuf>,

Save the module to a file using the provided file recorder.

fn load_file<FR, PB>( self, file_path: PB, recorder: &FR, device: &<B as Backend>::Device, ) -> Result<Self, RecorderError>

where FR: FileRecorder<B>, PB: Into<PathBuf>,

Load the module from a file using the provided file recorder.

fn quantize_weights(self, quantizer: &mut Quantizer) -> Self

Quantize the weights of the module.

impl<B: Backend> ModuleDisplay for Layer<B>

fn format(&self, passed_settings: DisplaySettings) -> String

Formats the module with provided display settings.

fn custom_settings(&self) -> Option<DisplaySettings>

Custom display settings for the module.

fn custom_content(&self, _content: Content) -> Option<Content>

Custom attributes for the module.

impl<B: Backend> ModuleDisplayDefault for Layer<B>

fn content(&self, content: Content) -> Option<Content>

Attributes of the module used for display purposes.

fn num_params(&self) -> usize

Gets the number of the parameters of the module.

impl<B: Backend> Op for Layer<B>

type Output = Value<B>

suggested output type to help with composition coherence. Ideally, Self should implement AI<X,Self::Output> for some X

fn abs(self) -> Abs<Self>

where Self: Sized, Abs<Self>: Op,

wraps with an absolute value operation

fn acc_q(self, gamma: f32) -> AccQ<Self>

where AccQ<Self>: Op, Self: Sized,

wraps with a accq operation

fn cat(self, dim: i32) -> Cat<Self>

where Cat<Self>: Op, Self: Sized,

wraps with a cat operation

fn chain<B>(self, b: B) -> Sequential<(Self, B)>

where Self: Sized, Sequential<(Self, B)>: Op,

sequences with another ai operation

fn cross_entropy(self, temperature: f32) -> CrossEntropy<Self>

where CrossEntropy<Self>: Op, Self: Sized,

wraps with a cross entropy operation. If temperature is a number it will be used to apply softmax to the logits before computing entropy with the target. if the input will already be a probability distribution instead of logits, put NaN temperature

fn duplicate(self) -> Duplicate<Self>

where Duplicate<Self>: Op, Self: Sized,

wraps with a duplicate operation

fn fix_type<Z>(self) -> SetType<Self, Z, Z>

where Self: AI<Z, Z> + Sized,

set type but with the same input and output

fn flatten<R: Clone>(self, args: R) -> Flatten<Self, R>

where Flatten<Self, R>: Op, Self: Sized,

wraps with a flatten layer

fn forward_fixed<Z>(&self, input: Z) -> Z

where Self: AI<Z, Z> + Sized,

applies to the input

fn forward_fixed_mut<Z>(&mut self, input: Z) -> Z

where Self: AI<Z, Z> + Sized,

applies to the input

fn forward_typed<W, Z>(&self, input: W) -> Z

where Self: AI<W, Z> + Sized,

applies to the input

fn forward_typed_mut<W, Z>(&mut self, input: W) -> Z

where Self: AI<W, Z> + Sized,

applies to the input, possibly updating internal caches

fn infer_autoregressive<X, Y>(self, input: X) -> Autoregression<Self, Y>

where Self: AI<X, Y> + AI<Y, Y> + Sized, Y: Clone,

creates an autoregressive inference

fn map<B>(self, b: B) -> Map<Sequential<(Self, B)>>

where Map<Sequential<(Self, B)>>: Op, Self: Sized, Sequential<(Self, B)>: Op,

applies the operation to every output

fn mean(self) -> Mean<Self>

where Mean<Self>: Op, Self: Sized,

wraps with a mean operation

fn optional(self) -> Option<Self>

where Self: Sized,

creates an optional operation

fn reshape<R: Clone>(self, args: R) -> Reshape<Self, R>

where Reshape<Self, R>: Op, Self: Sized,

wraps with a reshape layer

fn residual(self) -> Residual<Self>

where Residual<Self>: Op, Self: Sized,

wraps in a residual layer

fn set_type<W, Z>(self) -> SetType<Self, W, Z>

where Self: AI<W, Z> + Sized,

sets the input output types

fn soft_choose(self, temperature: f32) -> Choose<Self>

where Self: Sized, Choose<Self>: Op,

wraps with a choose operation

fn squared_error(self) -> SquaredError<Self>

where SquaredError<Self>: Op, Self: Sized,

wraps with a mse operation

fn squeeze(self, dim: i32) -> Squeeze<Self>

where Squeeze<Self>: Op, Self: Sized,

wraps with a squeeze operation

fn stack(self, dim: i32) -> Stack<Self>

where Stack<Self>: Op, Self: Sized,

wraps with a stack operation

fn to_each(self) -> Map<Self>

where Map<Self>: Op, Self: Sized,

wraps with a map operation

fn unsqueeze(self, dim: i32) -> Unsqueeze<Self>

where Unsqueeze<Self>: Op, Self: Sized,

wraps with a unsqueeze operation

fn sum(self) -> Sum<Self>

where Sum<Self>: Op, Self: Sized,

wraps with a sum operation

fn wrap_inner(self) -> Inner<Self>

where Self: Sized,

wraps the inner value so it can be unwrapped with unwrap inner

fn zip<B>(self, b: B) -> Zip<(Self, B)>

where Self: Sized, Zip<(Self, B)>: Op,

zips with another ai operation

impl<B: Backend> Serialize for Layer<B>

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl<B: Backend> Wrappable for Layer<B>

type B = B

type With<C: Backend> = Layer<C>