Fine-tune LLM on Hetzner - Githubissues

amosproj / amos2024ss08-cloud-native-llm

MIT License

6 stars 1 forks source link

Fine-tune LLM on Hetzner #99

Open grayJiaaoLi opened 1 week ago

grayJiaaoLi commented 1 week ago

User story

As a ML engineer
I want / need to fine-tune our LLM on Hetzner
So that our initially trained LLM can be improved

Acceptance criteria

A training of the model on Hetzner is performed
Follow the same procedure of we used to do on HPC
Document the results and feedback on satisfaction with resources
- Is the current resources enough?
- Do we have barriers to accessing Hetzner?

Definition of done (DoD)

Bill of Materials in the planning document has been updated
All feature branches have been merged and closed
New feature code has been documented
Potential new licenses have been checked
All GitHub Actions are passing
The requirement.txt is updated

DoD general criteria

Feature has been fully implemented
Feature has been merged into the mainline
All acceptance criteria were met
Product owner approved features
All tests are passing
Developers agreed to release

grayJiaaoLi commented 3 days ago

Considering still no resources on Hetzner, we have get to try it on HPC again for now.

dominic0df commented 3 days ago

The credentials were now provided by our industry partner (see discord chat)

dominic0df commented 2 days ago

already invested in HPC SP02, SP05 is still remaining as we had an issue with the credentials