Pre-release

This release is a pre-release and may not be stable for production use.

MLBricks Kit 1.0.0b1

MLBricks is a modular machine-learning library for building efficient language and vision models from reusable components, with PyTorch reference paths and optional native C++/CUDA acceleration.

Core components

ESA — Entangled State Attention for recurrent/state-based sequence processing.
Bolt / BoltAttention — optimized causal attention component.
BoltModel / BoltConfig — ready-made Bolt language-model interface.
FFNBricks — StateAwareFFN, VirtualStateAwareFFN, and MicroVirtualFFN.
ResController — adaptive residual control.
SOUP — state-oriented sequence processing with layerwise ESA/Bolt mixer selection.
ElasticBit — adaptive 4–32-bit CUDA runtime plus PyTorch-compatible quantization helpers.
VESA — ESA-based vision models.
VisualBolt — Bolt-based vision models through VisionBolt.
Bricks / Brick — heterogeneous model construction from MLBricks components.
Execution planner — automatic backend and execution-route planning.
Native acceleration — optional C++/CUDA kernels where supported.

The PyPI distribution is mlbricks-kit; the public Python import package remains mlbricks.

Version

import mlbricks
print(mlbricks.__version__)
# 1.0.0b1

Installation

Install mlbricks-kit from PyPI; use import mlbricks in Python.

From PyPI:

pip install mlbricks-kit

For development or installation directly from the repository root:

pip install -e .

For a CPU-only native build:

MLBRICKS_FORCE_CPU=1 pip install -e .

Quick start

import torch
from mlbricks import ESA, Bolt, StateAwareFFN, ResController

x = torch.randn(2, 128, 384)

esa = ESA(embd=384, head=6)
bolt = Bolt(d_model=384, num_heads=6, latent_dim=32)
ffn = StateAwareFFN(d_model=384)
residual = ResController(update_ratio=0.18)

MLBricks components use backend="auto" by default. The public backend choices are:

auto — qualify each element independently (for example ESA, Bolt, SAFFN, a vision scan, or ElasticLinear). PyTorch is used as the one-time correctness reference; native must match it before both routes are benchmarked. The fastest valid route is then frozen for that element. Composite models can mix routes.
native — require a supported MLBricks native implementation.
pytorch — force the PyTorch/reference path.

Example:

bolt.set_backend("native")
bolt.set_backend("pytorch")
bolt.set_backend("auto")

Unified model lifecycle

Saving, loading, training, resume, inference, compilation, and quantization are package-level MLBricks APIs. They are not owned by ESA or any other individual architecture.

import mlbricks as mlb

mlb.save(model, "my_model")
model = mlb.load("my_model", device="auto")
info = mlb.inspect("my_model")

The same calls work for ESA models, Bolt models, SOUP, VESA, Bricks, and mixed models built from multiple MLBricks components.

Train with the generic trainer:

trainer = mlb.Trainer(
    model,
    optimizer="adamw",
    lr=3e-4,
    checkpoint_dir="checkpoints",
    save_every=1000,
)

trainer.fit(
    train_loader,
    steps=10_000,
    val_loader=val_loader,
    validate_every=500,
)

Or use the one-call convenience API:

trainer = mlb.train(
    model,
    train_loader,
    steps=10_000,
    optimizer="adamw",
    lr=3e-4,
)

Resume an interrupted run with model weights, optimizer state, training step, scaler state, scheduler state (when supplied), and RNG state:

trainer = mlb.Trainer.resume("checkpoints/last", device="auto")
trainer.fit(train_loader, steps=20_000)

Unified inference and optimization helpers are also available:

y = mlb.predict(model, x)
text_or_tokens = mlb.generate(model, prompt, max_new_tokens=128)
model = mlb.compile(model)
model = mlb.quantize(model, method="elasticbit", bits=4)

ESAModel.save(), ESAModel.load(), and mlbricks.esa.Trainer are no longer public APIs.

ESA

from mlbricks import ESA

esa = ESA(
    embd=384,
    head=6,
)

For the ready-made ESA language-model architecture:

from mlbricks import ESAModel, ESAModelConfig

Bolt

Bolt and BoltAttention are the public attention names in MLBricks Kit 1.0.0b1.

CUDA FP16 Bolt now uses a compound Stage-1 execution path when the native extension is available: one packed Q/U/G GEMM followed by one fused gate/RMS postprocess emits only Q, C, and FP32 rho. Training keeps the same parameters/equations and uses normalized-key PyTorch SDPA; use_sdpa=False remains the explicit reference route.

from mlbricks import Bolt, BoltAttention

bolt = Bolt(
    d_model=384,
    num_heads=6,
    latent_dim=32,
)

You can also import directly from the component package:

from mlbricks.bolt import Bolt, BoltAttention

Bolt language model

from mlbricks import BoltModel

model = BoltModel(
    vocab_size=50_257,
    context=2048,
    layers=12,
    dim=768,
    heads=12,
    latent_dim=32,
    position="rope",
    ffn="saffn",
    residual="rescontroller",
    norm="rmsnorm",
)

FFNBricks

from mlbricks import StateAwareFFN, VirtualStateAwareFFN, MicroVirtualFFN

ffn = StateAwareFFN(
    d_model=384,
    state_dim=128,
    depth_embedding_dim=32,
    layer_index=0,
    total_layers=6,
)

ResController

from mlbricks import ResController

controller = ResController(update_ratio=0.18)

SOUP

from mlbricks import SOUP

model = SOUP(
    dim=512,
    width=[1116, 1116],
    depth=2,
    mixer=["esa", "bolt"],
    ffn=["saffn", "saffn"],
    backend="auto",
)

SOUP keeps backend="auto" element-wise. It does not force one backend for the whole SOUP model: an ESA layer can freeze to native while a Bolt layer freezes to PyTorch, and backend-aware FFN/residual elements make their own one-time decisions. Each native candidate must first match its PyTorch reference output; only parity-qualified routes are benchmarked for speed. Per-layer mixer and FFN choices remain supported.

