Core acceleration and memory

Cross-step caches, token pruning, fused kernels, VRAM lifecycle wrappers, disk maps, and memory stores.

On this page

Core separates exact implementation acceleration from approximation policy. Fused kernels and placement changes aim to preserve model semantics. Cross-step caches and token pruning intentionally trade computation for an approximation and therefore expose thresholds, retained ratios, events, and reset boundaries that integrations can evaluate.

Cross-step cache example

import torch

from worldfoundry.core.acceleration import FixedStepCache

cache = FixedStepCache(
    skip_steps={1, 3},
    dense_first=1,
    dense_last=1,
    total_steps=5,
)
calls = []

with torch.no_grad():
    outputs = []
    for step in range(5):
        def compute(step=step):
            calls.append(step)
            return torch.tensor([float(step)])

        outputs.append(cache.run(step, compute))

assert calls == [0, 2, 4]  # boundary steps remain dense
assert [event.hit for event in cache.events] == [False, True, False, True, False]

Caches disable replay while autograd is enabled. Call reset() between independent denoising trajectories; otherwise prior residuals become state for the next request. Treat the event stream as evidence during latency/quality evaluation rather than assuming requested skip steps were all used.

Token pruning lifecycle

select_token_indices is stateless. prune_tokens returns compact data plus TokenPruneState, and restore_tokens scatters processed tokens back, filling dropped positions from compensation or zeros. TokenPruner adds previous-step compensation: its first call records a dense segment, later calls can prune, and every prune must be paired with restore under the same key.

VRAM placement state

enable_vram_management replaces classes listed in module_map with wrappers such as AutoWrappedLinear. Each wrapper can use separate offload, onload, preparing, and computation dtype/device pairs. Disk-backed wrappers resolve parameter names through DiskMap; ordinary wrappers move or copy tensors between devices.

This transformation mutates module identity and should normally run once during model construction. The module-map order matters for overlapping classes. A placement configuration is not automatically safe for a new architecture: validate peak memory, transfer overlap, output equality/tolerance, and behavior when the configured VRAM limit is reached.

Memory records versus VRAM

The BaseMemory and MemoryStore APIs represent retrievable world-model memory records. They do not manage GPU allocation. VRAM wrappers manage parameter placement but do not provide semantic retrieval. Keeping those meanings separate prevents a “memory” setting from being applied at the wrong layer.

Complete reference

The blocks below are the generated signatures for this category. Use the on-page symbol index to jump; source links open the defining implementation behind each lazy export.

26 public symbols

class AdaptiveResidualCache(threshold: float,warmup_steps: int = 1,max_consecutive_hits: int = 3,dense_last: int = 1,total_steps: int | None = None,subsample: int = 1,eps: float = 1e-06)
worldfoundry.core.acceleration.AdaptiveResidualCachefrom worldfoundry.core.acceleration import AdaptiveResidualCache
source

Overview

Reuse a model residual while accumulated input change stays small. Belongs to Core acceleration and memory (caches, offload, VRAM, kernels).

Source docstring

Reuse a model residual while accumulated input change stays small.

Warm up densely, estimate normalized change from a cheap signal, accumulate it across steps, and replay the last dense residual until the threshold or consecutive-hit cap is reached.

Parameters

thresholdfloat
warmup_stepsint
default: 1
max_consecutive_hitsint
default: 3
dense_lastint
default: 1
total_stepsint | None
default: None
subsampleint
default: 1
epsfloat
default: 1e-06

Methods

methreset() -> Nonesource

Overview

Public method on this type.

Returns: None

methrun(step: int,signal: torch.Tensor,compute_residual: Callable[[], torch.Tensor],total_steps: int | None = None) -> torch.Tensorsource

Overview

Return a dense or cached residual for one denoising step.

Parameters

stepint
signaltorch.Tensor
compute_residualCallable[[], torch.Tensor]
total_stepsint | None
default: None

Returns: torch.Tensor

class AutoTorchModule(offload_dtype: torch.dtype = None,offload_device: Union[str, torch.device] = None,onload_dtype: torch.dtype = None,onload_device: Union[str, torch.device] = None,preparing_dtype: torch.dtype = None,preparing_device: Union[str, torch.device] = None,computation_dtype: torch.dtype = None,computation_device: Union[str, torch.device] = None,vram_limit: float = None)
worldfoundry.core.AutoTorchModulefrom worldfoundry.core import AutoTorchModule
source

Overview

Base state machine for modules that move between memory tiers. Belongs to Core acceleration and memory (caches, offload, VRAM, kernels).

