This Tweet is unavailable.
1
8
31
Follow
Click to Follow pcuenq
Pedro Cuenca
@pcuenq
ML Engineer at 
Hugging Face | Co-founder at LateNiteSoft (Camera+). I love AI and photography.
Hugging Face | Co-founder at LateNiteSoft (Camera+). I love AI and photography.SpainJoined May 2019
Pedro Cuenca’s Tweets
Some team members (including myself!) will be in the Bay Area from Feb 22 until the 28th. Who would like to meet? Should we do an informal "meetup" at Stanford? :D
7
2
44
read image description
ALT
1
5
55
- 1st place: Hugging Face merch of your choice + $500 Hugging Face compute credits
- 2nd place: Hugging Face merch of your choice + $200 Hugging Face compute credits
- 3rd place: Hugging Face socks + $50 Hugging Face compute credits 🚀
2
9
37
🙌 Credits 🙌
CompVis, and for Stable Diffusion.
Apple, for their amazing conversion and inference library: github.com/apple/ml-stabl
The Diffusers library : github.com/huggingface/di
OFA-Sys for their distilled model:
2
2
31
Show this thread
And, of course, it's open source!
Check out the repo: github.com/huggingface/sw
💓 Pull Requests are more than welcome, there's so much to do! Looking forward to including image-to-image, personal galleries, additional models, and more. Feel free to fork and contribute!
2
4
59
Show this thread
🔒 Complete privacy. Images are generated in your computer.
⬇️ Downloadable models. Optimized for CPU, GPU and Neural Engine.
🏎️ Fast, native app with Mac look and feel.
💧 Smaller distilled model by OFA-Sys available for not-so-powerful Macs.
⚙️ Flexible and fun.
2
2
50
Show this thread
🧨 Diffusers for Mac has just been released in the Mac App Store! Run Stable Diffusion easily on your Mac with our native and open-source Swift app 🚀
Being the Mac counterpart of the diffusers library, it allows you to download models from the Hub, in optimized Core ML format.
33
208
1,208
Show this thread
#diffuserslib v0.12 released with super cool 😎 features #instructpix2pix #DiT #LoRA and more 🎉 when I looked at the release note, I saw a super kind note 😳 Thank you so much ❤ team 🥰
2
19
Short sweet and full of useful info! Perfectly timed too - we've been realizing this as we experiment with noise schedules and input scaling for the latest lesson (recording with in a few hours!) 😀
Thank you , great work! twitter.com/_akhaliq/statu
This Tweet is unavailable.
1
8
46
LoRA DreamBooth Training Web UI
train and share your dreamboothed Stable Diffusion model in just a 3.3 MB file on the Lora library on
demo: huggingface.co/spaces/lora-li
4
62
262
Check the details in our blog post or docs, and get in touch with any questions. We'd love to see what you build!
3
23
Show this thread
🐏 Less RAM
💻 Less compute
🚀 Faster training
🏋️ Slim weights!
This enables new use cases for the community: share your models easily, quickly switch styles in production, use less resources for download and storage, ...
1
1
20
Show this thread
The good news is that it can be applied to Stable Diffusion too!
did it first, and now we brought it to diffusers
1
2
22
Show this thread
LoRA was proposed by Microsoft Research to deal with efficient fine-tuning of Large Language Models, by freezing everything but injecting small trainable layers in strategic places. It can achieve similar results to fine-tuning, but with 3 or 4 orders of magnitude less weights 🤯
2
1
22
Show this thread
What if you could share your dreamboothed or fine-tuned Stable Diffusion model in just a 3.3 MB file?
You can now, with LoRA
25
139
689
Show this thread
I'm happy to announce Hugging Face Competitions! With 🤗 Competitions, you can create public/private competitions with full control of datasets, metrics and evaluation! Check out our first competition: AI or Not huggingface.co/spaces/competi 1/N
19
121
471
Show this thread
DIffusers 0.12.0 is out with new magical features 🪄
📷InstructPix2Pix Edit images w/ language instructions
💡LoRA for fast & mem-efficient #stablediffusion fine-tuning
🚀New fast DEIS scheduler
🤝Transformer meets Diffusion w/ DiT
✨New API to easily customize attention layers
3
51
246
Show this thread
If anyone was affected by the BigTech layoffs and is ready to look for a new job: at we're still hiring for all kinds of roles (from developer advocacy to ML engineering). We're looking for growth and generalist mindsets and people excited to work in OS ML. Ping me🤗
10
56
384
The Instruct2Pix2Pix official code and model weights were released *yesterday*. Enjoy the full diffusers treatment today: new pipeline, docs, demo and Colab.
huggingface.co/docs/diffusers
Quote Tweet
The InstructPix2Pix #stablediffusion model, which allows editing images using language instructions, is now available in diffusers!
