1719 152 65

Stefan Schweter PRO

stefan-it

AI & ML interests

Flair Library 💕, NER & PoS Tagging, LM Pretraining (mostly encoder-only & encoder-decoder), Historical Language Models

Recent Activity

new activity 3 days ago

stefan-it/neobert-ner-conll03:Update model.py

new activity 5 days ago

dbmdz/electra-base-german-europeana-cased-discriminator:Adding `safetensors` variant of this model

reacted to csabakecskemeti's post with 🔥 6 days ago

-UPDATED- 4bit inference is working! The blogpost is updated with code snippet and requirements.txt https://devquasar.com/uncategorized/all-about-amd-and-rocm/ -UPDATED- I've played around with an MI100 and ROCm and collected my experience in a blogpost: https://devquasar.com/uncategorized/all-about-amd-and-rocm/ Unfortunately I've could not make inference or training work with model loaded in 8bit or use BnB, but did everything else and documented my findings.

View all activity

Organizations

stefan-it's activity

New activity in stefan-it/neobert-ner-conll03 3 days ago

Update model.py

#1 opened 6 days ago by

KoichiYasuoka

New activity in dbmdz/electra-base-german-europeana-cased-discriminator 5 days ago

Adding `safetensors` variant of this model

#1 opened 5 days ago by

SFconvertbot

reacted to csabakecskemeti's post with 🔥 6 days ago

Post

1906

-UPDATED-
4bit inference is working! The blogpost is updated with code snippet and requirements.txt
https://devquasar.com/uncategorized/all-about-amd-and-rocm/
-UPDATED-
I've played around with an MI100 and ROCm and collected my experience in a blogpost:
https://devquasar.com/uncategorized/all-about-amd-and-rocm/
Unfortunately I've could not make inference or training work with model loaded in 8bit or use BnB, but did everything else and documented my findings.

4 replies

replied to their post 6 days ago

Let's see if BERT5urk can make it into @merve 's weekly recap of open AI 🤗

posted an update 6 days ago

Post

768

🇹🇷 😍 I'm very happy to finally announce my new Turkish LM called "BERT5urk":

stefan-it/bert5urk

It is a 1.42B T5-based model, trained with UL2 pretraining objective on the Turkish part of the awesome HuggingFaceFW/fineweb-2 dataset.

Feel free to check it out!