update tax

wenerme · May 3, 2024 · b441539 · b441539
1 parent 7ff784e
commit b441539
Show file tree

Hide file tree

Showing 34 changed files with 1,189 additions and 486 deletions.
diff --git a/notes/ai/llm/token.md b/notes/ai/llm/token.md
@@ -0,0 +1,17 @@
+---
+title: Token
+---
+
+# LLM Tokenizer
+
+**ChatGPT 特殊 Token**
+
+```
+<|endoftext|>
+<|eos|>
+<|pad|>
+<|bos|>
+<|eol|>
+<|math|>
+<|doc|>
+```
diff --git a/notes/ai/service/openai/README.md b/notes/ai/service/openai/README.md
@@ -6,12 +6,17 @@ title: OpenAI
 
 **价格**
 
-| Service      | Input/1K tokens | Output/1K tokens |     Chinese | note               |
-| ------------ | --------------- | ---------------- | ----------: | ------------------ |
-| GPT-4 8K     | $0.03           | $0.06            | ~1.6 万字/$ |
-| GPT-4 32K    | $0.06           | $0.12            | ~0.8 万字/$ |
-| Chat 3.5 4K  | $0.0015         | $0.002           |  ~30 万字/$ | gpt-3.5-turbo      |
-| Chat 3.5 16K | $0.003          | $0.004           |  ~15 万字/$ | gpt-3.5-turbo-0613 |
+| Service                |             Input |             Output | Notes      |
+| ---------------------- | ----------------: | -----------------: | ---------- |
+| gpt-4-turbo 128k       | $10.00 /1M tokens |  $30.00 /1M tokens | 更好更便宜 |
+| gpt-4                  | $30.00 /1M tokens |  $60.00 /1M tokens |
+| gpt-4-32k              | $60.00 /1M tokens | $120.00 /1M tokens |
+| gpt-3.5-turbo-0125     |  $0.50 /1M tokens |   $1.50 /1M tokens |
+| gpt-3.5-turbo-instruct |  $1.50 /1M tokens |   $2.00 /1M tokens |
+
+- 1M tokens 大约 50 万字
+
+---
 
 - InstructGPT - 单个指令生成
   - Ada - $0.0004 / 1K tokens
@@ -114,7 +119,6 @@ Rate limit reached for default-gpt-3.5-turbo in organization org-xxx on requests
 
 ## 您的银行卡被拒绝了。
 
-
 ## Access denied
 
 尝试不用 VPN 或者换 VPN

diff --git a/notes/culture/game/baccarat.md b/notes/culture/game/baccarat.md
@@ -42,30 +42,31 @@ tags:
 1. Ace 为 1 点
 1. J, Q, K, 10 为 0 点
 1. 一靴牌
-  - 第21局+，不得投注 大小
-  - 第26局+，不得投注 超和
-  - 第46局+，不得投注 庄龙宝、闲龙宝、完美对子、任意对子、庄对、闲对
-  - 第51局+，不得投注 闲例牌/庄例牌
+
+- 第21局+，不得投注 大小
+- 第26局+，不得投注 超和
+- 第46局+，不得投注 庄龙宝、闲龙宝、完美对子、任意对子、庄对、闲对
+- 第51局+，不得投注 闲例牌/庄例牌
 
 **投注项目**
 
 | bet      | rate |
 | -------- | ---- |
-| 闲家     |1
-| 庄家     |0.95
-| 和局     |8
-| 闲对     |11
-| 庄对     |11
-| 大       |0.5
-| 小       |1.5
-| 任意对子 |5
-| 完美对子 |25
-| 闲例牌   |4
-| 庄例牌   |4
-超和 N|
+| 闲家     | 1    |
+| 庄家     | 0.95 |
+| 和局     | 8    |
+| 闲对     | 11   |
+| 庄对     | 11   |
+| 大       | 0.5  |
+| 小       | 1.5  |
+| 任意对子 | 5    |
+| 完美对子 | 25   |
+| 闲例牌   | 4    |
+| 庄例牌   | 4    |
+| 超和 N   |
 
 - 大小 - 4张牌为小，5张牌或6张牌为大
 - 完美对子 - 庄或闲的前两张牌的牌面，必须同花同色同数字或同英文字母
 - 例牌 - 庄或闲的前两张牌点数总和为8或9点并胜出
-- 超和  - 指庄和闲为和局时，押中和局点数
+- 超和 - 指庄和闲为和局时，押中和局点数
 - 超级六 - 庄家点数为 6 而且赢闲家
diff --git a/notes/culture/game/game-glossary.md b/notes/culture/game/game-glossary.md
@@ -0,0 +1,22 @@
+---
+tags:
+  - Glossary
+---
+
+# Game Glossary
+
+| abbr. | for                | cn     |
+| ----- | ------------------ | ------ |
+| 1B3P  | 1 Banker 3 Players | 1庄3闲 |
+
+| en          | cn   |
+| ----------- | ---- |
+| player      | 闲家 |
+| banker      | 庄家 |
+| tie         | 和局 |
+| player pair | 闲对 |
+| banker pair | 庄对 |
+| big         | 大   |
+| small       | 小   |
+| big pair    | 大对 |
+| small pair  | 小对 |
diff --git a/notes/db/relational/mysql/mysql-version.md b/notes/db/relational/mysql/mysql-version.md
@@ -5,10 +5,16 @@ tags:
 