ElasticBit

from mlbricks import (
    ElasticBit,
    ElasticBitConfig,
    ElasticLinear,
    ElasticEmbedding,
    quantize_tensor,
    dequantize_tensor,
)

ElasticBit now also exposes the standalone 0.2 adaptive 4–32-bit CUDA API through the MLBricks namespace:

from mlbricks import ElasticBit

analysis = ElasticBit.bitsAnaliser(weights, calibration, threshold=0.01)
matrix = ElasticBit.RuntimeMatrix(weights, analysis["selected_bits"], "compact")
y = matrix.forward(x)

The existing ElasticLinear, tensor quantization, and module-conversion helpers remain available as a PyTorch-compatible fallback surface.

VESA

from mlbricks import Vesa

model = Vesa(
    image_size=224,
    patch_size=16,
    dim=384,
    depth=12,
    engine="Serpentine",
    scan="cross",
    position=None,
    backend="auto",
)

Supported vision engine names include:

Serpentine
ViT / VisionTransformer
CNN
Diffusion
AR

VisualBolt

VisualBolt is exposed through the VisionBolt API:

from mlbricks import VisionBolt

model = VisionBolt(
    image_size=224,
    patch_size=16,
    dim=384,
    depth=12,
    heads=6,
    latent_dim=32,
    engine="Serpentine",
    scan="cross",
    position=None,
    backend="auto",
)

VESA and VisualBolt share the same high-level vision-engine choices while keeping their respective ESA and Bolt mixers.

Build a custom model with Bricks

import torch
from mlbricks import (
    Bricks,
    Brick,
    ESA,
    Bolt,
    Attention,
    StateAwareFFN,
    FFN,
    ResController,
)

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

model = Bricks(
    vocab_size=50_257,
    dim=384,
    context=2048,
    position="sinusoidal",
    layers=[
        Brick(
            # In a composed model, let Bricks own placement and move the parent once.
            mixer=ESA(embd=384, head=6, device=None),
            ffn=StateAwareFFN(
                d_model=384,
                state_dim=128,
                depth_embedding_dim=32,
                layer_index=0,
                total_layers=2,
            ),
            residual=ResController(update_ratio=0.18),
            dim=384,
        ),
        Brick(
            mixer=Bolt(384, 6, latent_dim=32),
            ffn=FFN(384, 1536, activation="gelu"),
            dim=384,
        ),
    ],
).to(device)

# Place inputs on the same device as the parent model.
input_ids = torch.randint(0, 50_257, (1, 32), device=device)
logits = model(input_ids)

ESA(device="auto") is convenient when ESA is used by itself. Inside Bricks or another parent nn.Module, prefer ESA(..., device=None) and move the complete parent model with .to(device) so embeddings, residual paths, mixers, FFNs, and inputs remain on one device.

Native vision runtime

VESA and VisualBolt can use the shared MLBricks vision runtime when the compiled backend is available. The public backend contract remains auto | native | pytorch.

Native-supported operations include selected scan/reorder paths, positional operations, normalization/dataflow helpers, recurrent ESA updates, and Bolt sequence paths. Operations already efficiently provided by PyTorch libraries such as GEMM, convolution, and standard tensor operations continue to use the corresponding PyTorch/ATen vendor implementations.

API documentation

See API.md for the public API reference and examples/ for runnable examples.

Component licenses

MLBricks Kit 1.0.0b1 contains component-specific license notices in addition to the repository-level licensing documents:

ESA — mlbricks/esa/LICENSE_ESA.txt
Bolt — mlbricks/bolt/LICENSE_BOLT.txt
ElasticBit — mlbricks/elasticbit/LICENSE_ELASTICBIT.txt
VESA — mlbricks/vesa/LICENSE_VESA.txt
VisualBolt — mlbricks/LICENSE_VISUALBOLT.txt
FFNBricks — mlbricks/ffnbrick/LICENSE_FFNBRICK.txt
ResController — mlbricks/residualbrick/LICENSE_RESIDUALBRICK.txt
SOUP — mlbricks/soup/LICENSE_SOUP.txt

Also see the repository-level LICENSE.md, LICENSING_NOTICE.md, and COMMERCIAL_LICENSE.md.

Licensing

MLBricks Kit 1.0.0b1 is source-available software distributed under the PolyForm Noncommercial License 1.0.0.

The public license permits noncommercial use, including personal use, education, academic study, noncommercial research, experimentation, benchmarking, and hobby projects, subject to the complete PolyForm license terms.

Commercial use is not granted by the public license and requires a separate written commercial license. This includes commercial products, paid software, SaaS/cloud services, enterprise deployment, paid client work, and other commercial or revenue-generating use where applicable.

Commercial licensing inquiries: licensing@mlbricks.io

See LICENSE.md, LICENSING_NOTICE.md, and COMMERCIAL_LICENSE.md.

Release

MLBricks Kit 1.0.0b1 is the beta release line for the package and its current component APIs.

Training compilation

The default remains training_compile_mode="default".

Bolt native build control

To install MLBricks without compiling the optional Bolt native extension, set:

MLBRICKS_BUILD_BOLT_NATIVE=0

Bolt remains available through its PyTorch/fallback implementation; this flag only disables native extension compilation during installation.

SOUP component license

The bundled SOUP component license is shipped at mlbricks/soup/LICENSE_SOUP.txt.

