HackerTrans
TopNewTrendsCommentsPastAskShowJobs

jumpCastle

no profile record

comments

jumpCastle
·l’année dernière·discuss
It was a good benchmark until it entered the training set.
jumpCastle
·l’année dernière·discuss
But open source like aider
jumpCastle
·l’année dernière·discuss
Deepseek v3 was FP8
jumpCastle
·l’année dernière·discuss
It's quite similar to muP

https://github.com/microsoft/mup
jumpCastle
·il y a 2 ans·discuss
Stock prices cannot go up without if the planet is destroyed
jumpCastle
·il y a 2 ans·discuss
Attention was invented because Bengio lab had to be disciplined about a black box (google had more compute)
jumpCastle
·il y a 2 ans·discuss
Aren't services like runpod solve half of these concerns?
jumpCastle
·il y a 2 ans·discuss
Are the updates really slow?