Lorenzo Ampil

@AND__SO

Problem solving tech. + for life sciences, and finance. Modelling humor for fun. ML Product Manager & Data Scientist at .

MANILA
linkedin.com/in/lorenzoampi…
Menyertai Mac 2009
131 Foto atau video 131 Foto dan video Foto dan video

Tweet

Anda telah menyekat @AND__SO

Adakah anda pasti anda mahu melihat Tweet ini? Melihat Tweet tidak akan menyekat @AND__SO

  1. Tweet Dipinkan
    22 Apr

    Excited to share that my PR to add a text generation pipeline to Transformers has been merged! It will be included in a coming release😀 Soon, generating text with transformer models (e.g. GPT-2) will be possible with only 3 lines of code!

    6 balasan6 balasan 34 tweet semula34 tweet semula 196 suka
    Tunjukkan thread ini
    Tunjukkan thread ini
    Buat asal
  2. 6 jam lalu

    Was an absolute pleasure sharing my recent work on T5 in ’s webinar series!

    Yesterday, we had an amazing session by Lorenzo on T5 and its application on 's current twitter sentiment competition. If you missed it, you can watch the recording here: Thank you everyone who joined and stay tuned for more amazing talks!
    8 suka
    Buat asal
  3. 11 jam lalu

    I think I'm getting asked for resources enough to start writing up a blog post on a learning plan for specifically how to learn NLP from scratch 😀

    3 balasan3 balasan 23 suka
    Buat asal
  4. Mengetweet Semula
    24 Mei

    Using transformers models (BERT, XLNet, Bart, Electra, Roberta, XLM-Roberta) to predict next word or a masked word in a sentence. Just to quickly compare those models. Repo:

    12 balasan12 balasan 250 tweet semula250 tweet semula 1,035 suka
    Buat asal
  5. Mengetweet Semula
    23 Mei

    Below is a thread about the future of remote work after the COVID-19 pandemic is over. I predict that remote will go through a trough of sorrow due to hybrid not working out, and most companies will return to being office based. But many all remote companies will see success.

    76 balasan76 balasan 1,029 tweet semula1,029 tweet semula 3,944 suka
    Tunjukkan thread ini
    Tunjukkan thread ini
    Buat asal
  6. 22 Mei

    Amazing work guys!

    Extremely happy to have such an important line of research – modeling of extremely long sequences – make their way into 🤗 transformers! Amazing work led by 🚀🔥👏🏻
    Buat asal
  7. Mengetweet Semula
    22 Mei

    Going LIVE on YouTube with Lorenzo in 2 hours from now. He is going to talk about T5 and its application in sentiment span detection. Don't forget to join, learn and ask questions! :) Live link: Calendar: Also, SUBSCRIBE NOW!

    2 tweet semula2 tweet semula 11 suka
    Buat asal
  8. Mengetweet Semula
    19 Mei

    Happy to announce we partenered with to make state-of-the-art inference up to 5x faster 🚀. NLP for every people and organizations!

    3 balasan3 balasan 118 tweet semula118 tweet semula 476 suka
    Buat asal
  9. 19 Mei

    Multi-task learning is turning into the main NLP buzz word in my head

    3 suka
    Buat asal
  10. 19 Mei

    If you're free this May 22 (Friday) 5PM CET, and are interested, do watch me discuss about how to use T5 for sentiment span extraction in 's Talks # 3! 😀 Youtube Link: Google Calendar reminder link:

    7 suka
    Buat asal
  11. 18 Mei

    I am going back to my habit of listening to an podcast episode every morning.

    2 suka
    Buat asal
  12. Mengetweet Semula
    18 Mei

    I’m so fucking long human ingenuity. 🤞🏽🤞🏽🤞🏽

    1:03
    BREAKING: Moderna announces Phase 1 results: After two doses all patients in the trial developed antibodies for the virus reports
    168 balasan168 balasan 432 tweet semula432 tweet semula 3,857 suka
    Buat asal
  13. Mengetweet Semula
    18 Mei

    I took 30min to find another proof of this - . 3% of weights, 95%+ accuracy. (paper by , , from )

    Our ML models are getting bigger every day, but we're just scratching the surface! 👇 is a great example of that! Facebook managed to remove 96% of the model weights, without degrading perceived audio quality.
    Tunjukkan thread ini
    1 balasan1 balasan 9 tweet semula9 tweet semula 37 suka
    Buat asal
  14. 18 Mei

    Wow this is amazing!

    Victor is releasing his new research work on extreme pruning of pretrained models! I really loved this project! A very deep dive to understand why & how standard pruning methods fail in the context of Transfer Learning and how we can do a lot better! Check his detailed thread👇
    1 suka
    Buat asal
  15. 17 Mei

    Another merged PR to Transformers for their pipelines module 😀 This time, it's for an "entity grouping" feature for the Named Entity Recognition (NER) Pipeline. You can now extract entities that consist of multiple words & word pieces 🙂

    5 balasan5 balasan 23 tweet semula23 tweet semula 125 suka
    Buat asal
  16. Mengetweet Semula
    15 Mei

    CALLING ALL SENIOR DEVELOPERS! Join us for an exclusive webinar with our best Big Data Engineers, Solution Architects, and Software Engineers. Fill up the form to be shortlisted:

    5 tweet semula5 tweet semula 6 suka
    Buat asal
  17. Mengetweet Semula
    15 Mei

    Surviving every AI wave, two kernels have consistently been the beating hearts of Natural Language Processing: Datasets and Metrics Today we release "nlp", a library to easily share & load data/metrics already providing access to 99+ datasets! Try it👉

    , , dan ≺a class="request-tagging-popup" data-activity-popup-title="Tagged in this photo" 1 3 yang lain
    26 balasan26 balasan 447 tweet semula447 tweet semula 1,600 suka
    Tunjukkan thread ini
    Tunjukkan thread ini
    Buat asal
  18. 15 Mei

    Will be speaking at 's Talks webinar series next week Friday about T5 and how to fine-tune it for Sentiment Span Extraction using transformers 🙂 Please do watch if you're interested!

    Text to text transfer transformer aka T5. Talk by Lorenzo, next week. LIVE on 22nd May, 5PM CET Add to your calendar: YouTube: He used and its for an ongoing competiton!!!
    4 tweet semula4 tweet semula 19 suka
    Buat asal
  19. Mengetweet Semula
    13 Mei

    "Spatial goes free, aiming to become the Zoom of virtual collaboration!" Thank you & !!! 🙌

    Spatial goes free, aiming to become the Zoom of virtual collaboration
    5 balasan5 balasan 34 tweet semula34 tweet semula 133 suka
    Buat asal
  20. 14 Mei

    Anyone else having intermittent disconnection issues w/ Skype from a Macbook Pro?

    Buat asal
  21. 13 Mei

    I wish there was a way to put detailed, noob friendly comments on source code without having to flood the code with comments. Currently, I just make a fork + branch that adds detailed comments. Could "official" commentaries at a line / code block level be a feature on ?

    Buat asal

@AND__SO belum lagi Tweet.

Pemuatan nampaknya mengambil sedikit masa.

Twitter mungkin melebihi kapasiti atau mengalami gangguan seketika. Cuba sekali lagi atau lawati Status Twitter untuk mendapatkan maklumat lanjut.

    Anda juga mungkin menyukai

    ·