Versioning note

MLBricks Kit is released as 1.0.0b1. Experimental SOUP retains its independent component version 0.1.0a3.

Beta native wheel distribution

MLBricks 1.0.0b1 is distributed with prebuilt native wheels for supported Linux, Windows, and macOS targets plus a py3-none-any fallback wheel. On a matching platform, pip selects the platform-specific native wheel; otherwise it installs the portable PyTorch fallback instead of compiling C++/CUDA locally.

For release builds, GitHub CI compiles CUDA wheels with a fat architecture list covering NVIDIA compute capabilities 7.0, 7.5, 8.0, 8.6, 8.9, 9.0, 10.0, and 12.0+PTX. The beta native ABI is validated against PyTorch 2.10.x; the package therefore requires torch>=2.10,<2.11 until the native bindings migrate to the PyTorch stable ABI.

Source installations default native compilation off. Developers who intentionally want to compile from source can set the relevant MLBRICKS_BUILD_*_NATIVE=1 flags. The corresponding =0 values explicitly disable each extension.

Beta native platform support

For 1.0.0b1, official prebuilt native wheels target Linux x86_64 with CUDA 12.8, Windows x86_64 with CUDA 12.8, and macOS Apple Silicon (arm64) for CPU-native acceleration. Intel macOS is not included because the PyTorch 2.10 binary line used by this beta does not provide the required current x86_64 macOS wheels.

Windows CUDA-native users must use the CUDA 12.8 PyTorch build from the official PyTorch cu128 index. The MLBricks wheel itself remains precompiled, so MLBricks does not compile native code on the user's machine. If CUDA-native prerequisites are unavailable, MLBricks' PyTorch implementation remains the portable fallback.

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

mlbricks_kit-1.0.0b1.tar.gz (291.8 kB view details)

Uploaded Sep 4, 2026 Source

Built Distributions

If you're not sure about the file name format, learn more about wheel file names.

The dropdown lists show the available interpreters, ABIs, and platforms. Enable javascript to be able to filter the list of wheel files.

mlbricks_kit-1.0.0b1-py3-none-any.whl (304.2 kB view details)

Uploaded Sep 4, 2026 Python 3

mlbricks_kit-1.0.0b1-cp313-cp313-win_amd64.whl (4.0 MB view details)

Uploaded Sep 4, 2026 CPython 3.13Windows x86-64

mlbricks_kit-1.0.0b1-cp313-cp313-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl (26.9 MB view details)

Uploaded Sep 4, 2026 CPython 3.13manylinux: glibc 2.27+ x86-64manylinux: glibc 2.28+ x86-64

mlbricks_kit-1.0.0b1-cp313-cp313-macosx_11_0_arm64.whl (706.4 kB view details)

Uploaded Sep 4, 2026 CPython 3.13macOS 11.0+ ARM64

mlbricks_kit-1.0.0b1-cp312-cp312-win_amd64.whl (4.0 MB view details)

Uploaded Sep 4, 2026 CPython 3.12Windows x86-64

mlbricks_kit-1.0.0b1-cp312-cp312-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl (26.9 MB view details)

Uploaded Sep 4, 2026 CPython 3.12manylinux: glibc 2.27+ x86-64manylinux: glibc 2.28+ x86-64

mlbricks_kit-1.0.0b1-cp312-cp312-macosx_11_0_arm64.whl (706.0 kB view details)

Uploaded Sep 4, 2026 CPython 3.12macOS 11.0+ ARM64

mlbricks_kit-1.0.0b1-cp311-cp311-win_amd64.whl (4.0 MB view details)

Uploaded Sep 4, 2026 CPython 3.11Windows x86-64

mlbricks_kit-1.0.0b1-cp311-cp311-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl (26.9 MB view details)

Uploaded Sep 4, 2026 CPython 3.11manylinux: glibc 2.27+ x86-64manylinux: glibc 2.28+ x86-64

mlbricks_kit-1.0.0b1-cp311-cp311-macosx_11_0_arm64.whl (699.9 kB view details)

Uploaded Sep 4, 2026 CPython 3.11macOS 11.0+ ARM64

mlbricks_kit-1.0.0b1-cp310-cp310-win_amd64.whl (4.0 MB view details)

Uploaded Sep 4, 2026 CPython 3.10Windows x86-64

mlbricks_kit-1.0.0b1-cp310-cp310-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl (26.8 MB view details)

Uploaded Sep 4, 2026 CPython 3.10manylinux: glibc 2.27+ x86-64manylinux: glibc 2.28+ x86-64

mlbricks_kit-1.0.0b1-cp310-cp310-macosx_11_0_arm64.whl (693.7 kB view details)

Uploaded Sep 4, 2026 CPython 3.10macOS 11.0+ ARM64

File details

Details for the file mlbricks_kit-1.0.0b1.tar.gz.

File metadata

Download URL: mlbricks_kit-1.0.0b1.tar.gz
Upload date: Sep 4, 2026
Size: 291.8 kB
Tags: Source
Uploaded using Trusted Publishing? Yes
Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for mlbricks_kit-1.0.0b1.tar.gz
Algorithm	Hash digest
SHA256	`28d4113f98556e353466835319072dd5c61c56cc74f9e31b487b0ce56a248ada`
MD5	`2e625cd9ee5b74260fb4a76fadf871f4`
BLAKE2b-256	`de53bf8f9ce2df4a967227a2075f1235e383f7d741a9825432fd287174e8aa54`

See more details on using hashes here.

