There is a total of six submissions, three for the track 4090, three for the track A100. As we are currently not in the top 10 in the leaderboard anymore and one of us is in vacation, we do not consider a top 2 place likely, so we just provide the zip files to eval the models without detailed instructions. Should we be in top 2 or all others fail, we will provide instructions that explain the scripts in the folder ...detailedinstructions.
You find the following folders:
Probably it is easiest just to use the folder zip files with the submissions as provided to evalbot.
Track 4090:
- submission_4090_1_fine_tune which refers to the following submission in track 4090: (12.) attkap (Oct 11, 07:37:31AM)
- submission_4090_2_base which refers to the following submission in track 4090: (20.) hannesvienna (Oct 02, 06:59:13PM)
- submission_4090_3_fine_tune_2
Track A100:
- submission_A100_4_70b (!!! this needs around two hours of downloading time during runtime !!!)
- submission_A100_5_base
- submission_A100_6_fine_tune
Hi, Setup and onboarding with toy submission was great, thank you for that. We will try via the following approaches:
- https://www.youtube.com/watch?v=g68qlo9Izf0
- https://blog.langchain.dev/using-langsmith-to-support-fine-tuning-of-open-source-llms/
- https://learn.deeplearning.ai/finetuning-large-language-models/lesson/2/why-finetune
and then see how far we get with that tomorrow. As Attila and I are currently not employed in any company with access to GPUs we would love the AWS credits.
Best, Hannes
PS: Don't get thrown off by our repository name, it is a reference to "Odysseus nobody". ;
HI, IF YOU READ THIS, THERE IS NOTHING HERE YET, WE ARE JUST SETTING UP OUR TEAM APPLICATION AND REPOSITORY