Source docstring

Base state machine for modules that move between memory tiers.

`state is 0 while offloaded, 1 while onloaded, and 2` while kept on the computation device. Subclasses decide whether transitions copy tensors, materialize them from disk, or use temporary computation weights.

Parameters

offload_dtypetorch.dtype
Storage dtype while the module is inactive.default: None
offload_deviceUnion[str, torch.device]
Storage device while inactive; subclasses may also accept the sentinel `"disk"`.default: None
onload_dtypetorch.dtype
Dtype after the first prefetch transition.default: None
onload_deviceUnion[str, torch.device]
Device used for prefetched weights.default: None
preparing_dtypetorch.dtype
Dtype used by the optional second prefetch stage.default: None
preparing_deviceUnion[str, torch.device]
Device used by the preparing stage.default: None
computation_dtypetorch.dtype
Dtype used for the actual forward operation.default: None
computation_deviceUnion[str, torch.device]
Device used for the actual forward operation.default: None
vram_limitfloat
Optional used-memory limit in GiB. Below the limit, wrappers may keep prepared weights resident.default: None

Methods

methset_dtype_and_device(offload_dtype: torch.dtype = None,offload_device: Union[str, torch.device] = None,onload_dtype: torch.dtype = None,onload_device: Union[str, torch.device] = None,preparing_dtype: torch.dtype = None,preparing_device: Union[str, torch.device] = None,computation_dtype: torch.dtype = None,computation_device: Union[str, torch.device] = None,vram_limit: float = None)source

Overview

Update lifecycle placement, defaulting omitted stages to computation placement.

Parameters

offload_dtypetorch.dtype
default: None
offload_deviceUnion[str, torch.device]
default: None
onload_dtypetorch.dtype
default: None
onload_deviceUnion[str, torch.device]
default: None
preparing_dtypetorch.dtype
default: None
preparing_deviceUnion[str, torch.device]
default: None
computation_dtypetorch.dtype
default: None
computation_deviceUnion[str, torch.device]
default: None
vram_limitfloat
default: None
methcast_to(weight,dtype,device)source

Overview

Copy one tensor to `dtype and device` without mutating the source.

Parameters

weight
dtype
device
methcheck_free_vram()source

Overview

Return whether current accelerator usage is below `vram_limit`.

methoffload()source

Overview

Move managed parameters to the inactive placement and set state 0.

methonload()source

Overview

Move managed parameters to the first prefetch placement and set state 1.

methkeep()source

Overview

Keep managed parameters at computation placement and set state 2.

methparam_name(name)source

Overview

Return the fully qualified checkpoint key for a local parameter name.

Parameters

name
class AutoWrappedLinear(module: torch.nn.Linear,offload_dtype: torch.dtype = None,offload_device: Union[str, torch.device] = None,onload_dtype: torch.dtype = None,onload_device: Union[str, torch.device] = None,preparing_dtype: torch.dtype = None,preparing_device: Union[str, torch.device] = None,computation_dtype: torch.dtype = None,computation_device: Union[str, torch.device] = None,vram_limit: float = None,name: str = '',disk_map: DiskMap = None,**kwargs)
worldfoundry.core.AutoWrappedLinearfrom worldfoundry.core import AutoWrappedLinear
source

Overview

Linear-layer wrapper with staged placement, disk loading, FP8, and LoRA support. Belongs to Core acceleration and memory (caches, offload, VRAM, kernels).

Parameters

moduletorch.nn.Linear
Source `torch.nn.Linear` whose parameters are reused.
offload_dtypetorch.dtype
Inactive storage dtype or `"disk"`.default: None
offload_deviceUnion[str, torch.device]
Inactive storage device or `"disk"`.default: None
onload_dtypetorch.dtype
First prefetch dtype.default: None
onload_deviceUnion[str, torch.device]
First prefetch device.default: None
preparing_dtypetorch.dtype
Second prefetch dtype.default: None
preparing_deviceUnion[str, torch.device]
Second prefetch device.default: None
computation_dtypetorch.dtype
Matmul dtype; supported FP8 dtypes use scaled MM.default: None
computation_deviceUnion[str, torch.device]
Matmul device.default: None
vram_limitfloat
Optional used-memory limit in GiB.default: None
namestr
Qualified checkpoint prefix for weight and bias.default: ''
disk_mapDiskMap
Required lazy tensor map when disk offload is selected.default: None
kwargs
Reserved for module-map compatibility.

