ternaus

Albucore #219: writable LUT allocation

Baseline: 4d798c7. Candidate: 81a87ea, PR #220. Measurements: 2026-11-05.

Each host ran 66 paired, complete public NumPy apply_uint8_lut workloads with uint8 data and exactly one CPU thread. The grid covers canonical non-square HWC C=1/3/9, tiny inputs, rank-4/rank-5 arrays, strides, Fortran order, read-only inputs and supplied outputs. Every candidate output is checked against table indexing and unchanged input. Timings use 31 interleaved samples with calibrated batches targeting 3 ms; raw per-call samples and complete package/source/thread/ISA inventories are in the JSON files.

Host	Median new / baseline time	Initial maximum
Apple M4 Max / macOS ARM64	0.9989x	1.0117x
Xeon

ternaus

Biome 2.5.15 package main reproduction

Create these files in an empty directory:

biome.json

{"linter":{"rules":{"correctness":{"noUnresolvedImports":"error"}}}}

node_modules/example/package.json

ternaus

albu-spec Contributor License Agreement

Version 2.0 — September 9, 2026

This Contributor License Agreement (the Agreement) is between the person or legal entity accepting it (You) and Albumentations, LLC, a Delaware limited liability company (Albumentations, We, or Us).

Plain-English Summary

ternaus

import albumentations as A
import cv2

bgr_image = cv2.imread("cat1.jpeg")
image = cv2.cvtColor(bgr_image, cv2.COLOR_BGR2RGB)

transform = A.Compose([A.RandomResizedCrop(size=(512, 512), 
                                     scale=(0.08, 1), 
                                     ratio=(0.75, 1.33), 
                                     interpolation=cv2.INTER_AREA, p=0.5)])

ternaus

# https://raw.githubusercontent.com/ternaus/retinaface/master/setup.py

import io
import os
import re
import sys
from shutil import rmtree
from typing import Tuple, List

from setuptools import Command, find_packages, setup

ternaus

# https://github.com/ternaus/retinaface/blob/master/retinaface/pre_trained_models.py

from collections import namedtuple
from torch.utils import model_zoo
from retinaface.predict_single import Model

model = namedtuple("model", ["url", "model"])

models = {
    "resnet50_2020-07-20": model(

ternaus

import albumentations as A

transform = A.Compose([A.HorizontalFlip(p=0.5),
                       A.ShiftScaleRotate(border_mode=cv2.BORDER_CONSTANT, scale_limit=0.3, p=0.5)],
                         bbox_params=albu.BboxParams(format='pascal_voc', label_fields=['category_ids']),
                         keypoint_params=albu.KeypointParams(format='xy'),
                         additional_targets={
                             "image1": "image",
                             "bboxes1": "bboxes",
                             "mask1": "mask",

ternaus

import albumentations as A

transform = A.Compose([
    A.HorizontalFlip(p=0.5),
    A.GridDistortion(p=0.5),    
    A.RandomCrop(height=1024, width=1024, p=0.5),    
], p=1)

transformed = transform(image=image, masks=[mask, mask2])

ternaus

import albumentations as A

transform = A.Compose([
    A.HorizontalFlip(p=0.5),
    A.ShiftScaleRotate(border_mode=cv2.BORDER_CONSTANT, 
                          scale_limit=0.3,
                          rotate_limit=(10, 30),
                          p=0.5)
], p=1)

ternaus

black .
isort