Provenance

The following attestation bundles were made for mlbricks_kit-1.0.0b1.tar.gz:

Publisher: native-wheels-beta.yml on MlBricks/MLBricks

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Statement:
- Statement type: https://in-toto.io/Statement/v1
- Predicate type: https://docs.pypi.org/attestations/publish/v1
- Subject name: mlbricks_kit-1.0.0b1.tar.gz
- Subject digest: 28d4113f98556e353466835319072dd5c61c56cc74f9e31b487b0ce56a248ada
- Sigstore transparency entry: 2715383379
- Sigstore integration time: Sep 4, 2026
Source repository:
- Permalink: MlBricks/MLBricks@d86fccc6645640d98b7db33916ef2021564b5d0b
- Branch / Tag: refs/tags/v1.0.0b1
- Owner: https://github.com/MlBricks
- Access: public
Publication detail:
- Token Issuer: https://token.actions.githubusercontent.com
- Runner Environment: github-hosted
- Publication workflow: native-wheels-beta.yml@d86fccc6645640d98b7db33916ef2021564b5d0b
- Trigger Event: push

File details

Details for the file mlbricks_kit-1.0.0b1-py3-none-any.whl.

File metadata

Download URL: mlbricks_kit-1.0.0b1-py3-none-any.whl
Upload date: Sep 4, 2026
Size: 304.2 kB
Tags: Python 3
Uploaded using Trusted Publishing? Yes
Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for mlbricks_kit-1.0.0b1-py3-none-any.whl
Algorithm	Hash digest
SHA256	`95ca8f253d02c524ebcf8d7f442a0e517a3f76e916caa0a7aef582e7bff40427`
MD5	`6d0d3e7abd96db9cb7bd2c3a2bbf1fe9`
BLAKE2b-256	`bbf9ae72af7c4028c3771da00d1ca5d28a20d50e9318fa2ac3f51b2d274c78e1`

See more details on using hashes here.

Provenance

The following attestation bundles were made for mlbricks_kit-1.0.0b1-py3-none-any.whl:

Publisher: native-wheels-beta.yml on MlBricks/MLBricks

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Statement:
- Statement type: https://in-toto.io/Statement/v1
- Predicate type: https://docs.pypi.org/attestations/publish/v1
- Subject name: mlbricks_kit-1.0.0b1-py3-none-any.whl
- Subject digest: 95ca8f253d02c524ebcf8d7f442a0e517a3f76e916caa0a7aef582e7bff40427
- Sigstore transparency entry: 2715383678
- Sigstore integration time: Sep 4, 2026
Source repository:
- Permalink: MlBricks/MLBricks@d86fccc6645640d98b7db33916ef2021564b5d0b
- Branch / Tag: refs/tags/v1.0.0b1
- Owner: https://github.com/MlBricks
- Access: public
Publication detail:
- Token Issuer: https://token.actions.githubusercontent.com
- Runner Environment: github-hosted
- Publication workflow: native-wheels-beta.yml@d86fccc6645640d98b7db33916ef2021564b5d0b
- Trigger Event: push

File details

Details for the file mlbricks_kit-1.0.0b1-cp313-cp313-win_amd64.whl.

File metadata

Download URL: mlbricks_kit-1.0.0b1-cp313-cp313-win_amd64.whl
Upload date: Sep 4, 2026
Size: 4.0 MB
Tags: CPython 3.13, Windows x86-64
Uploaded using Trusted Publishing? Yes
Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for mlbricks_kit-1.0.0b1-cp313-cp313-win_amd64.whl
Algorithm	Hash digest
SHA256	`c9c70e77ccf3f967a9ca2c3e89d96225f0136f6cd1a5ed5e221f33e67ef115d8`
MD5	`40d6c948e0d2ee76a60747c7d1cb5b50`
BLAKE2b-256	`f830233ba1ef84b31aa929f7d6c045ecbb03252b246742159f8906d26a976e97`

See more details on using hashes here.

Provenance

The following attestation bundles were made for mlbricks_kit-1.0.0b1-cp313-cp313-win_amd64.whl:

Publisher: native-wheels-beta.yml on MlBricks/MLBricks

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Statement:
- Statement type: https://in-toto.io/Statement/v1
- Predicate type: https://docs.pypi.org/attestations/publish/v1
- Subject name: mlbricks_kit-1.0.0b1-cp313-cp313-win_amd64.whl
- Subject digest: c9c70e77ccf3f967a9ca2c3e89d96225f0136f6cd1a5ed5e221f33e67ef115d8
- Sigstore transparency entry: 2715383456
- Sigstore integration time: Sep 4, 2026
Source repository:
- Permalink: MlBricks/MLBricks@d86fccc6645640d98b7db33916ef2021564b5d0b
- Branch / Tag: refs/tags/v1.0.0b1
- Owner: https://github.com/MlBricks
- Access: public
Publication detail:
- Token Issuer: https://token.actions.githubusercontent.com
- Runner Environment: github-hosted
- Publication workflow: native-wheels-beta.yml@d86fccc6645640d98b7db33916ef2021564b5d0b
- Trigger Event: push

File details

Details for the file mlbricks_kit-1.0.0b1-cp313-cp313-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl.

File metadata

Download URL: mlbricks_kit-1.0.0b1-cp313-cp313-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl
Upload date: Sep 4, 2026
Size: 26.9 MB
Tags: CPython 3.13, manylinux: glibc 2.27+ x86-64, manylinux: glibc 2.28+ x86-64
Uploaded using Trusted Publishing? Yes
Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for mlbricks_kit-1.0.0b1-cp313-cp313-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl
Algorithm	Hash digest
SHA256	`678c0dcd249e1641f0a4ef85eda65dbacf1c254dc30b7017a498dd36f06c7133`
MD5	`c67658ce8925735cb116a26673f2e7dd`
BLAKE2b-256	`2a894687d7c596a2814a624dda4ad310099f5743a75ae27b027f2eb6dd6f7129`

See more details on using hashes here.