Methods

methfp8_linear(input: torch.Tensor,weight: torch.Tensor,bias: torch.Tensor = None) -> torch.Tensorsource

Overview

Public method on this type.

Parameters

inputtorch.Tensor
weighttorch.Tensor
biastorch.Tensor
default: None

Returns: torch.Tensor

methload_from_disk(torch_dtype,device,assign = True)source

Overview

Public method on this type.

Parameters

torch_dtype
device
assign
default: True
methoffload()source

Overview

Public method on this type.

methonload()source

Overview

Public method on this type.

methpreparing()source

Overview

Public method on this type.

methcomputation()source

Overview

Public method on this type.

methlinear_forward(x,weight,bias)source

Overview

Public method on this type.

Parameters

x
weight
bias
methlora_forward(x, out)source

Overview

Public method on this type.

Parameters

x
out
methforward(x,*args,**kwargs)source

Overview

Public method on this type.

Parameters

x
args
kwargs
class AutoWrappedModule(module: torch.nn.Module,offload_dtype: torch.dtype = None,offload_device: Union[str, torch.device] = None,onload_dtype: torch.dtype = None,onload_device: Union[str, torch.device] = None,preparing_dtype: torch.dtype = None,preparing_device: Union[str, torch.device] = None,computation_dtype: torch.dtype = None,computation_device: Union[str, torch.device] = None,vram_limit: float = None,name: str = '',disk_map: DiskMap = None,**kwargs)
worldfoundry.core.AutoWrappedModulefrom worldfoundry.core import AutoWrappedModule
source

Overview

Wrap an arbitrary module with staged CPU/GPU/disk weight movement. Belongs to Core acceleration and memory (caches, offload, VRAM, kernels).

Source docstring

Wrap an arbitrary module with staged CPU/GPU/disk weight movement.

Forward calls opportunistically prepare weights, materialize a computation copy only when placement differs, and then delegate to the wrapped module. Attribute access falls through to the wrapped module so model code can keep using its original interface.

Parameters

moduletorch.nn.Module
Original PyTorch module.
offload_dtypetorch.dtype
Inactive storage dtype or `"disk"`.default: None
offload_deviceUnion[str, torch.device]
Inactive storage device or `"disk"`.default: None
onload_dtypetorch.dtype
First-stage prefetch dtype.default: None
onload_deviceUnion[str, torch.device]
First-stage prefetch device.default: None
preparing_dtypetorch.dtype
Second-stage prefetch dtype.default: None
preparing_deviceUnion[str, torch.device]
Second-stage prefetch device.default: None
computation_dtypetorch.dtype
Forward-pass dtype.default: None
computation_deviceUnion[str, torch.device]
Forward-pass device.default: None
vram_limitfloat
Optional used-memory limit in GiB.default: None
namestr
Qualified module name used to resolve disk-map keys.default: ''
disk_mapDiskMap
Lazy checkpoint tensor mapping required for disk offload.default: None
kwargs
Reserved for wrapper-compatible module maps.

Methods

methload_from_disk(torch_dtype,device,copy_module = False)source

Overview

Public method on this type.

Parameters

torch_dtype
device
copy_module
default: False
methoffload_to_disk(model: torch.nn.Module)source

Overview

Public method on this type.

Parameters

modeltorch.nn.Module
methoffload()source

Overview

Public method on this type.

methonload()source

Overview

Public method on this type.

methpreparing()source

Overview

Public method on this type.

methcast_to(module,dtype,device)source

Overview

Public method on this type.

Parameters

module
dtype
device
methcomputation()source

Overview

Public method on this type.

methforward(args, **kwargs)source

Overview

Public method on this type.

Parameters

args
kwargs
class AutoWrappedNonRecurseModule(module: torch.nn.Module,offload_dtype: torch.dtype = None,offload_device: Union[str, torch.device] = None,onload_dtype: torch.dtype = None,onload_device: Union[str, torch.device] = None,preparing_dtype: torch.dtype = None,preparing_device: Union[str, torch.device] = None,computation_dtype: torch.dtype = None,computation_device: Union[str, torch.device] = None,vram_limit: float = None,name: str = '',disk_map: DiskMap = None,**kwargs)
worldfoundry.core.AutoWrappedNonRecurseModulefrom worldfoundry.core import AutoWrappedNonRecurseModule
source

Overview

Manage only a module's direct parameters while its children are wrapped separately. Belongs to Core acceleration and memory (caches, offload, VRAM, kernels).

