robocrunch
Julien Chaumond
@julien_c
Co-founder and CTO at @huggingface 🤗. Natural language processing, ML/AI for everyone, building products to propel communities fwd. @Stanford + @Polytechnique
Tweets by Julien Chaumond
@krasul and @NielsRogge are 100% the best people on time-series (w or without transformers)
Shared by
Julien Chaumond
at
5/11/2022
I love those sections on the experimental nitty gritty of LLM training. I think they could be titled "Is this Machine learning, or Alchemy" 😅 From the OPT paper https://t.co/mw8ny2Yf6j by @stephenroller and team 🔥 PS/ Hopefully weights will be soon on @huggingface!
Shared by
Julien Chaumond
at
5/3/2022
And @mervenoyann is working on some Keras and TF training scripts!! 🔥🔥
Shared by
Julien Chaumond
at
4/6/2022
We just started our first ever ✨ Computer Vision ✨ community event. Register on Discord, or just jump straight to the really awesome material/code from @NielsRogge @_nateraw @osanseviero, at https://t.co/HM2EQsBDXI Hoping to upload a few models and demos too 🤞
Shared by
Julien Chaumond
at
4/5/2022
This is super, super cool: Nix-TTS 🐤: An incredibly lightweight text-to-speech via non end-to-end distillation 📄 Paper: https://t.co/c2uuvFxBfd 📦 Repository: https://t.co/74Sa0botmF 🤗 Space: https://t.co/A6k9TLCW6Y
Shared by
Julien Chaumond
at
4/4/2022
The training of @BigScienceLLM aka. the largest *open source* language model has started a few days ago We track the training's progress in real-time on @huggingface homepage 🔥🔥 Let's see if that loss converges 👀 H/t @severo_dev @StasBekman and the whole @BigscienceW team!
Shared by
Julien Chaumond
at
3/16/2022
Hat/tip to our ML Optimization team @ellacharlaix @MorganFunto @Jhuaplin Michael B. with a lot of support from @_philschmid @_lewtun @jeffboudier 😎😎 Ping us for any feedback on Inference acceleration!
Shared by
Julien Chaumond
at
3/8/2022
Are you interested in Inference optimization? 🔥 Step 1: export your model to ONNX directly from transformers: https://t.co/fpkNozOCHZ 🏎 Step 2: Use our new repo Optimum https://t.co/GcgwzGzf2R to optimize your ONNX (operator fusion etc), including to a target hardware
Shared by
Julien Chaumond
at
3/8/2022
if you have any custom model implementation, it now takes 1 line of code to have a `from_pretrained` method that loads pretrained weights from https://t.co/BW2WwWwU96 ❤️ Doc page: https://t.co/syIoDUwv0P cc @GuggerSylvain @LysandreJik 💙
Shared by
Julien Chaumond
at
2/22/2022
Phil Wang aka `lucidrains` wrote yet another awesome PyTorch implementation, this time of Enformer, Deepmind's model for predicting gene expression. Thanks to @NielsRogge we can now load the pretrained models from the huggingface hub directly from the Enformer codebase 🔥
Shared by
Julien Chaumond
at
2/22/2022
Had a great lunch with @mnlpariente creator of the Asteroid open source library for audio source separation 🔥 Fun fact: along with @ai2_allennlp Asteroid was the first third-party library that supported loading models from the @huggingface Hub 😎
Shared by
Julien Chaumond
at
12/23/2021
Great PyTorch intro + while at the same time a small peek into some internals of PyTorch by @marksaroufim
Shared by
Julien Chaumond
at
11/15/2021
Anyone knows what a Feature store is (in the context of ML) and can ELI5 it to me?
Shared by
Julien Chaumond
at
11/15/2021
It’s been one year to the day that we implemented git hosting for models and datasets on @huggingface 🎂1️⃣ Before that we had our own custom solution on top of S3. I think it was a really good decision to switch 🙂
Shared by
Julien Chaumond
at
11/10/2021
GAME-CHANGING FEATURE! So proud of the whole team for shipping this.
Shared by
Julien Chaumond
at
10/29/2021
if you try this new feature let us know of your feedback here
Shared by
Julien Chaumond
at
10/27/2021
Controlling a 3D model using natural language, built with @huggingface Inference API Deployed by community member @pierreant_p on HuggingFace Spaces 🔥🔥🔥
Shared by
Julien Chaumond
at
10/19/2021
personally a bit bored by 530 billion parameters models I'd be more interested in a model that can be used or researched by 530 million humans
Shared by
Julien Chaumond
at
10/11/2021
Congrats on the 1.0 release, @streamlit! 🔥🔥 And very timely blog from @mervenoyann on how to build cool Streamlit apps for dataset visualization or model inference, on https://t.co/BW2WwWfiKw ⤵️
Shared by
Julien Chaumond
at
10/5/2021
Woohoo! Some of my models are now compatible with @huggingface AutoNLP! 😍 Fine-tuning them on your own domain or task is now just a matter of minutes Try it out by clicking on "Train" on these model pages: https://t.co/SaDXXl17Qy
Shared by
Julien Chaumond
at
10/5/2021
🔥 Hub Update: We now validate the license identifiers you can use in your models & dataset cards. We also migrated the existing repos to those standardized identifiers, hat/tip @elias_showk 🙏. Full list of licenses is at https://t.co/0s5eU1vun9 ➡️we now have cleaner filter:
Shared by
Julien Chaumond
at
9/24/2021
We just added support for **Tensorboard for private models** 🔥 Just push your TensorBoard traces to any model repo (public or private) and we automatically spawn a TensorBoard server to visualize them. See latest models that include TensorBoard traces➡️ https://t.co/weJdTmjroI
Shared by
Julien Chaumond
at
9/13/2021
Reminder that if you're building a @gradio app (for instance on @huggingface Spaces) and computation time is long, you should add `enable_queue=True` to your gr.Interface() so user requests are queued (with a position indicator) 🔥 See screencap to see the UI in action
Shared by
Julien Chaumond
at
9/9/2021
📸🖼 Introducing HugsVision: an easy-to-use wrapper to @huggingface for Healthcare Computer Vision Tutorial on How to Train a Vision Transformer (ViT) Image Classifier to help endoscopists: https://t.co/rccNitC5t1 GitHub repo: https://t.co/JYOKfTSjuy hat/tip @ParcolletT
Shared by
Julien Chaumond
at
9/3/2021
🔥 Had a great time on the @mlopscommunity podcast recently chatting w/ @dpbrinkm on @aponteanalytics about upcoming features on the Hugging Face Model Hub =) You can check out the whole conversation at https://t.co/YEIFsQBHva Thanks for having me!
Shared by
Julien Chaumond
at
7/27/2021
Am i the only one on earth not using 42 as their random seed 😈
Shared by
Julien Chaumond
at
7/7/2021
I am SUPER excited about this. `transformers v4.8` = Seamless integration of training <> the @huggingface Hub 🔥 Upload your @TensorBoard traces and eval results to your model repo ♥️ https://t.co/T81uzbaCUF GIF courtesy of the incredible @LysandreJik and @GuggerSylvain ⤵️ h
Shared by
Julien Chaumond
at
6/25/2021
Twitter Sentiment Emotion Phrase Extractor - Flask + Pytorch + @huggingface A cool video by @1littlecoder https://t.co/BdAc2TpK8F
Shared by
Julien Chaumond
at
6/10/2021
you can now filter on _just_ the transformers models on https://t.co/AjudkByqJX
Shared by
Julien Chaumond
at
6/9/2021