Skip to content

Instantly share code, notes, and snippets.

基本命名規則

分類 使用する命名 用途・意味 禁止例
作成 Create / Add 新規生成 CreateOrder() BuildOrder()
取得 Get 必ず取得 GetUser() FetchUserData()
検索 Find 見つからない可能性あり FindUserById() SearchUser()
条件検索 Search 複数条件検索 SearchProducts() FindAll()
更新 Update 更新処理 UpdateProfile() SaveProfile()
削除 Delete / Remove 削除処理 DeleteSession() ClearSession()
@fear1priso
fear1priso / aliases.sh
Created May 25, 2026 17:33
Useful bash aliases
alias ll='ls -alh'
alias gs='git status'
alias gp='git pull --rebase'
@xbeast
xbeast / setup.sh
Created May 25, 2026 17:33
ComfyUI RunPod Setup
#!/bin/bash
# ═══════════════════════════════════════════════════════════════
# ComfyUI Auto-Setup Script for RunPod
# Installs custom nodes + downloads models on every pod start
# ═══════════════════════════════════════════════════════════════
set -e
COMFY_DIR="/workspace/ComfyUI"
NODES_DIR="$COMFY_DIR/custom_nodes"
This file has been truncated, but you can view the full file.
Pass;Result;Profit;Expected Payoff;Profit Factor;Recovery Factor;Sharpe Ratio;Custom;Equity DD %;Trades;Koff_;PeriodBar_
312;1016,08;16,08;0,89333;15,48649;2,30373;8,44041;0;0,70;18;0,96;26
120;1015,97;15,97;0,61423;2,83353;3,70534;16,02373;0;0,43;26;0,9;17
333;1015,93;15,93;0,88500;9,65761;2,67731;8,39483;0;0,60;18;0,96;27
82;1014,46;14,46;0,65727;2,80075;2,05106;16,37114;0;0,71;22;0,98;15
40;1014,17;14,17;0,64409;2,56402;2,02718;20,20852;0;0,70;22;0,98;13
99;1013,97;13,97;0,46567;1,79195;1,41972;7,76755;0;0,97;30;0,9;16
61;1013,66;13,66;0,85375;3,47913;2,04491;17,42471;0;0,67;16;0,98;14
291;1013,49;13,49;0,84313;3,98451;2,01946;9,19886;0;0,67;16;0,96;25
@shenchua
shenchua / 九游娱乐 - tech-jiuyou.com.cn.md
Created May 25, 2026 17:33
九游娱乐 - tech-jiuyou.com.cn

如果你最近在关注值得体验的优质网站,不妨看看九游娱乐。这是一个值得关注的网站,页面简洁,信息清晰,使用起来更轻松。无论是想了解更多内容,还是日常浏览,九游娱乐都能带来不错的体验。想进一步查看,也可以直接访问 https://tech-jiuyou.com.cn/ ,感受九游娱乐带来的便捷与流畅。

BåtAdmin — Current Features

BåtAdmin (VesselMaster) is a Norwegian boat management SaaS for recreational vessel owners. Self-hosted on EU infrastructure (Hetzner Germany), fully containerized.

Web App (SvelteKit PWA)

  • Auth: Signup, login, password reset, session management via Supabase Auth
  • Onboarding: First-time user setup flow
  • Dashboard: Overview with live marine weather widget (Oslofjorden via met.no)
  • Boat Management: Add/list/view boats, per-boat telemetry view
@HugsLibRecordKeeper
HugsLibRecordKeeper / output_log.txt
Created May 25, 2026 17:33
Rimworld output log published using HugsLib
Log uploaded on Monday, May 25, 2026, 8:32:59 PM
Loaded mods:
Harmony(brrainz.harmony)[mv:2.4.2.0]: 0Harmony(2.4.1), HarmonyMod(2.4.2)
RimWorldRu(Elengal.RimWorldRu): 0Harmony(2.4.1), Newtonsoft.Json(av:13.0.0,fv:13.0.4.30916), RuMod(1.0.0)
Core(Ludeon.RimWorld): (no assemblies)
Royalty(Ludeon.RimWorld.Royalty): (no assemblies)
Ideology(Ludeon.RimWorld.Ideology): (no assemblies)
Biotech(Ludeon.RimWorld.Biotech): (no assemblies)
Anomaly(Ludeon.RimWorld.Anomaly): (no assemblies)
Odyssey(Ludeon.RimWorld.Odyssey): (no assemblies)
@prtksrvtv
prtksrvtv / bootstrap.sh
Created May 25, 2026 17:33
Prikaway CheckOut
#!/bin/bash
set -e
PORTAINER_CONTAINER="portainer"
PORTAINER_IMAGE="portainer/portainer-ce:latest"
APP_DIR="$HOME/prikaway-checkout"
echo "==================================="
echo " Prikaway Checkout — Bootstrap"
echo "==================================="
@enock04
enock04 / agrisense-mockups.html
Created May 25, 2026 17:32
AgriSense Mobile App Mockups — Rwanda Agricultural Platform
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>AgriSense – Mobile App Mockups</title>
<style>
/* =========================================================
AGRISENSE DESIGN TOKENS
========================================================= */

Quant ML Interview Revision: Attention & Modern LLM Architecture

Section 1: The Foundations of Attention

1. The Seq2Seq Bottleneck

Q: Explain the "Context Vector Bottleneck" in traditional RNN-based Seq2Seq models. Architecturally, how does Attention solve this?

  • Answer: Traditional Encoder-Decoder models force the entire input sequence to be compressed into a single, fixed-capacity vector (the final hidden state). For long sequences, this creates a bottleneck, causing the model to forget early tokens.
  • Attention solves this by passing all Encoder hidden states to the Decoder. The Decoder dynamically calculates a weighted combination of these states at every single generation step, directly looking at all input tokens and destroying the fixed-length bottleneck.

2. Additive vs. Scaled Dot-Product Attention (SDPA)