Parameters

moduletorch.nn.Module
offload_dtypetorch.dtype
default: None
offload_deviceUnion[str, torch.device]
default: None
onload_dtypetorch.dtype
default: None
onload_deviceUnion[str, torch.device]
default: None
preparing_dtypetorch.dtype
default: None
preparing_deviceUnion[str, torch.device]
default: None
computation_dtypetorch.dtype
default: None
computation_deviceUnion[str, torch.device]
default: None
vram_limitfloat
default: None
namestr
default: ''
disk_mapDiskMap
default: None
kwargs

Methods

methload_from_disk(torch_dtype,device,copy_module = False)source

Overview

Public method on this type.

Parameters

torch_dtype
device
copy_module
default: False
methoffload_to_disk(model: torch.nn.Module)source

Overview

Public method on this type.

Parameters

modeltorch.nn.Module
methcast_to(module,dtype,device)source

Overview

Public method on this type.

Parameters

module
dtype
device
class BaseMemory(capacity = None, **kwargs)
worldfoundry.core.memory.BaseMemoryfrom worldfoundry.core.memory import BaseMemory
source

Overview

Generic multimodal memory template for VLM and generative tasks. Belongs to Core acceleration and memory (caches, offload, VRAM, kernels).

Source docstring

Generic multimodal memory template for VLM and generative tasks.

Subclasses implement the five-stage pipeline:

Command methods (mutate state): - `record(data, ...) — ingest raw interaction data. - manage()` — evict, merge, or consolidate memories.

Query methods (read state): - `select(context_query, ...) — retrieve relevant snippets. - compress(memory_items, ...) — distill selected memories. - process(refined_data, ...)` — adapt memories to model input formats.

Parameters

capacity
default: None
kwargs

Methods

propstorage -> list[dict[str, Any]]source

Overview

Public property on this type.

Parameters

self

Returns: list[dict[str, Any]]

methstorage(records: Iterable[Mapping[str, Any]]) -> Nonesource

Overview

Public method on this type.

Parameters

recordsIterable[Mapping[str, Any]]

Returns: None

methcheck_template(**kwargs)source

Overview

Return required record keys and supported content types.

Parameters

kwargs
methappend_record(content: Any,kind: str = 'other',timestamp: int | float | None = None,metadata: Mapping[str, Any] | None = None) -> dict[str, Any]source

Overview

Public method on this type.

Parameters

contentAny
kindstr
default: 'other'
timestampint | float | None
default: None
metadataMapping[str, Any] | None
default: None

Returns: dict[str, Any]

methlatest_record(prefer_type: str | None = None,metadata: Mapping[str, Any] | None = None) -> dict[str, Any] | Nonesource

Overview

Public method on this type.

Parameters

prefer_typestr | None
default: None
metadataMapping[str, Any] | None
default: None

Returns: dict[str, Any] | None

methreset_records() -> Nonesource

Overview

Public method on this type.

Returns: None

methrecord(data,metadata = None,**kwargs)source

Overview

Ingest raw interaction data and assign metadata tags.

Parameters

data
metadata
default: None
kwargs
methselect(context_query, **kwargs)source

Overview

Retrieve memory snippets relevant to the current task context.

Parameters

context_query
kwargs
methcompress(_memory_items, **kwargs)source

Overview

Distill selected memories to reduce dimensionality or token count.

Parameters

_memory_items
kwargs
methprocess(_refined_data,_target_format = 'kv_cache',**kwargs)source

Overview

Convert refined memories into a model-ready format such as KV cache.

Parameters

_refined_data
_target_format
default: 'kv_cache'
kwargs
methmanage(**kwargs)source

Overview

Maintain memory lifecycle: eviction, merging, and STM→LTM transfer.

Parameters

kwargs
class DiskMap(path,device,torch_dtype = None,state_dict_converter = None,buffer_size = 10 ** 9)
worldfoundry.core.DiskMapfrom worldfoundry.core import DiskMap
source

Overview

Lazy mapping from checkpoint parameter names to materialized tensors. Belongs to Core acceleration and memory (caches, offload, VRAM, kernels).

Source docstring

Lazy mapping from checkpoint parameter names to materialized tensors.

Safetensors files remain memory-mapped and individual weights are loaded on lookup. PyTorch binary checkpoints use an in-memory compatibility loader. The mapping may apply a state-dict converter and periodically reopen files after `buffer_size` tensor elements have been materialized.