Provenance

The following attestation bundles were made for mlbricks_kit-1.0.0b1-cp313-cp313-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl:

Publisher: native-wheels-beta.yml on MlBricks/MLBricks

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Statement:
- Statement type: https://in-toto.io/Statement/v1
- Predicate type: https://docs.pypi.org/attestations/publish/v1
- Subject name: mlbricks_kit-1.0.0b1-cp313-cp313-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl
- Subject digest: 678c0dcd249e1641f0a4ef85eda65dbacf1c254dc30b7017a498dd36f06c7133
- Sigstore transparency entry: 2715383812
- Sigstore integration time: Sep 4, 2026
Source repository:
- Permalink: MlBricks/MLBricks@d86fccc6645640d98b7db33916ef2021564b5d0b
- Branch / Tag: refs/tags/v1.0.0b1
- Owner: https://github.com/MlBricks
- Access: public
Publication detail:
- Token Issuer: https://token.actions.githubusercontent.com
- Runner Environment: github-hosted
- Publication workflow: native-wheels-beta.yml@d86fccc6645640d98b7db33916ef2021564b5d0b
- Trigger Event: push

File details

Details for the file mlbricks_kit-1.0.0b1-cp313-cp313-macosx_11_0_arm64.whl.

File metadata

Download URL: mlbricks_kit-1.0.0b1-cp313-cp313-macosx_11_0_arm64.whl
Upload date: Sep 4, 2026
Size: 706.4 kB
Tags: CPython 3.13, macOS 11.0+ ARM64
Uploaded using Trusted Publishing? Yes
Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for mlbricks_kit-1.0.0b1-cp313-cp313-macosx_11_0_arm64.whl
Algorithm	Hash digest
SHA256	`54580804ef3a2713162bc451ccf7eca26d353cfea2a0ce66c02a42eabd051192`
MD5	`5287b4ebd216011d8eb3446f16cb2ce2`
BLAKE2b-256	`5aa5632a8f780ef3c50a21fd46b66a2671abdd425f499143d7a371272f73899b`

See more details on using hashes here.

Provenance

The following attestation bundles were made for mlbricks_kit-1.0.0b1-cp313-cp313-macosx_11_0_arm64.whl:

Publisher: native-wheels-beta.yml on MlBricks/MLBricks

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Statement:
- Statement type: https://in-toto.io/Statement/v1
- Predicate type: https://docs.pypi.org/attestations/publish/v1
- Subject name: mlbricks_kit-1.0.0b1-cp313-cp313-macosx_11_0_arm64.whl
- Subject digest: 54580804ef3a2713162bc451ccf7eca26d353cfea2a0ce66c02a42eabd051192
- Sigstore transparency entry: 2715383504
- Sigstore integration time: Sep 4, 2026
Source repository:
- Permalink: MlBricks/MLBricks@d86fccc6645640d98b7db33916ef2021564b5d0b
- Branch / Tag: refs/tags/v1.0.0b1
- Owner: https://github.com/MlBricks
- Access: public
Publication detail:
- Token Issuer: https://token.actions.githubusercontent.com
- Runner Environment: github-hosted
- Publication workflow: native-wheels-beta.yml@d86fccc6645640d98b7db33916ef2021564b5d0b
- Trigger Event: push

File details

Details for the file mlbricks_kit-1.0.0b1-cp312-cp312-win_amd64.whl.

File metadata

Download URL: mlbricks_kit-1.0.0b1-cp312-cp312-win_amd64.whl
Upload date: Sep 4, 2026
Size: 4.0 MB
Tags: CPython 3.12, Windows x86-64
Uploaded using Trusted Publishing? Yes
Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for mlbricks_kit-1.0.0b1-cp312-cp312-win_amd64.whl
Algorithm	Hash digest
SHA256	`b50ec1b86621c64f4fdc4901557ff8cbd27c2377c580e9b029115ea0aadfa8c6`
MD5	`8eb4665dfe20f4248ac885d7f8f679a0`
BLAKE2b-256	`cd79a7acc4a5edd20ad9d51a2e063e8bb07af06accdc97d93a49e895c868bb8e`

See more details on using hashes here.

Provenance

The following attestation bundles were made for mlbricks_kit-1.0.0b1-cp312-cp312-win_amd64.whl:

Publisher: native-wheels-beta.yml on MlBricks/MLBricks

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Statement:
- Statement type: https://in-toto.io/Statement/v1
- Predicate type: https://docs.pypi.org/attestations/publish/v1
- Subject name: mlbricks_kit-1.0.0b1-cp312-cp312-win_amd64.whl
- Subject digest: b50ec1b86621c64f4fdc4901557ff8cbd27c2377c580e9b029115ea0aadfa8c6
- Sigstore transparency entry: 2715383929
- Sigstore integration time: Sep 4, 2026
Source repository:
- Permalink: MlBricks/MLBricks@d86fccc6645640d98b7db33916ef2021564b5d0b
- Branch / Tag: refs/tags/v1.0.0b1
- Owner: https://github.com/MlBricks
- Access: public
Publication detail:
- Token Issuer: https://token.actions.githubusercontent.com
- Runner Environment: github-hosted
- Publication workflow: native-wheels-beta.yml@d86fccc6645640d98b7db33916ef2021564b5d0b
- Trigger Event: push

File details

Details for the file mlbricks_kit-1.0.0b1-cp312-cp312-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl.