 # MySQL Version
 
-| version   | GA         | eol        |
-| --------- | ---------- | ---------- |
-| MySQL 8.0 | 2018-04-19 | 2026-04    |
-| MySQL 5.7 | 2015-10-21 | 2023-10    |
-| MySQL 5.6 | 2013-02-05 | 2021-02-05 |
+| version     | GA         | eol        |
+| ----------- | ---------- | ---------- |
+| [MySQL 8.0] | 2018-04-19 | 2026-04    |
+| [MySQL 5.7] | 2015-10-21 | 2023-10    |
+| [MySQL 5.6] | 2013-02-05 | 2021-02-05 |
 
 - [MySQL 5.7 Release Notes](https://dev.mysql.com/doc/relnotes/mysql/5.7/en/)
+
+## MySQL 8.0
+
+## MySQL 5.7
+
+## MySQL 5.6
diff --git a/notes/dev/compression/README.md b/notes/dev/compression/README.md
@@ -0,0 +1,8 @@
+---
+title: 压缩
+---
+
+# 压缩
+
+- [Awesome](./compression-awesome.md)
+- [FAQ](./compression-faq.md)
diff --git a/notes/dev/compression/compression-awesome.md b/notes/dev/compression/compression-awesome.md
@@ -0,0 +1,164 @@
+---
+title: Compression
+tags:
+  - Awesome
+---
+
+# Compression
+
+- [google/zopfli](https://github.com/google/zopfli)
+  - Zopfli Compression Algorithm is a compression library programmed in C to perform very good, but slow, deflate or zlib compression.
+- [google/brotli](https://github.com/google/brotli)
+  - 通用无损压缩
+  - combination of a modern variant of the LZ77 algorithm, Huffman coding and 2nd order context modeling, with a compression ratio comparable to the best currently available general-purpose compression methods
+  - 速度与 deflate 相似, 但压缩率更高
+  - [kothar/brotli-go](https://github.com/kothar/brotli-go)
+    - Go bindings for the Brotli compression library
+- [google/draco](https://github.com/google/draco)
+  - Draco is a library for compressing and decompressing 3D geometric meshes and point clouds. It is intended to improve the storage and transmission of 3D graphics.
+- [dsnet/compress](https://github.com/dsnet/compress)
+  - Collection of compression related Go packages.
+- [facebook/zstd](https://github.com/facebook/zstd)
+  - [zstd](http://facebook.github.io/zstd/)
+  - Zstandard - Fast real-time compression algorithm
+- [peazip/PeaZip](https://github.com/peazip/PeaZip)
+  - LGPLv3
+  - 跨平台
+- [mcmilk/7-Zip-zstd](https://github.com/mcmilk/7-Zip-zstd)
+  - LGPL v2.1-or-later
+  - 7-Zip + Brotli, Fast-LZMA2, Lizard, LZ4, LZ5, Zstandard
+  - Windows
+  - [p7zip-project/p7zip](https://github.com/p7zip-project/p7zip)
+    - Linux, macOS
+- Benchmark
+  - [Squash Compression Benchmark](https://quixdb.github.io/squash-benchmark/)
+  - [Large Text Compression Benchmark](http://mattmahoney.net/dc/text.html)
+  - [Gzip vs Bzip2 vs LZMA vs XZ vs LZ4 vs LZO](https://catchchallenger.first-world.info/wiki/Quick_Benchmark:_Gzip_vs_Bzip2_vs_LZMA_vs_XZ_vs_LZ4_vs_LZO)
+- JS
+  - [nodeca/pako](https://github.com/nodeca/pako)
+    - high speed zlib port to javascript, works in browser & node.js
+- [pigz](https://zlib.net/pigz/) - 并行 Gzip - 命令兼容 gzip
+- [MacPaw/XADMaster](https://github.com/MacPaw/XADMaster)
+  - LGPL-2.1
+  - Zip, Tar, Gzip, Bzip2, 7-Zip, Rar, LhA, StuffIt, CAB, LZX
+  - unar
+  - [MacPaw/unar](https://github.com/MacPaw/unar)
+  - https://theunarchiver.com/
+  - https://github.com/Homebrew/homebrew-core/blob/master/Formula/unar.rb
+  - https://theunarchiver.com/command-line
+- unrar
+  - Alpine 3.14 有
+  - https://www.rarlab.com/download.htm