Parameters

path
Checkpoint path or list of paths.
device
Device on which fetched tensors are materialized.
torch_dtype
Optional dtype conversion applied on lookup.default: None
state_dict_converter
Optional callable that remaps public model keys to keys stored in the checkpoint.default: None
buffer_size
Number of fetched tensor elements after which file handles are refreshed.default: 10 ** 9

Methods

methflush_files()source

Overview

Open or refresh checkpoint readers and reset the materialized-element count.

methfetch_rename_dict(state_dict_converter)source

Overview

Build the optional model-key to checkpoint-key mapping.

Parameters

state_dict_converter
def enable_layerwise_cpu_offload(model: nn.Module,layer_container: str | None = None,device: torch.device | str | None = None,pin_memory: bool = True) -> LayerwiseOffloadHandle
worldfoundry.core.enable_layerwise_cpu_offloadfrom worldfoundry.core import enable_layerwise_cpu_offload
source

Overview

Attach layerwise CPU offload hooks to the first or named `ModuleList. Belongs to Core acceleration and memory (caches, offload, VRAM, kernels). Annotated return type: LayerwiseOffloadHandle`.

Source docstring

Attach layerwise CPU offload hooks to the first or named `ModuleList`.

The helper keeps parameters on CPU and moves one layer at a time to CUDA. The next layer is prefetched on a separate CUDA stream while the current layer runs. It is intentionally opt-in and returns a disabled handle on non-CUDA systems.

Parameters

modelnn.Module
layer_containerstr | None
default: None
devicetorch.device | str | None
default: None
pin_memorybool
default: True

Returns: LayerwiseOffloadHandle

def enable_vram_management(model: torch.nn.Module,module_map: dict,vram_config: dict,vram_limit = None,disk_map = None,max_num_param = None,overflow_vram_config: dict | None = None,**kwargs)
worldfoundry.core.vram.enable_vram_managementfrom worldfoundry.core.vram import enable_vram_management
source

Overview

Install staged VRAM management on a model and return that model. Belongs to Core acceleration and memory (caches, offload, VRAM, kernels).

Parameters

modeltorch.nn.Module
Model to wrap. It may be replaced when its own class appears in `module_map`; otherwise matching descendants are mutated.
module_mapdict
Mapping from original module classes to wrapper classes, for example `{torch.nn.Linear: AutoWrappedLinear}`.
vram_configdict
Dictionary containing offload, onload, preparing, and computation dtype/device pairs.
vram_limit
Optional used-memory limit in GiB.default: None
disk_map
Lazy checkpoint mapping used by disk-backed wrappers.default: None
max_num_param
Optional parameter budget for the primary policy.default: None
overflow_vram_configdict | None
Placement used after `max_num_param`.default: None
kwargs
Extra wrapper constructor arguments.

Notes

`module_map` order matters when source classes overlap. Configure complete placement keys before calling; this function changes module identity and is normally run once during model construction.

def enable_vram_management_recursively(model: torch.nn.Module,module_map: dict,vram_config: dict,vram_limit = None,name_prefix = '',disk_map = None,max_num_param = None,overflow_vram_config: dict | None = None,total_num_param = 0,**kwargs)
worldfoundry.core.vram.enable_vram_management_recursivelyfrom worldfoundry.core.vram import enable_vram_management_recursively
source

Overview

Replace matching descendants with lifecycle-aware wrapper modules. Belongs to Core acceleration and memory (caches, offload, VRAM, kernels).

Parameters

modeltorch.nn.Module
Root module mutated in place.
module_mapdict
Mapping from source module classes to wrapper classes.
vram_configdict
Default offload/onload/preparing/computation placements.
vram_limit
Optional used-memory limit in GiB forwarded to wrappers.default: None
name_prefix
Prefix used when resolving checkpoint keys through a `DiskMap`.default: ''
disk_map
Optional lazy checkpoint mapping for disk-backed wrappers.default: None
max_num_param
Parameter budget that switches later modules to `overflow_vram_config`.default: None
overflow_vram_configdict | None
Placement policy used after the parameter budget.default: None
total_num_param
Running parameter count for recursive calls.default: 0
kwargs
Extra wrapper constructor arguments.

Notes

The function mutates child modules. Most callers should use `enable_vram_management so the root itself is handled correctly and the vram_management_enabled` marker is installed.

def fill_vram_config(model, vram_config)
worldfoundry.core.vram.fill_vram_configfrom worldfoundry.core.vram import fill_vram_config
source

Overview

Collapse a placement policy when the root module is wrapped as one unit. Belongs to Core acceleration and memory (caches, offload, VRAM, kernels).

