Hugging Face's logo

Spaces:
open-llm-leaderboard
/
open_llm_leaderboard
Running on CPU Upgrade

App Files Community
1049

14B model detected as 7B

#1049
by djuna - opened
Discussion
djuna

I've been working on merging a 14 billion parameter model recently, but when it comes time to evaluate the model, the system indicates that the model has only 7 billion parameter instead of the expected 14 billion. It's funny when the top 7 billion model is actually 14 billion

alozowski
 Open LLM Leaderboard org

Hi @djuna ,

Could you please provide the request file for the model you submitted so we will be able to check the number of parameters?

djuna

@alozowski this one
https://huggingface.co/datasets/open-llm-leaderboard/requests/blob/main/djuna/Q2.5-Veltha-14B-0.5_eval_request_False_bfloat16_Original.json

djuna

When you filter 7-8B size on the spaces, more than 10+ model is actually 14B

brankor-mcom

There are quite a few models in the leaderboard where the indicated size is half the actual size:

  • maldv/Qwentile2.5-32B-Instruct
  • CultriX/Qwen2.5-14B-Wernickev3

...and many others, most of them Qwen-derived.

Sign up or log in to comment