[U] Update channel content
This commit is contained in:
github-actions
+11
-1
@@ -2,11 +2,21 @@
|
||||
<feed xmlns="http://www.w3.org/2005/Atom" xml:lang="zh-cn">
|
||||
<id>https://aza.moe/life</id>
|
||||
<title>小桂桂的回忆录 📒</title>
|
||||
<updated>2024-07-10T08:11:20.387577+00:00</updated>
|
||||
<updated>2024-07-10T10:10:28.149276+00:00</updated>
|
||||
<link href="https://aza.moe/life" rel="alternate"/>
|
||||
<generator uri="https://lkiesow.github.io/python-feedgen" version="0.9.0">python-feedgen</generator>
|
||||
<logo>https://aza.moe/meru_256px.png</logo>
|
||||
<subtitle>「我们所经历的每个平凡的日常，也许就是连续发生的奇迹」</subtitle>
|
||||
<entry>
|
||||
<id>4115</id>
|
||||
<title>小桂桂的回忆录 📒 #4115</title>
|
||||
<updated>2024-07-10T10:05:26+00:00</updated>
|
||||
<content><p>训练好了 Sekai LLM 之后现在正在训练 Sekai TTS，目前训练到第一代 250 epoch，效果意外不错 <i class="custom-emoji" emoji-src="emoji/6170139473288890346.webm">🤩</i></p>
|
||||
<p>开放了一个测试网页大家可以来玩玩 <a href="https://sekai-test.aquadx.net/">https://sekai-test.aquadx.net/</a></p>
|
||||
<p>架构用的是 2021 年的 <a href="https://arxiv.org/abs/2106.06103">VITS</a>，训练用了 <a href="https://github.com/Plachtaa/VITS-fast-fine-tuning">Plachtaa 的微调仓库</a>，把 pjsk 拆好的资源下下来把文本音频对应上之后就开始训练啦，都是别人走过的路。第一代用了每人 200 个音频，看起来有些数据集外的音频听起来怪怪的，下一代打算把游戏里的全部 210308 条音频都喂进去（虽然那样应该会需要训练更久</p>
|
||||
<p>以及感谢 menci 借给我 4090 机器 qwq</p></content>
|
||||
<link href="https://aza.moe/life?post=4115" rel="alternate"/>
|
||||
</entry>
|
||||
<entry>
|
||||
<id>4113</id>
|
||||
<title>小桂桂的回忆录 📒 #4113</title>
|
||||
|
||||
File diff suppressed because one or more lines are too long
Binary file not shown.
|
After Width: | Height: | Size: 65 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 20 KiB |
+29
-10
@@ -55910,14 +55910,14 @@
|
||||
"id": 4098,
|
||||
"date": "2024-07-04T21:36:28",
|
||||
"text": "在加拿大自杀报警会被带到医院急诊，等待心理医生过来进行诊断，判断是不是需要在精神科医院住院或者其他的治疗，不需要就可以回去了\n\n在中国自杀报警会被带到派出所，等熟人来，签文件盖指纹录一个视频，接下来出什么事就都是这个人的责任了",
|
||||
"views": 98,
|
||||
"views": 99,
|
||||
"forwards": 1
|
||||
},
|
||||
{
|
||||
"id": 4099,
|
||||
"date": "2024-07-05T15:48:35",
|
||||
"text": "拍到了睡着的蝴蝶 <i class=\"custom-emoji\" emoji-src=\"emoji/6170139473288890346.webm\">🤩</i>",
|
||||
"views": 74,
|
||||
"views": 75,
|
||||
"forwards": 0,
|
||||
"media_group_id": 13761556124439541,
|
||||
"images": [
|
||||
@@ -55957,7 +55957,7 @@
|
||||
"id": 4102,
|
||||
"date": "2024-07-05T17:24:36",
|
||||
"text": "折千纸鹤",
|
||||
"views": 77,
|
||||
"views": 78,
|
||||
"forwards": 0,
|
||||
"forwarded_from": {
|
||||
"name": "Azalea",
|
||||
@@ -55980,7 +55980,7 @@
|
||||
"id": 4103,
|
||||
"date": "2024-07-06T19:20:37",
|
||||
"text": "发一些最近可能有趣的照片",
|
||||
"views": 75,
|
||||
"views": 76,
|
||||
"forwards": 1,
|
||||
"media_group_id": 13762349096239221,
|
||||
"images": [
|
||||
@@ -56040,21 +56040,21 @@
|
||||
"id": 4108,
|
||||
"date": "2024-07-07T20:17:04",
|
||||
"text": "下楼散散心，遇到了猫猫但是没带猫粮\n\n上楼拿了猫粮放在同一个位置等，结果猫猫没喂到反而喂饱了很多蚊子 <i class=\"custom-emoji\" emoji-src=\"emoji/6170293817233638918.webm\">😐</i>",
|
||||
"views": 68,
|
||||
"views": 69,
|
||||
"forwards": 0
|
||||
},
|
||||
{
|
||||
"id": 4109,
|
||||
"date": "2024-07-07T20:59:17",
|
||||
"text": "世界通用语言：emoji 🤗",
|
||||
"views": 69,
|
||||
"views": 70,
|
||||
"forwards": 0
|
||||
},
|
||||
{
|
||||
"id": 4110,
|
||||
"date": "2024-07-08T07:33:46",