Parameters

model
vram_config
class FixedStepCache(skip_steps: Iterable[int] = (),delta_scale: float = 0.0,dense_first: int = 1,dense_last: int = 1,total_steps: int | None = None)
worldfoundry.core.acceleration.FixedStepCachefrom worldfoundry.core.acceleration import FixedStepCache
source

Overview

Reuse a previous denoiser output on an explicit set of steps. Belongs to Core acceleration and memory (caches, offload, VRAM, kernels).

Source docstring

Reuse a previous denoiser output on an explicit set of steps.

`delta_scale=0` replays the last dense output. A non-zero value performs first-order extrapolation from the two latest dense outputs.

Parameters

skip_stepsIterable[int]
default: ()
delta_scalefloat
default: 0.0
dense_firstint
default: 1
dense_lastint
default: 1
total_stepsint | None
default: None

Methods

methreset() -> Nonesource

Overview

Public method on this type.

Returns: None

methrun(step: int,compute: Callable[[], T],total_steps: int | None = None) -> Tsource

Overview

Compute or replay one step output.

Parameters

stepint
computeCallable[[], T]
total_stepsint | None
default: None

Returns: T

def init_weights_on_device(device = torch.device('meta'), include_buffers: bool = False)
worldfoundry.core.init_weights_on_devicefrom worldfoundry.core import init_weights_on_device
source

Overview

Temporarily redirect newly registered module weights to one device. Belongs to Core acceleration and memory (caches, offload, VRAM, kernels).

Parameters

device
Destination for parameters created inside the context. The default `meta` device skips real allocation and initialization.default: torch.device('meta')
include_buffersbool
Also redirect registered buffers and common tensor constructors. Leave disabled unless module construction allocates large persistent buffers.default: False

Notes

Global PyTorch hooks are restored in `finally`. This context is not intended to overlap across threads.

def layer_norm_scale_shift(x: torch.Tensor,scale: torch.Tensor,shift: torch.Tensor,eps: float = 1e-06,upcast: bool = False) -> torch.Tensor
worldfoundry.core.kernels.layer_norm_scale_shiftfrom worldfoundry.core.kernels import layer_norm_scale_shift
source

Overview

Fuse affine-free LayerNorm with AdaLN scale and shift. Belongs to Core acceleration and memory (caches, offload, VRAM, kernels). Annotated return type: torch.Tensor.

Parameters

xtorch.Tensor
scaletorch.Tensor
shifttorch.Tensor
epsfloat
default: 1e-06
upcastbool
default: False

Returns: torch.Tensor

def layerwise_offload_mutation_scope(module: nn.Module) -> Iterator[None]
worldfoundry.core.layerwise_offload_mutation_scopefrom worldfoundry.core import layerwise_offload_mutation_scope
source

Overview

Temporarily materialize offloaded parameters for in-place mutations. Belongs to Core acceleration and memory (caches, offload, VRAM, kernels). Annotated return type: Iterator[None].

Parameters

modulenn.Module

Returns: Iterator[None]

class LayerwiseOffloadHandle(enabled: bool,layer_count: int,reason: str = '')
worldfoundry.core.LayerwiseOffloadHandlefrom worldfoundry.core import LayerwiseOffloadHandle
source

Overview

Handle returned by `enable_layerwise_cpu_offload`. Belongs to Core acceleration and memory (caches, offload, VRAM, kernels).

Attributes

enabledbool
layer_countint
reasonstr
default: ''
class MemoryStore(capacity: int | None = None,records: Iterable[Mapping[str, Any] | MemoryRecord] = ())
worldfoundry.core.memory.MemoryStorefrom worldfoundry.core.memory import MemoryStore
source

Overview

Bounded in-process memory store used by all concrete WorldFoundry memories. Belongs to Core acceleration and memory (caches, offload, VRAM, kernels).

Parameters

capacityint | None
default: None
recordsIterable[Mapping[str, Any] | MemoryRecord]
default: ()

Methods

methappend(content: Any,kind: str = 'other',timestamp: int | float | None = None,metadata: Mapping[str, Any] | None = None,score: float | None = None) -> dict[str, Any]source

Overview

Public method on this type.

Parameters

contentAny
kindstr
default: 'other'
timestampint | float | None
default: None
metadataMapping[str, Any] | None
default: None
scorefloat | None
default: None

Returns: dict[str, Any]

methappend_record(record: Mapping[str, Any] | MemoryRecord) -> dict[str, Any]source

Overview

Public method on this type.