File metadata

Download URL: mlbricks_kit-1.0.0b1-cp312-cp312-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl
Upload date: Sep 4, 2026
Size: 26.9 MB
Tags: CPython 3.12, manylinux: glibc 2.27+ x86-64, manylinux: glibc 2.28+ x86-64
Uploaded using Trusted Publishing? Yes
Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for mlbricks_kit-1.0.0b1-cp312-cp312-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl
Algorithm	Hash digest
SHA256	`765115b3f05be1e6e4b1eb23d5b4c41474fbb4ea6fbbe605be79b0dc9d5da05f`
MD5	`8ad8689060e56b45c403a02a7b862921`
BLAKE2b-256	`29f66c6e4f21a8b29d5baffc8781ceab595387c229107bc9555db4fd651952e2`

See more details on using hashes here.

Provenance

The following attestation bundles were made for mlbricks_kit-1.0.0b1-cp312-cp312-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl:

Publisher: native-wheels-beta.yml on MlBricks/MLBricks

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Statement:
- Statement type: https://in-toto.io/Statement/v1
- Predicate type: https://docs.pypi.org/attestations/publish/v1
- Subject name: mlbricks_kit-1.0.0b1-cp312-cp312-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl
- Subject digest: 765115b3f05be1e6e4b1eb23d5b4c41474fbb4ea6fbbe605be79b0dc9d5da05f
- Sigstore transparency entry: 2715383759
- Sigstore integration time: Sep 4, 2026
Source repository:
- Permalink: MlBricks/MLBricks@d86fccc6645640d98b7db33916ef2021564b5d0b
- Branch / Tag: refs/tags/v1.0.0b1
- Owner: https://github.com/MlBricks
- Access: public
Publication detail:
- Token Issuer: https://token.actions.githubusercontent.com
- Runner Environment: github-hosted
- Publication workflow: native-wheels-beta.yml@d86fccc6645640d98b7db33916ef2021564b5d0b
- Trigger Event: push

File details

Details for the file mlbricks_kit-1.0.0b1-cp312-cp312-macosx_11_0_arm64.whl.

File metadata

Download URL: mlbricks_kit-1.0.0b1-cp312-cp312-macosx_11_0_arm64.whl
Upload date: Sep 4, 2026
Size: 706.0 kB
Tags: CPython 3.12, macOS 11.0+ ARM64
Uploaded using Trusted Publishing? Yes
Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for mlbricks_kit-1.0.0b1-cp312-cp312-macosx_11_0_arm64.whl
Algorithm	Hash digest
SHA256	`ae6f191d9728b7635b05b36a6e76f461fc42232ff4fb1f34995aa3b207b7d67d`
MD5	`eeb6819b7aac3d15c064defc466f0c0e`
BLAKE2b-256	`490080dba638d2ae8ff83996e2b73d3306c773b193e9f981fefd92b6c984f5a9`

See more details on using hashes here.

Provenance

The following attestation bundles were made for mlbricks_kit-1.0.0b1-cp312-cp312-macosx_11_0_arm64.whl:

Publisher: native-wheels-beta.yml on MlBricks/MLBricks

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Statement:
- Statement type: https://in-toto.io/Statement/v1
- Predicate type: https://docs.pypi.org/attestations/publish/v1
- Subject name: mlbricks_kit-1.0.0b1-cp312-cp312-macosx_11_0_arm64.whl
- Subject digest: ae6f191d9728b7635b05b36a6e76f461fc42232ff4fb1f34995aa3b207b7d67d
- Sigstore transparency entry: 2715383556
- Sigstore integration time: Sep 4, 2026
Source repository:
- Permalink: MlBricks/MLBricks@d86fccc6645640d98b7db33916ef2021564b5d0b
- Branch / Tag: refs/tags/v1.0.0b1
- Owner: https://github.com/MlBricks
- Access: public
Publication detail:
- Token Issuer: https://token.actions.githubusercontent.com
- Runner Environment: github-hosted
- Publication workflow: native-wheels-beta.yml@d86fccc6645640d98b7db33916ef2021564b5d0b
- Trigger Event: push

File details

Details for the file mlbricks_kit-1.0.0b1-cp311-cp311-win_amd64.whl.

File metadata

Download URL: mlbricks_kit-1.0.0b1-cp311-cp311-win_amd64.whl
Upload date: Sep 4, 2026
Size: 4.0 MB
Tags: CPython 3.11, Windows x86-64
Uploaded using Trusted Publishing? Yes
Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for mlbricks_kit-1.0.0b1-cp311-cp311-win_amd64.whl
Algorithm	Hash digest
SHA256	`ce700534cb25e7d7fef82502358b24aa572adf3332f24124dd194dca81251e3b`
MD5	`096a1666fe763fa1ae28498f25d2a4a9`
BLAKE2b-256	`f5ca648507eb9f0bb043af711be4cd8873c363a468f5c45923ca82bb4bee0d3c`

See more details on using hashes here.

Provenance

The following attestation bundles were made for mlbricks_kit-1.0.0b1-cp311-cp311-win_amd64.whl:

Publisher: native-wheels-beta.yml on MlBricks/MLBricks

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Statement:
- Statement type: https://in-toto.io/Statement/v1
- Predicate type: https://docs.pypi.org/attestations/publish/v1
- Subject name: mlbricks_kit-1.0.0b1-cp311-cp311-win_amd64.whl
- Subject digest: ce700534cb25e7d7fef82502358b24aa572adf3332f24124dd194dca81251e3b
- Sigstore transparency entry: 2715384046
- Sigstore integration time: Sep 4, 2026
Source repository:
- Permalink: MlBricks/MLBricks@d86fccc6645640d98b7db33916ef2021564b5d0b
- Branch / Tag: refs/tags/v1.0.0b1
- Owner: https://github.com/MlBricks
- Access: public
Publication detail:
- Token Issuer: https://token.actions.githubusercontent.com
- Runner Environment: github-hosted
- Publication workflow: native-wheels-beta.yml@d86fccc6645640d98b7db33916ef2021564b5d0b
- Trigger Event: push

File details

Details for the file mlbricks_kit-1.0.0b1-cp311-cp311-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl.

