ucalyptus
@ucalyd
Followers
1K
Following
9K
Media
286
Statuses
2K
Throwback to trying out WebGPU for @premai_io 's SLM
On-device deployment of LLMs is more important than ever. Today we’re releasing SmolLM a new SOTA series of 135M, 360M and 1.7B models:. - Outperforming MobileLLM, Phi1.5 and Qwen2 small models.- Trained on SmolLM-corpus, of high quality web, code and synthetic data
1
1
7
last commit 4 months ago.
github.com
🤗 AutoTrain Advanced. Contribute to huggingface/autotrain-advanced development by creating an account on GitHub.
0
0
0
lets go.
The new GRPO stabilisation trick from @PrimeIntellect has just landed in TRL!. To use it, install from main with. pip install git+ and then set `delta > 1 + epsilon` in GRPOConfig :)
1
0
1
now merged in TRL.
Two-Sided GRPO Clipping: Stabilizes training by mitigating gradient spikes with two-sided token probability ratio clipping.
1
0
6
maintenant ce compte x est principalement pour pratiquer mon français.
0
0
1
tout le monde prétend qu'ils sont "cracked", mais seuls quelques-uns le sont.
0
0
1