Colab with @Gradio demo:
colab.research.google.com/github/hugging
Read more about the model here: timothybrooks.com/instruct-pix2p
7
32
Excited to share what I've been working on for the past 6 months: Text-to-Audio Diffusion.
More samples and info here: bit.ly/audio-diffusion
Enjoy!
🔊 Prompt: "Dubstep Insane Drop Remix (Deluxe Edition), 2 of 4"
1:27
191.3K views
281
1,023
2,306
Woohoo 🥳 my PR merged to ✨ now we can convert Flax models to PyTorch 🔥
thanks to 🤗 patrickvonplaten ❤ ❤ and ❤ for teaching me lots of things 🥰 I will soon add the converter here:
1
3
18
10 years without this face, without this voice, without this mind, without this love, without this light.
Ten years.
RIP my darling boy.
298
2,861
19.9K
The future of generative art is *personalization*. Rather than "pic of a kitten at a party", we may want "pic of MY kitten at MY party".
New paper enables us to quickly teach Stable Diffusion many new concepts at once, and combine them on the fly: arxiv.org/abs/2212.04488
1/2
31
138
861
Show this thread
🖼️ Listen to explain diffusion!
In this week's Gradient Dissent, he also discusses large models with , revisits an old conversation on Python vs. Julia, and talks about his scientific advocacy during the early days of COVID-19.
✨ wandb.me/gd-jeremy-howa
18
87
Results from the Lambda x 🤗 sprint to fine-tune 's Whisper model:
- 20K+ A100 hours provided for community use via Lambda Cloud credits
- 100+ languages covered
- 650 models created
Beyond impressed by the community. Giveaway announcement coming Friday!
16
123
Are you into both Game Dev and AI? Check out my latest blog post: AI for Game Development
8
58
274
Show this thread
We're 1 week into the DreamBooth Hackathon and we've already got some great submissions across the 5 themes of 🐨🔬🍔🏔🔥!
You can view all the models on the Hub and give them a ❤️: huggingface.co/models?other=d
Here's a 🧵on some of the models that made me 🤯
1/7
1
6
31
Show this thread
Running Stable Diffusion on iPhone 13 Pro in 23s 🤯
Code coming soon to github.com/huggingface/sw (built on top of Apple's Core ML stable diffusion library).
13
65
496
Unit 3 of the 🤗 diffusers class is out! If you've been waiting to learn about #stablediffusion this is your chance!
Learn how to use SD via pipelines, deep dive into the individual components, and train a custom DreamBooth model for the hackathon:
github.com/huggingface/di
read image description
ALT
read image description
ALT
5
35
178
Show this thread
I just learned that this uses the diffusers library under the hood. It’s amazing to see what people can create!
Quote Tweet
#StableDiffusion2 depth to image can texture entire scenes in Blender automatically.
Try it for yourself: github.com/carson-katri/d
0:42
236.9K views
1
13
65
📣 CRFM announces PubMedGPT, a new 2.7B language model that achieves a new SOTA on the US medical licensing exam. The recipe is simple: a standard Transformer trained from scratch on PubMed (from The Pile) using on the MosaicML Cloud, then fine-tuned for the QA task.
40
391
1,556
Show this thread
Requires the latest betas: macOS Ventura 13.1, Xcode 14.2, iOS/iPadOS 16.2. They are all in Release Candidate state, so they should be released very soon!
PRs welcome! Or just use the source code for inspiration and build something great 🎈
2
2
17
Show this thread
We are open sourcing a native Swift app for Stable Diffusion!
github.com/huggingface/sw
It integrates Apple's Core ML library (released last week) and the fastest scheduler (DPM-Solver++), which we ported to Swift from diffusers!
Generation time: ~10s in my M1 Max 🚀
9
51
310
Show this thread
1) YAAY EMNLP has recognized our demo of measurement of data and models!! 🤗🤗 WTG & everyone in between!
2) ...But we're still not distinguishing between industry demo and grad student demo? Kinda stacks the odds against students. 😕 twitter.com/douwekiela/sta
This Tweet is unavailable.
5
8
75
We partnered w/ W&B on this course that simulates a real-world process of iterating on a model as a team.
- Uses for image segmentation
- We show you how to use fastai w/ Sweeps, Reports, Tables, and a Model Registry in addition to experiment tracking. twitter.com/dk21/status/16
This Tweet is unavailable.
3
32
123
Show this thread