GSoC 2024: Summary of LLM Hyperparameter Optimization API Project #154
Conversation
Signed-off-by: helenxie-bit <helenxiehz@gmail.com>
Signed-off-by: helenxie-bit <helenxiehz@gmail.com>
|
Ref: kubeflow/katib#2339 |
|
Please review when you have time and any suggestions are welcome! Thanks! @andreyvelich @johnugeorge @terrytangyuan |
Signed-off-by: helenxie-bit <helenxiehz@gmail.com>
There was a problem hiding this comment.
Thank you for working on this @helenxie-bit, and sorry for the late reply!
/assign @varodrig @hbelmiro @franciscojavierarceo @kubeflow/wg-training-leads @Electronic-Waste
Please can you help us with the review, so we can merge this great blog post ?
|
This is awesome! We'll make sure to review these sooner going forward :) |
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: franciscojavierarceo The full list of commands accepted by this bot can be found here. The pull request process is described here
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
|
|
||
| Hyperparameter optimization is a crucial but time-consuming task in fine-tuning machine learning models, especially for LLMs that involve billions of parameters. This API aims to streamline this process by abstracting the complexity of Kubernetes infrastructure, enabling data scientists to focus on model performance instead of system configuration. | ||
|
|
||
|  |
There was a problem hiding this comment.
This image doesn't work for me on preview:
https://deploy-preview-154--infallible-murdock-d05108.netlify.app/gsoc-2024-project-4/
|
|
||
| Hyperparameter optimization is a crucial but time-consuming task in fine-tuning machine learning models, especially for LLMs that involve billions of parameters. This API aims to streamline this process by abstracting the complexity of Kubernetes infrastructure, enabling data scientists to focus on model performance instead of system configuration. | ||
|
|
||
|  |
There was a problem hiding this comment.
It would be nice to share a little bit about the feature and why it is useful for Kubeflow Katib end-users.
Maybe we can take something from your proposal or documentation PR: kubeflow/website#3952
| Hyperparameter optimization is a crucial but time-consuming task in fine-tuning machine learning models, especially for LLMs that involve billions of parameters. This API aims to streamline this process by abstracting the complexity of Kubernetes infrastructure, enabling data scientists to focus on model performance instead of system configuration. | ||
|
|
||
|  | ||
|
|
There was a problem hiding this comment.
Let's also cross-reference the docs for this feature, since we will merge this PR soon: kubeflow/website#3952
This PR adds a detailed summary of my GSoC 2024 Project 4: Developing the LLM Hyperparameter Optimization API in Kubeflow's Katib. It highlights the motivation, goals, my contributions, and key lessons learned from the project.