国内从 Docker Hub 拉取镜像有时会遇到困难,此时可以配置镜像加速器。Docker 官方和国内很多云服务商都提供了国内加速器服务。
Dockerized 实践 https://github.com/y0ngb1n/dockerized
Ubuntu 16.04+、Debian 8+、CentOS 7+
国内从 Docker Hub 拉取镜像有时会遇到困难,此时可以配置镜像加速器。Docker 官方和国内很多云服务商都提供了国内加速器服务。
Dockerized 实践 https://github.com/y0ngb1n/dockerized
Ubuntu 16.04+、Debian 8+、CentOS 7+
Good question! I am collecting human data on how quantization affects outputs. See here for more information: ggerganov/llama.cpp#5962
In the meantime, use the largest that fully fits in your GPU. If you can comfortably fit Q4_K_S, try using a model with more parameters.
See the wiki upstream: https://github.com/ggerganov/llama.cpp/wiki/Feature-matrix
git_current_branch () { | |
local ref | |
ref=$(command git symbolic-ref --quiet HEAD 2> /dev/null) | |
local ret=$? | |
if [[ $ret != 0 ]] | |
then | |
[[ $ret == 128 ]] && return | |
ref=$(command git rev-parse --short HEAD 2> /dev/null) || return | |
fi | |
echo ${ref#refs/heads/} |
//short code to get the woocommerce recently viewed products | |
<?php function custom_track_product_view() { | |
if ( ! is_singular( 'product' ) ) { | |
return; | |
} | |
global $post; | |
if ( empty( $_COOKIE['woocommerce_recently_viewed'] ) ) | |
$viewed_products = array(); |
@Composable | |
fun FlowRow( | |
horizontalGap: Dp = 0.dp, | |
verticalGap: Dp = 0.dp, | |
alignment: Alignment.Horizontal = Alignment.Start, | |
content: @Composable () -> Unit, | |
) = Layout(content = content) { measurables, constraints -> | |
val horizontalGapPx = horizontalGap.toPx().roundToInt() | |
val verticalGapPx = verticalGap.toPx().roundToInt() |
.editor-group-watermark { | |
max-width: 400px !important; | |
} | |
.editor-group-watermark > .letterpress { | |
aspect-ratio: 1920/1080 !important; | |
background-image:url('https://github.com/SAWARATSUKI/ServiceLogos/blob/c2cf29211b93efe48e5c34451bac5c56f979ab94/VisualStudioCode/VisualStudioCodeRound.png?raw=true') !important; | |
background-position: center; | |
} |
This is a compiled collection of common Japanese SFXs (sound effects or onomatopoeia).
This collection uses mostly Romaji transliterations. Some SFXs can be used in combination with other SFXs; some of them, or combinations thereof, can be written separately for emphasis within the same panel (example: do-ki-, ドッ キッ ), but can be considered part of the same action; some use repeated sub-elements or prolongation characters (ー) within them. All interjections can be written differently, usually to indicate particularly strong or repeated stimulation. For example, a cho in isolation is almost certainly a truncated version of chotto, where the speaker was distracted from finishing the word. A prolongation of a character, on the other hand, often implies lazy, harsh, or sloppy pronunciation, or possibly even a scream. Because of this, many SFXs can have multiple meanings, depending on the action drawn, as well as one's interpretation of that action.
a: general interjection and versa
import os | |
import argparse | |
from PIL import Image | |
from diffusers import AutoPipelineForText2Image | |
import torch | |
args = None | |
# Define the models | |
SDXL_MODEL = "/models/stable-diffusion/sdxl/photoStableXL.safetensors" |
! 2021-12-29 https://139.99.17.109 | |
139.99.17.109###top-banner | |
139.99.17.109###after-menu-2 | |
! 2021-12-29 https://www.facebook.com | |
||facebook.com/video/unified_cvc/*$xhr,domain=facebook.com | |
! 2022-01-06 https://139.99.17.109 | |
139.99.17.109##.post > .custom-html-widget.textwidget > .row | |
139.99.17.109###middle-banner |