a10y

TurboQuant: Online Vector Quantization with Near-optimal Distortion Rate

Paper Summary | arXiv:2504.19874

Authors: Amir Zandieh (Google Research), Majid Daliri (NYU), Majid Hadian (Google DeepMind), Vahab Mirrokni (Google Research)

---

What This Paper Is About

a10y

0000000000040470 t _GLOBAL__sub_I_nvcompManagerFactory.cpp
00000000000e5780 t __device_stub__ZN6nvcomp20fill_cudecomp_paramsEPKPKvPKmPKPvPmP24CUmemDecompressParams_stP14nvcompStatus_tmPi(void const* const*, unsigned long const*, void* const*, unsigned long*, CUmemDecompressParams_st*, nvcompStatus_t*, unsigned long, int*)
00000000000abc30 t __device_stub__ZN6nvcomp38compact_comp_buffers_and_header_outputEPhPKPKhPKmP12CommonHeaderPmmmm10FormatTypeS9_m(unsigned char*, unsigned char const* const*, unsigned long const*, CommonHeader*, unsigned long*, unsigned long, unsigned long, unsigned long, FormatType, unsigned long*, unsigned long)
0000000000115ac0 t __device_stub__ZN14nvcomp_deflate18fill_params_kernelEP24CUmemDecompressParams_stPKPKvPKmPmPKPvbP14nvcompStatus_tmPKi(CUmemDecompressParams_st*, void const* const*, unsigned long const*, unsigned long*, void* const*, bool, nvcompStatus_t*, unsigned long, int const*)
00000000000e3770 t __device_stub__ZN6nvcomp8lowlevel20fill_cudecomp_paramsEPKPKvPKmPKPvPmP24CUmem

a10y

# find all refs with remotes that are deleted on the server, and delete them locally
git branch -vv | grep ': gone]' | awk '{print $1}' | xargs -r git branch -D

a10y

config:
    flowchart:
        defaultRenderer: elk
flowchart TD
   %% Data Types
   Integer[Integer]:::datatype
   Float[Float]:::datatype
   String[String]:::datatype

a10y

/// Create a new struct type. For example:
///
/// ```
/// use vortex_dtype::{struct_type, DType, PType};
///
/// let the_struct = struct_type! {
///     "x" => DType::from(PType::F64),
///     "y" => DType::from(PType::F64),
/// };
///

a10y

FROM dustynv/nano_llm:r36.4.0

RUN apt update && apt install -y gstreamer1.0-nice

a10y

package org.apache.iceberg.parquet;

import java.io.IOException;
import java.util.ArrayList;
import java.util.List;
import org.apache.avro.Schema;
import org.apache.avro.generic.GenericRecord;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.FileSystem;
import org.apache.hadoop.fs.Path;

a10y

import json
import os
import glob
import pytesseract
import multiprocessing
from pdf2image import convert_from_path


def ocr_pdf_to_json(pdf_path):
    # Convert PDF to images (one per page)

a10y

// use rand::Rng;
use std::time::Instant;
use zstd::{decode_all, stream::encode_all};

fn main() -> Result<(), Box<dyn std::error::Error>> {
    // Generate sample data - 1 million random u32 integers
    let count = 64_000;
    println!("Generating {} random integers...", count);

    // let mut rng = rand::rng();

a10y

plugins {
    `java-library`
    `maven-publish`
    `signing`
}

val tokenizerSharedLibrary by configurations.creating {
    isCanBeConsumed = false
}