Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
20230322 - Generative AI 小聚 ft. Happy Designer
Search
Sponsored
·
Your Podcast. Everywhere. Effortlessly.
Share. Educate. Inspire. Entertain. You do you. We'll handle the rest.
→
蒼時弦や
March 22, 2023
Technology
480
0
Share
Embed
Copy iframe code
Copy JS code
Copy link
Start on current slide
20230322 - Generative AI 小聚 ft. Happy Designer
TLDR
蒼時弦や
March 22, 2023
More Decks by 蒼時弦や
See All by 蒼時弦や
2024 - COSCUP - Clean Architecture in Rails
elct9620
2
210
2023 - RubyConfTW - Rethink Rails Architecture
elct9620
0
260
20230916 - DDDTW - 導入 Domain-Driven Design 的最佳時機
elct9620
0
500
2023 - WebConf - 選擇適合你的技能組合
elct9620
0
720
2022 - 默默會 - 重新學習 MVC 的 Model
elct9620
1
520
MOPCON 2022 - 從 Domain-Driven Design 看網站開發框架隱藏
elct9620
1
540
2022 - COSCUP - 我想慢慢寫程式該怎麼辦?
elct9620
0
310
2022 - COSCUP - 打造高速 Ruby 專案開發流程
elct9620
0
340
2021 - RubyKaigi - It is time to build your mruby VM on the microcontroller?
elct9620
0
310
Other Decks in Technology
See All in Technology
AI研修(Day1)【MIXI 26新卒技術研修】
mixi_engineers
PRO
2
3.2k
[MIRU26] To What Extent Does MLLM-as-a-Judge Exhibit Cross-Model Preference Bias?
keio_smilab
PRO
0
100
クラウドを使う側から、作る側へ / 大吉祥寺.pm 2026前夜祭
fujiwara3
8
2k
新しい SLO が良い感じにハマっている話
z63d
4
1.9k
toio・myCobotでフィジカルAIっぽいことを行うための検討(とりあえず調査) / フィジカルAI LT(IoTLTによる開催)
you
PRO
0
220
Claude Mythos、Fable...フロンティアAIの最新動向と企業のセキュリティ対策
flatt_security
0
180
AIエージェントの知識表現と推論に なぜグラフが使われるのか - 記号的AIの復権とニューラルAIとの統合
yohei1126
1
250
PLaMo 3.0 Primeの構造化出力サポート
pfn
PRO
0
160
テックカンファレンス三大ステークホルダーの文化人類学 ─ 違いを認め合う関係性作り
bash0c7
5
1.5k
「待ち時間」の消滅と「自我消耗」の加速:生成AI時代のエンジニアを救うメンタル・リソース管理
poropinai1966
0
400
もう一度考える SRE チームの作り方・育て方 / Rethinking SRE #1: Building and Growing SRE Teams
rrreeeyyy
4
620
現場で使える AWS DevOps Agent 活用ノウハウ - Release Management 機能の検証結果を添えて / AWS DevOps Agent Release Management and Know-How
kinunori
5
840
Featured
See All Featured
The Spectacular Lies of Maps
axbom
PRO
1
880
Embracing the Ebb and Flow
colly
88
5.1k
Designing for Performance
lara
611
70k
Google's AI Overviews - The New Search
badams
0
1.1k
For a Future-Friendly Web
brad_frost
183
10k
Getting science done with accelerated Python computing platforms
jacobtomlinson
2
370
Introduction to Domain-Driven Design and Collaborative software design
baasie
1
920
Leo the Paperboy
mayatellez
8
2k
The Cost Of JavaScript in 2023
addyosmani
55
10k
[RailsConf 2023] Rails as a piece of cake
palkan
59
6.8k
Darren the Foodie - Storyboard
khoart
PRO
3
3.5k
個人開発の失敗を避けるイケてる考え方 / tips for indie hackers
panda_program
123
22k
Transcript
除了串 API 之外工程師還能做些什麼 Generative AI Meetup ft. Happy Designer
蒼時弦也 Software Developer https://blog.aotoki.me
開始之前 生態系進化太快,都不知道該怎麼辦⋯⋯
Diffusers 簡單使用 Stable Diffusion 的方法
Hugging Face AI 領域的 GitHub,大部分開源模型都可以在上面找到
import from import = = = = = torch diffusers
StableDiffusionPipeline pipe StableDiffusionPipeline.from_pretrained( , torch.float16) pipe pipe.to( ) prompt image pipe(prompt).images[ ] # 載入模型 # 進行運算 "runwayml/stable- diffusion-v1-5" "cuda" "a photo of an astronaut riding a horse on mars" torch_dtype 0
# ... # 載入不同模型 # 多重處理 prompt image txt2image(prompt).images[ ]
image upscale(image).images[ ] image image2image(image).images[ ] = = = = "a photo of an astronaut riding a horse on mars" 0 0 0
None
http://bit.ly/3JJf16o
Replicate 目前最好用的 Serverless GPU 服務
None
from import import class def def = = = =
= return cog BasePredictor, Path, Input torch ( ): (self): (self, image: Path Input( ), scale: Input( , ) ) -> Path: output Predictor BasePredictor setup predict # 初始化、載入模型 # 處理 e.g. Stable Diffusion Pipeline description description default "Image to enlarge" "Factor to scale image by" float 1.5
$ $ cog login cog push r8.im your username your
model name /< - >/< - - >
有 GPU 的伺服器、API 都有了,剩下就是套應用
AI Gacha 使用 Diffusers、Replicate 的實驗性專案
None
http://bit.ly/3yImmgc
我的看法 能做的事情其實不多,大部分應用的成本也偏高,但是商用已經沒 有問題,剩下要處理的大概會在 跟讓更多人能 合法的版權 輕鬆上手