Hacker News new | ask | show | jobs
by Squeeze2664 323 days ago
How do you determine the importance of a layer in this case?
2 comments
smallerize 322 days ago
https://unsloth.ai/blog/dynamic-v2
link
danielhanchen 322 days ago
Yes also https://unsloth.ai/blog/deepseekr1-dynamic, https://unsloth.ai/blog/dynamic-4bit, https://docs.unsloth.ai/basics/unsloth-dynamic-2.0-ggufs
link
kkzz99 322 days ago
Afaik they have a test bench that they use and take the activation data from that.
link
danielhanchen 322 days ago
Yes we have around 1 to 3 million tokens of high quality self verified data that we use to calibrate models!
link