model resubmit needed (1 model returning as finished)

#380
by sequelbox - opened

Hi! We submitted 4 models to the leaderboard this afternoon - two from Valiant Labs, two from sequelbox.
3 of the 4 are running, but 1 (ValiantLabs/ShiningValiant) is immediately returning as 'finished' and not running.
It hasn't been run, though, it's not on the leaderboard.

Can you start this run for us? Thanks!

most recent re-submission:
https://huggingface.co/datasets/open-llm-leaderboard/requests/commit/340023e9835a855737461592a2c2bc5158e6b04b
https://huggingface.co/datasets/open-llm-leaderboard/requests/commit/3f02fdf956c7f0d0bd6ba3260ff4b6c7aca82cfc

Open LLM Leaderboard org
edited Nov 15, 2023

I checked your logs and request file - why is the "revision" commit 1.3?

It's not a valid commit hash, hence the failure (I'll check why it appears as finished though) - I'll also make it clearer

Open LLM Leaderboard org

Thank you for providing the links to the request file!

@clefourrier I appreciate the new wording on the submit form, thank you :)

sequelbox changed discussion status to closed

@clefourrier (CC: @SaylorTwift )

hmm, the 'FINISHED' problem is still happening for Shining Valiant.

tried submitting with blank revision field as well as the exact commit.

most recent:

https://huggingface.co/datasets/open-llm-leaderboard/requests/commit/216245b04a0b542e8ed0a695b01b63d4c249abd6
https://huggingface.co/datasets/open-llm-leaderboard/requests/commit/e286862b8e96fa54e67cad075fe75b6c323bdcd3

sequelbox changed discussion status to open
Open LLM Leaderboard org
edited Nov 20, 2023

Your model is running fine (about 10h remaining if it doesn't crash or get cancelled :) )

Interestingly, it returned as FINISHED when it got cancelled (and before it was re-ran) not FAILED because you already submitted a model of the same name 2 months ago - and it is checking the results file against names. We'll fix this.

thanks for clarifying what was happening!

we're still waiting for results; assuming there's been some crashes? :) no worries, we know there's a lot of people excited about their models.
I'll keep an eye on the results repo and watch out there.

thanks again!

sequelbox changed discussion status to closed

Sign up or log in to comment