Malte Ostendorff
@XYOU
Followers
767
Following
2K
Statuses
618
@occiglot @openlegaldata @deutschetelekom
Berlin, Germany
Joined June 2009
10/ Even basic things like language identification work poorly for those languages, which decreases the amount of data already at the stage of Web crawls.
1
0
0
The very first edition of the Conference on Language Modeling is kicking off 🚀#COLM2024 @COLM_conf
1
1
16
@BSC_CNS just released a series of European LLMs: 2b, 7b and soon 40b 🚀
We are launching Salamandra 2B & 7B multilingual LLMs trained at @BSC_CNS from scratch with nearly 8 trillion tokens in 35 EU languages+code. Spanish languages have been carefully curated, with Romance languages comprising >30% of the training dataset. 👇
0
0
2
@MalteLandwehr @HamelHusain The assumption is that you can simply put all new data into the prompt as soon as the context length becomes really big.
0
0
1
There are even more Axolotl configs being shared on @huggingface
One of the most underrated things about @winglian's Axolotl is the database of config yamls that exist showcasing how to train so, so many models. I hope some day soon we can emulate a tenth of this into Accelerate to help people understand DDP configs more
0
0
1
All three videos look so similar. Do they all overfit to the same training data?
We are LOVING the generations from @LumaLabsAI Dream Machine. Now it’s time to answer the question, how does it compare to these other big name video generators. For most of these shots, Luma wins. #lumadreammachine #aifilmmaking #aivideo #filmmaking
0
0
1