Parameters

recordMapping[str, Any] | MemoryRecord

Returns: dict[str, Any]

methlatest(prefer_type: str | None = None,metadata: Mapping[str, Any] | None = None) -> dict[str, Any] | Nonesource

Overview

Public method on this type.

Parameters

prefer_typestr | None
default: None
metadataMapping[str, Any] | None
default: None

Returns: dict[str, Any] | None

methselect(query: MemoryQuery | None = None, **overrides: Any) -> MemorySelectionsource

Overview

Rank stored records and return the top-*query.top_k* matches.

Parameters

queryMemoryQuery | None
default: None
overridesAny

Returns: MemorySelection

methevict() -> Nonesource

Overview

Public method on this type.

Returns: None

methreset() -> Nonesource

Overview

Public method on this type.

Returns: None

methreplace(records: Iterable[Mapping[str, Any] | MemoryRecord]) -> Nonesource

Overview

Public method on this type.

Parameters

recordsIterable[Mapping[str, Any] | MemoryRecord]

Returns: None

def prune_tokens(hidden_states: torch.Tensor,keep_ratio: float,method: TokenScoreMethod | str = 'feat_norm',seq_dim: int = 1,start: int = 0,end: int | None = None,previous_velocity: torch.Tensor | None = None) -> tuple[torch.Tensor, TokenPruneState]
worldfoundry.core.acceleration.prune_tokensfrom worldfoundry.core.acceleration import prune_tokens
source

Overview

Gather the selected part of a token segment before expensive blocks. Belongs to Core acceleration and memory (caches, offload, VRAM, kernels). Annotated return type: tuple[torch.Tensor, TokenPruneState].

Parameters

hidden_statestorch.Tensor
keep_ratiofloat
methodTokenScoreMethod | str
default: 'feat_norm'
seq_dimint
default: 1
startint
default: 0
endint | None
default: None
previous_velocitytorch.Tensor | None
default: None

Returns: tuple[torch.Tensor, TokenPruneState]

def residual_gate_add(residual: torch.Tensor,update: torch.Tensor,gate: torch.Tensor) -> torch.Tensor
worldfoundry.core.kernels.residual_gate_addfrom worldfoundry.core.kernels import residual_gate_add
source

Overview

Return `residual + update * gate with broadcast-aware fusion. Belongs to Core acceleration and memory (caches, offload, VRAM, kernels). Annotated return type: torch.Tensor`.

Parameters

residualtorch.Tensor
updatetorch.Tensor
gatetorch.Tensor

Returns: torch.Tensor

def restore_tokens(processed: torch.Tensor,state: TokenPruneState,compensation: torch.Tensor | None = None) -> torch.Tensor
worldfoundry.core.acceleration.restore_tokensfrom worldfoundry.core.acceleration import restore_tokens
source

Overview

Scatter processed tokens and restore dropped tokens from prior state. Belongs to Core acceleration and memory (caches, offload, VRAM, kernels). Annotated return type: torch.Tensor.

Source docstring

Scatter processed tokens and restore dropped tokens from prior state.

`compensation` is the previous full segment in the original tensor layout. If omitted, dropped tokens are zero-filled.

Parameters

processedtorch.Tensor
compensationtorch.Tensor | None
default: None

Returns: torch.Tensor

def routed_swiglu_moe(hidden_states: torch.Tensor,routing_weights: torch.Tensor,selected_experts: torch.Tensor,gate_weight: torch.Tensor,up_weight: torch.Tensor,down_weight: torch.Tensor,workspace: dict[str, torch.Tensor] | Callable[[], dict[str, torch.Tensor]] | None = None,allow_triton: bool | None = None) -> torch.Tensor
worldfoundry.core.routed_swiglu_moefrom worldfoundry.core import routed_swiglu_moe
source

Overview

Dispatch packed SwiGLU MoE to Triton or the portable PyTorch reference. Belongs to Core acceleration and memory (caches, offload, VRAM, kernels). Annotated return type: torch.Tensor.