File metadata

Download URL: mlbricks_kit-1.0.0b1-cp311-cp311-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl
Upload date: Sep 4, 2026
Size: 26.9 MB
Tags: CPython 3.11, manylinux: glibc 2.27+ x86-64, manylinux: glibc 2.28+ x86-64
Uploaded using Trusted Publishing? Yes
Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for mlbricks_kit-1.0.0b1-cp311-cp311-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl
Algorithm	Hash digest
SHA256	`48ff6c10a932689211a7cf1e0cf6a34b258d99ca4b7a02496fad1105e7bcd93e`
MD5	`7f285cb4b003f95d24b72fdec4389954`
BLAKE2b-256	`87bf9fa332bfbafd456d4e22a977154c5e7484a47c864d76afa62645a7981407`

See more details on using hashes here.

Provenance

The following attestation bundles were made for mlbricks_kit-1.0.0b1-cp311-cp311-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl:

Publisher: native-wheels-beta.yml on MlBricks/MLBricks

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Statement:
- Statement type: https://in-toto.io/Statement/v1
- Predicate type: https://docs.pypi.org/attestations/publish/v1
- Subject name: mlbricks_kit-1.0.0b1-cp311-cp311-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl
- Subject digest: 48ff6c10a932689211a7cf1e0cf6a34b258d99ca4b7a02496fad1105e7bcd93e
- Sigstore transparency entry: 2715383991
- Sigstore integration time: Sep 4, 2026
Source repository:
- Permalink: MlBricks/MLBricks@d86fccc6645640d98b7db33916ef2021564b5d0b
- Branch / Tag: refs/tags/v1.0.0b1
- Owner: https://github.com/MlBricks
- Access: public
Publication detail:
- Token Issuer: https://token.actions.githubusercontent.com
- Runner Environment: github-hosted
- Publication workflow: native-wheels-beta.yml@d86fccc6645640d98b7db33916ef2021564b5d0b
- Trigger Event: push

File details

Details for the file mlbricks_kit-1.0.0b1-cp311-cp311-macosx_11_0_arm64.whl.

File metadata

Download URL: mlbricks_kit-1.0.0b1-cp311-cp311-macosx_11_0_arm64.whl
Upload date: Sep 4, 2026
Size: 699.9 kB
Tags: CPython 3.11, macOS 11.0+ ARM64
Uploaded using Trusted Publishing? Yes
Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for mlbricks_kit-1.0.0b1-cp311-cp311-macosx_11_0_arm64.whl
Algorithm	Hash digest
SHA256	`c26ac2b254116d32e8976b57e691476f125b74b85ff0230185c71684f756a7fb`
MD5	`34dcfa14e94bdd2e7ef7b01f56d3da6e`
BLAKE2b-256	`b137408d5b88762ff91face0540bc2a9a1f19d15065a7aa728d2697fb13f5122`

See more details on using hashes here.

Provenance

The following attestation bundles were made for mlbricks_kit-1.0.0b1-cp311-cp311-macosx_11_0_arm64.whl:

Publisher: native-wheels-beta.yml on MlBricks/MLBricks

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Statement:
- Statement type: https://in-toto.io/Statement/v1
- Predicate type: https://docs.pypi.org/attestations/publish/v1
- Subject name: mlbricks_kit-1.0.0b1-cp311-cp311-macosx_11_0_arm64.whl
- Subject digest: c26ac2b254116d32e8976b57e691476f125b74b85ff0230185c71684f756a7fb
- Sigstore transparency entry: 2715383410
- Sigstore integration time: Sep 4, 2026
Source repository:
- Permalink: MlBricks/MLBricks@d86fccc6645640d98b7db33916ef2021564b5d0b
- Branch / Tag: refs/tags/v1.0.0b1
- Owner: https://github.com/MlBricks
- Access: public
Publication detail:
- Token Issuer: https://token.actions.githubusercontent.com
- Runner Environment: github-hosted
- Publication workflow: native-wheels-beta.yml@d86fccc6645640d98b7db33916ef2021564b5d0b
- Trigger Event: push

File details

Details for the file mlbricks_kit-1.0.0b1-cp310-cp310-win_amd64.whl.

File metadata

Download URL: mlbricks_kit-1.0.0b1-cp310-cp310-win_amd64.whl
Upload date: Sep 4, 2026
Size: 4.0 MB
Tags: CPython 3.10, Windows x86-64
Uploaded using Trusted Publishing? Yes
Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for mlbricks_kit-1.0.0b1-cp310-cp310-win_amd64.whl
Algorithm	Hash digest
SHA256	`be7f17da71f603c4bf26a45f6970a0a0e468af56a1d78c8b670abd824c0660fc`
MD5	`3775267e9d101bdf998fc532dde40921`
BLAKE2b-256	`cfd87414c6c561174083eaf8997b947c9ab081acb4608ca6eb7973c8230f8580`

See more details on using hashes here.

