Struct Mamba2Layers 

pub struct Mamba2Layers<B: Backend> {
    pub n_real_layers: usize,
    pub n_virtual_layers: Option<(usize, Schedule)>,
    pub real_layers: Vec<Mamba2Layer<B>>,
    pub ignore_first_residual: bool,
    pub ignore_last_residual: bool,
}

A stack of Mamba-2 layers with optional virtual-layer scheduling.

The stack maintains n_real_layers distinct weight sets but can execute n_virtual_layers logical forward passes, cycling through weights according to the provided Schedule.

Fields

n_real_layers: usize

Number of real (weight-bearing) layers.

n_virtual_layers: Option<(usize, Schedule)>

Optional (n_virtual_layers, schedule) for weight-sharing.

When None, the virtual layer count falls back to n_real_layers (no sharing). Marked module(skip) so Burn does not treat it as a trainable parameter.

real_layers: Vec<Mamba2Layer<B>>

The actual weight-bearing layer instances.

Length: n_real_layers.

ignore_first_residual: bool

When true, the residual connection of the first virtual layer is scaled to zero (i.e. the first block acts as a pure projection, not a residual update).

ignore_last_residual: bool

When true, the residual connection of the last virtual layer is scaled to zero.

Implementations

impl<B: Backend + Mamba2BackendExt> Mamba2Layers<B>

pub fn forward( &self, x: Tensor<B, 3>, caches: Option<Mamba2Caches<B>>, ssd_path: Mamba2SsdPath, ) -> (Tensor<B, 3>, Mamba2Caches<B>)

Process a full sequence through every (virtual) layer.

Internally each layer calls Mamba2::forward, which runs the chunkwise SSD algorithm. This is efficient for training because the intra-chunk products can exploit GEMM / tensor cores.

If caches is None, zero-initialised caches are created automatically.

Arguments

• x — input tensor, shape [batch, sequence, d_model]
• caches — optional pre-filled layer caches (useful for prefill followed by decode)
• ssd_path — SSD algorithm and chunk length selection.

Returns

(output, updated_caches) where output has shape [batch, sequence, d_model].

pub fn step( &self, x: Tensor<B, 2>, caches: Option<Mamba2Caches<B>>, ) -> (Tensor<B, 2>, Mamba2Caches<B>)

Process a single token through every (virtual) layer.

Each layer calls Mamba2::step, which runs one tick of the recurrent SSM: hₜ = Āₜ hₜ₋₁ + B̄ₜ xₜ, yₜ = Cₜᵀ hₜ + D xₜ. This is O(H·P·N) per step — independent of sequence length — and requires no KV-cache.

Arguments

• x — current token embedding, shape [batch, d_model]
• caches — layer caches from the previous step (or None for the first token, in which case zero caches are created)

Returns

(output, updated_caches) where output has shape [batch, d_model].

Trait Implementations

impl<B> AutodiffModule<B> for Mamba2Layers<B>

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

type InnerModule = Mamba2Layers<<B as AutodiffBackend>::InnerBackend>

Inner module without auto-differentiation.

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

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

fn from_inner(module: Self::InnerModule) -> Self

Wraps an inner module back into an auto-diff module.

impl<B: Backend> Clone for Mamba2Layers<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 Mamba2Layers<B>

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

Formats the value using the given formatter.

impl<B: Backend> Display for Mamba2Layers<B>

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

Formats the value using the given formatter.

impl<B> HasAutodiffModule<B> for Mamba2Layers<B::InnerBackend>

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

type TrainModule = Mamba2Layers<B>

The module with auto-differentiation.

impl<B: Backend> Module<B> for Mamba2Layers<B>

type Record = Mamba2LayersRecord<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 BackendTypes>::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 train<AB>(self) -> Self::TrainModule

where AB: AutodiffBackend<InnerBackend = B>, Self: HasAutodiffModule<AB>,

Move the module and all of its sub-modules to the autodiff backend.

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

Quantize the weights of the module.

impl<B: Backend> ModuleDisplay for Mamba2Layers<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 Mamba2Layers<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.