|
||||
"text": "日出日落作息：日出睡觉，日落起床 <i class=\"custom-emoji\" emoji-src=\"emoji/6170358057059485759.webm\">😪</i>",
|
||||
"views": 66,
|
||||
"views": 68,
|
||||
"forwards": 0,
|
||||
"images": [
|
||||
{
|
||||
@@ -56073,7 +56073,7 @@
|
||||
"id": 4111,
|
||||
"date": "2024-07-08T14:24:46",
|
||||
"text": "syama 精神状态 be like...",
|
||||
"views": 63,
|
||||
"views": 65,
|
||||
"forwards": 0,
|
||||
"images": [
|
||||
{
|
||||
@@ -56093,7 +56093,7 @@
|
||||
"id": 4112,
|
||||
"date": "2024-07-08T23:07:46",
|
||||
"text": "人生第一次训练大语言模型 <i class=\"custom-emoji\" emoji-src=\"emoji/6170139473288890346.webm\">🤩</i>",
|
||||
"views": 58,
|
||||
"views": 61,
|
||||
"forwards": 0,
|
||||
"images": [
|
||||
{
|
||||
@@ -56112,7 +56112,7 @@
|
||||
"id": 4113,
|
||||
"date": "2024-07-09T09:47:13",
|
||||
"text": "训练了一天 Qwen2-7B pretrain lora 之后似乎可以和 pjsk 的大家说话了！\n\n（虽然还有很多要优化，但是可行性通过了 quq",
|
||||
"views": 38,
|
||||
"views": 43,
|
||||
"forwards": 0,
|
||||
"media_group_id": 13764147471161677,
|
||||
"images": [
|
||||
@@ -56137,5 +56137,24 @@
|
||||
"thumb": "media/4114.jpg_thumb.jpg"
|
||||
}
|
||||
]
|
||||
},
|
||||
{
|
||||
"id": 4115,
|
||||
"date": "2024-07-10T10:05:26",
|
||||
"text": "训练好了 Sekai LLM 之后现在正在训练 Sekai TTS，目前训练到第一代 250 epoch，效果意外不错 <i class=\"custom-emoji\" emoji-src=\"emoji/6170139473288890346.webm\">🤩</i>\n\n开放了一个测试网页大家可以来玩玩 <a href=\"https://sekai-test.aquadx.net/\">https://sekai-test.aquadx.net/</a>\n\n架构用的是 2021 年的 <a href=\"https://arxiv.org/abs/2106.06103\">VITS</a>，训练用了 <a href=\"https://github.com/Plachtaa/VITS-fast-fine-tuning\">Plachtaa 的微调仓库</a>，把 pjsk 拆好的资源下下来把文本音频对应上之后就开始训练啦，都是别人走过的路。第一代用了每人 200 个音频，看起来有些数据集外的音频听起来怪怪的，下一代打算把游戏里的全部 210308 条音频都喂进去（虽然那样应该会需要训练更久\n\n以及感谢 menci 借给我 4090 机器 qwq",
|
||||
"views": 13,
|
||||
"forwards": 1,
|
||||
"images": [
|
||||
{
|
||||
"width": 1888,
|
||||
"height": 880,
|
||||
"date": "2024-07-10T10:05:25",
|
||||
"media_type": "photo",
|
||||
"original_name": "photo_2024-07-10_10-05-25.jpg",
|
||||
"url": "media/4115.jpg",
|
||||
"size": 66532,
|
||||
"thumb": "media/4115.jpg_thumb.jpg"
|
||||
}
|
||||
]
|
||||
}
|
||||
]
|
||||
+10
-1
@@ -12,7 +12,16 @@
|
||||
<link>https://aza.moe/life</link>
|
||||
</image>
|
||||
<language>zh-cn</language>
|
||||
<lastBuildDate>Wed, 10 Jul 2024 08:11:20 +0000</lastBuildDate>
|
||||
<lastBuildDate>Wed, 10 Jul 2024 10:10:28 +0000</lastBuildDate>
|
||||
<item>
|
||||
<title>小桂桂的回忆录 📒 #4115</title>
|
||||
<link>https://aza.moe/life?post=4115</link>
|
||||
<description><p>训练好了 Sekai LLM 之后现在正在训练 Sekai TTS，目前训练到第一代 250 epoch，效果意外不错 <i class="custom-emoji" emoji-src="emoji/6170139473288890346.webm">🤩</i></p>
|
||||
<p>开放了一个测试网页大家可以来玩玩 <a href="https://sekai-test.aquadx.net/">https://sekai-test.aquadx.net/</a></p>
|
||||
<p>架构用的是 2021 年的 <a href="https://arxiv.org/abs/2106.06103">VITS</a>，训练用了 <a href="https://github.com/Plachtaa/VITS-fast-fine-tuning">Plachtaa 的微调仓库</a>，把 pjsk 拆好的资源下下来把文本音频对应上之后就开始训练啦，都是别人走过的路。第一代用了每人 200 个音频，看起来有些数据集外的音频听起来怪怪的，下一代打算把游戏里的全部 210308 条音频都喂进去（虽然那样应该会需要训练更久</p>
|
||||
<p>以及感谢 menci 借给我 4090 机器 qwq</p></description>
|
||||
<guid isPermaLink="false">4115</guid>
|
||||
</item>
|
||||
<item>
|
||||
<title>小桂桂的回忆录 📒 #4113</title>
|
||||
<link>https://aza.moe/life?post=4113</link>
|
||||
|
||||
Reference in New Issue
Block a user