Provenance

The following attestation bundles were made for mlbricks_kit-1.0.0b1-cp310-cp310-win_amd64.whl:

Publisher: native-wheels-beta.yml on MlBricks/MLBricks

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Statement:
- Statement type: https://in-toto.io/Statement/v1
- Predicate type: https://docs.pypi.org/attestations/publish/v1
- Subject name: mlbricks_kit-1.0.0b1-cp310-cp310-win_amd64.whl
- Subject digest: be7f17da71f603c4bf26a45f6970a0a0e468af56a1d78c8b670abd824c0660fc
- Sigstore transparency entry: 2715383720
- Sigstore integration time: Sep 4, 2026
Source repository:
- Permalink: MlBricks/MLBricks@d86fccc6645640d98b7db33916ef2021564b5d0b
- Branch / Tag: refs/tags/v1.0.0b1
- Owner: https://github.com/MlBricks
- Access: public
Publication detail:
- Token Issuer: https://token.actions.githubusercontent.com
- Runner Environment: github-hosted
- Publication workflow: native-wheels-beta.yml@d86fccc6645640d98b7db33916ef2021564b5d0b
- Trigger Event: push

File details

Details for the file mlbricks_kit-1.0.0b1-cp310-cp310-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl.

File metadata

Download URL: mlbricks_kit-1.0.0b1-cp310-cp310-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl
Upload date: Sep 4, 2026
Size: 26.8 MB
Tags: CPython 3.10, manylinux: glibc 2.27+ x86-64, manylinux: glibc 2.28+ x86-64
Uploaded using Trusted Publishing? Yes
Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for mlbricks_kit-1.0.0b1-cp310-cp310-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl
Algorithm	Hash digest
SHA256	`5ce753e089cb0b4f40f64a81de6d5afe37c46984eb241bb4d4b743aac8792de3`
MD5	`d267e9c97ca69c9fea14e8de86fe2f9e`
BLAKE2b-256	`c98a0de9c8690c5805e0ddde5a05e99bc157a03eb4f088eb902bb1e654797bd6`

See more details on using hashes here.

Provenance

The following attestation bundles were made for mlbricks_kit-1.0.0b1-cp310-cp310-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl:

Publisher: native-wheels-beta.yml on MlBricks/MLBricks

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Statement:
- Statement type: https://in-toto.io/Statement/v1
- Predicate type: https://docs.pypi.org/attestations/publish/v1
- Subject name: mlbricks_kit-1.0.0b1-cp310-cp310-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl
- Subject digest: 5ce753e089cb0b4f40f64a81de6d5afe37c46984eb241bb4d4b743aac8792de3
- Sigstore transparency entry: 2715383626
- Sigstore integration time: Sep 4, 2026
Source repository:
- Permalink: MlBricks/MLBricks@d86fccc6645640d98b7db33916ef2021564b5d0b
- Branch / Tag: refs/tags/v1.0.0b1
- Owner: https://github.com/MlBricks
- Access: public
Publication detail:
- Token Issuer: https://token.actions.githubusercontent.com
- Runner Environment: github-hosted
- Publication workflow: native-wheels-beta.yml@d86fccc6645640d98b7db33916ef2021564b5d0b
- Trigger Event: push

File details

Details for the file mlbricks_kit-1.0.0b1-cp310-cp310-macosx_11_0_arm64.whl.

File metadata

Download URL: mlbricks_kit-1.0.0b1-cp310-cp310-macosx_11_0_arm64.whl
Upload date: Sep 4, 2026
Size: 693.7 kB
Tags: CPython 3.10, macOS 11.0+ ARM64
Uploaded using Trusted Publishing? Yes
Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for mlbricks_kit-1.0.0b1-cp310-cp310-macosx_11_0_arm64.whl
Algorithm	Hash digest
SHA256	`489ee42f1815a5cdf4885c7bf9b73ad485ac5c8e549a60628ee59093d390a89a`
MD5	`178d53547ae2b3101d9494c2868687da`
BLAKE2b-256	`1e57f1d47c7412e4c6818e849c115239adaa8abf2cf6343398c4682ce830c3f6`

See more details on using hashes here.

Provenance

The following attestation bundles were made for mlbricks_kit-1.0.0b1-cp310-cp310-macosx_11_0_arm64.whl:

Publisher: native-wheels-beta.yml on MlBricks/MLBricks

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Statement:
- Statement type: https://in-toto.io/Statement/v1
- Predicate type: https://docs.pypi.org/attestations/publish/v1
- Subject name: mlbricks_kit-1.0.0b1-cp310-cp310-macosx_11_0_arm64.whl
- Subject digest: 489ee42f1815a5cdf4885c7bf9b73ad485ac5c8e549a60628ee59093d390a89a
- Sigstore transparency entry: 2715383862
- Sigstore integration time: Sep 4, 2026
Source repository:
- Permalink: MlBricks/MLBricks@d86fccc6645640d98b7db33916ef2021564b5d0b
- Branch / Tag: refs/tags/v1.0.0b1
- Owner: https://github.com/MlBricks
- Access: public
Publication detail:
- Token Issuer: https://token.actions.githubusercontent.com
- Runner Environment: github-hosted
- Publication workflow: native-wheels-beta.yml@d86fccc6645640d98b7db33916ef2021564b5d0b
- Trigger Event: push

Release history Release notifications | RSS feed

Pre-release

1.0.0b2

Sep 5, 2026 14 files

This release

Pre-release

1.0.0b1 This release

Sep 4, 2026 14 files

mlbricks-kit 1.0.0b1

MLBricks Kit 1.0.0b1

Core components

Version

Installation

Quick start

Unified model lifecycle

ESA

Bolt

Bolt language model

FFNBricks

ResController

SOUP

ElasticBit

VESA

VisualBolt

Build a custom model with Bricks

Native vision runtime

API documentation

Component licenses

Licensing

Release

Training compilation

Bolt native build control

SOUP component license

Versioning note

Beta native wheel distribution

Beta native platform support

Project links

Key dates

1 maintainer

Credits

GitHub Statistics

License expression

Requires

Provides Extra

Tags

Classifiers

Download files

Source Distribution

Built Distributions

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes