[U] Update channel content

This commit is contained in:
github-actions
2025-04-09 18:13:59 +00:00
parent 2167c79459
commit 0c42005e93
4 changed files with 6 additions and 6 deletions
+1 -1
View File
@@ -2,7 +2,7 @@
<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>2025-04-09T16:11:31.380151+00:00</updated>
<updated>2025-04-09T18:13:54.876528+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>
File diff suppressed because one or more lines are too long
+3 -3
View File
@@ -70680,7 +70680,7 @@
{
"id": 5112,
"date": "2025-04-01T15:47:46",
"views": 1596,
"views": 1598,
"forwards": 26,
"images": [
{
@@ -70729,7 +70729,7 @@
"id": 5115,
"date": "2025-04-02T15:11:19",
"text": "今日热搜",
"views": 401,
"views": 403,
"forwards": 6,
"images": [
{
@@ -71420,7 +71420,7 @@
"id": 5168,
"date": "2025-04-09T00:50:00",
"text": "图像生成榜上又有新的很不错的开源模型 HiDream I1 了,不过这个模型居然比 Flux1 还重,40GB 显存都跑不起来...\n\n花了一天用了各种办法终于让它能只用 16GB 显存跑起来了!然后打包了一下发到了 pypi 上可以直接 pip 安装了 qwq\n\n(具体是用了 bnb nf4 量化,之后开 sequential offloading 让它在不用的时候把权重切回 cpu,然后文字理解部分它原本用了 llama 3.1 8b 我换成了 gptq 4bit 版本并且用 device map 让其中不需要在显卡上的部分卸载到 cpu 了)\n\n仓库在这里: <a href=\"https://github.com/hykilpikonna/HiDream-I1-nf4\">https://github.com/hykilpikonna/HiDream-I1-nf4</a>",
"views": 80,
"views": 83,
"forwards": 0,
"media_group_id": 13953278404348069,
"images": [
+1 -1
View File
@@ -12,7 +12,7 @@
<link>https://aza.moe/life</link>
</image>
<language>zh-cn</language>
<lastBuildDate>Wed, 09 Apr 2025 16:11:31 +0000</lastBuildDate>
<lastBuildDate>Wed, 09 Apr 2025 18:13:54 +0000</lastBuildDate>
<item>
<title>小桂桂的回忆录 📒 #5168</title>
<link>https://aza.moe/life?post=5168</link>