Parameters

hidden_statestorch.Tensor
routing_weightstorch.Tensor
selected_expertstorch.Tensor
gate_weighttorch.Tensor
up_weighttorch.Tensor
down_weighttorch.Tensor
workspacedict[str, torch.Tensor] | Callable[[], dict[str, torch.Tensor]] | None
default: None
allow_tritonbool | None
default: None

Returns: torch.Tensor

def routed_swiglu_moe_pytorch(hidden_states: torch.Tensor,routing_weights: torch.Tensor,selected_experts: torch.Tensor,gate_weight: torch.Tensor,up_weight: torch.Tensor,down_weight: torch.Tensor) -> torch.Tensor
worldfoundry.core.routed_swiglu_moe_pytorchfrom worldfoundry.core import routed_swiglu_moe_pytorch
source

Overview

Evaluate a token-routed SwiGLU MoE using portable PyTorch operators. Belongs to Core acceleration and memory (caches, offload, VRAM, kernels). Annotated return type: torch.Tensor.

Parameters

hidden_statestorch.Tensor
Flattened token activations with shape `[T, D]`.
routing_weightstorch.Tensor
Per-route weights with shape `[T, K]`.
selected_expertstorch.Tensor
Expert indices with shape `[T, K]`.
gate_weighttorch.Tensor
Packed gate weights with shape `[E, I, D]`.
up_weighttorch.Tensor
Packed up-projection weights with shape `[E, I, D]`.
down_weighttorch.Tensor
Packed down-projection weights with shape `[E, D, I]`. Only tokens routed to an expert are evaluated for that expert. This keeps the fallback substantially smaller than materializing every expert output, while retaining autograd support and working on CPU, CUDA, and other PyTorch devices.

Returns: torch.Tensor

def select_token_indices(hidden_states: torch.Tensor,keep_ratio: float,method: TokenScoreMethod | str = 'feat_norm',seq_dim: int = 1,previous_velocity: torch.Tensor | None = None,random_seed: int = 42) -> torch.Tensor
worldfoundry.core.acceleration.select_token_indicesfrom worldfoundry.core.acceleration import select_token_indices
source

Overview

Return ascending indices for tokens retained by a pruning policy. Belongs to Core acceleration and memory (caches, offload, VRAM, kernels). Annotated return type: torch.Tensor.

Parameters

hidden_statestorch.Tensor
keep_ratiofloat
methodTokenScoreMethod | str
default: 'feat_norm'
seq_dimint
default: 1
previous_velocitytorch.Tensor | None
default: None
random_seedint
default: 42

Returns: torch.Tensor

def skip_model_initialization(device = torch.device('meta'))
worldfoundry.core.skip_model_initializationfrom worldfoundry.core import skip_model_initialization
source

Overview

Return an allocation-skipping context for constructing checkpoint-backed models. Belongs to Core acceleration and memory (caches, offload, VRAM, kernels).

Parameters

device
default: torch.device('meta')
class TokenPruner(keep_ratio: float, , method: TokenScoreMethod | str = 'feat_norm')
worldfoundry.core.acceleration.TokenPrunerfrom worldfoundry.core.acceleration import TokenPruner
source

Overview

Stateful previous-step compensation for repeated denoising calls. Belongs to Core acceleration and memory (caches, offload, VRAM, kernels).

Parameters

keep_ratiofloat
methodTokenScoreMethod | str
default: 'feat_norm'

Methods

methreset(key: object | None = None) -> Nonesource

Overview

Public method on this type.

Parameters

keyobject | None
default: None

Returns: None

methprune(hidden_states: torch.Tensor,key: object = 'default',seq_dim: int = 1,start: int = 0,end: int | None = None) -> tuple[torch.Tensor, TokenPruneState | None]source

Overview

Prune after a dense seed call; first use only records compensation.

Parameters

hidden_statestorch.Tensor
keyobject
default: 'default'
seq_dimint
default: 1
startint
default: 0
endint | None
default: None

Returns: tuple[torch.Tensor, TokenPruneState | None]

methrestore(processed: torch.Tensor,state: TokenPruneState | None,key: object = 'default') -> torch.Tensorsource

Overview

Public method on this type.

Parameters

processedtorch.Tensor
stateTokenPruneState | None
keyobject
default: 'default'

Returns: torch.Tensor

class TokenPruneState(indices: torch.Tensor,start: int,end: int,full_length: int,seq_dim: int)
worldfoundry.core.acceleration.TokenPruneStatefrom worldfoundry.core.acceleration import TokenPruneState
source

Overview

Metadata needed to reconstruct a pruned token segment. Belongs to Core acceleration and memory (caches, offload, VRAM, kernels).

Attributes

indicestorch.Tensor
startint
endint
full_lengthint
seq_dimint

Methods

propkept_length -> intsource

Overview

Public property on this type.

Parameters

self

Returns: int