+
+```bash
+curl -O https://www.rarlab.com/rar/rarlinux-x64-612.tar.gz
+tar zxvf rarlinux-x64-612.tar.gz
+./rar/unrar
+```
+
+- 参考
+  - [Modern LZ Compression](https://glinscott.github.io/lz/index.html)
+  - [MiloszKrajewski/SilesiaCorpus](https://github.com/MiloszKrajewski/SilesiaCorpus)
+  - [hxhb/oodle-compression](https://github.com/hxhb/oodle-compression)
+    - http://www.radgametools.com/oodle.htm
+    - 压缩慢，解压快，压缩率高
+  - [List of archive formats](https://en.wikipedia.org/wiki/List_of_archive_formats)
+  - [Comparison of file archivers](https://en.wikipedia.org/wiki/Comparison_of_file_archivers)
+
+:::tip
+
+- 文件归档 区别于 压缩算法 - 归档 -> 容器
+- gzip 头包含 mtime，因此压缩结果是不一致的
+
+:::
+
+---
+
+- 7z
+  - https://documentation.help/7-Zip/
+
+```bash
+cd staged && mytar cf - . | ssh root@localhost "cd / && tar xvf -"
+
+# 7z
+# https://github.com/p7zip-project/p7zip
+apk add p7zip
+7z a dir.7z -m0=zstd -mx7 dir
+
+# bz2
+# ==========
+# 解压
+bzip2 -d filename.bz2
+# 解压并保留原始压缩包
+bzip2 -dk filename.bz2
+
+# 并行 Gzip
+# http://zlib.net/pigz/
+# 提亚提升不大, 因为解压的结构不适用于并发
+# 支持 -11 zopfli
+# 利用 pv 可以查看压缩进度
+pv dict.db | pigz -vfk9 > dict.db.gz
+
+# 可以直接替换
+ln -s /usr/bin/pigz /usr/local/bin/gzip
+ln -s /usr/bin/unpigz /usr/local/bin/gunzip
+
+# 强制覆写, 暴露原始文件, 9 级压缩
+gzip -vfk9 data.db > data.db.gz
+
+# -w 0 自动选择优化的窗口大小
+# -9 压缩率, 如果不指定, 默认为 -Z 即 11, 会非常慢
+pv dict.db | brotli -9 -w 0 > dict.db.br
+
+# 编码
+# ====
+# https://www.iana.org/assignments/character-sets/character-sets.xhtml
+unzip -O cp936 -l my.zip # 使用指定的编码, 有些 unzip 不一定有该选项 - unzip-iconv
+# 也可以先解压, 然后再转换
+LANG=C 7za x your-zip-file.zip
+convmv -f GBK -t utf8 --notest -r .
+# brew install unar
+unar -e gb18030 gb18030.zip
+
+# 批量
+# ====
+unzip \*.zip
+```
+
+# FAQ
+
+## 解压 7z exe
+
+```bash
+ls *.exe | xargs -n 1 -d "\n" 7z x
+```
+
+## Multi part
+
+```bash
+# 1. merge
+cat in.zip* > ~/in.zip
+unzip in.zip
+
+# 2. 7z
+7z x in.zip.001
+```
+
+## Password
+
+```bash
+7z -mhc=on -mhe=on -pPASSWORD a out.7z in.txt
+
+zip -P password -r encrypted.zip folderIWantToZip
+unzip -P password encrypted.zip
+```
+
+## zstd level
+
+- 默认 3
+- 可以考虑 6-9
+- 影响 window size
+  - level 1 = 2^18 = 256KB
+  - zlib 最大 32KB
+- 1-22
+  - https://github.com/facebook/zstd/blob/dev/lib/compress/clevels.h
diff --git a/notes/dev/compression/compression-faq.md b/notes/dev/compression/compression-faq.md
@@ -0,0 +1,28 @@
+---
+tags:
+  - FAQ
+---
+
+# Compression Format FAQ
+
+## gzip vs zlib
+
+| -                     | gzip      | deflate/zlib |
+| --------------------- | --------- | ------------ |
+| Header size           | 10 bytes  | 2 bytes      |
+| Footer size           | 4 bytes   | 0            |
+| Checksum              | CRC32     | Adler-32     |
+| Compression algorithm | DEFLATE   | DEFLATE      |
+| Specification         | [RFC1952] | [RFC1950]    |
+
+- [rfc1951] deflate 算法
+
+[rfc1950]: https://datatracker.ietf.org/doc/html/rfc1950
+[rfc1951]: https://datatracker.ietf.org/doc/html/rfc1951
+[rfc1952]: https://datatracker.ietf.org/doc/html/rfc1952
+
+- content-encoding deflate -> zlib
+  - 为什么大多数网站使用 gzip
+    - 因为早期 MS Server 会直接发送 deflate 数据，而不是 zlib - deflate 在这里有一点混乱
+  - https://stackoverflow.com/a/9186091/1870054
+- https://stackoverflow.com/a/68538037/1870054