Skip to content

Conversation

seanshi-scale
Copy link
Contributor

@seanshi-scale seanshi-scale commented Oct 12, 2023

Also update docs to reflect that mistral-7b is available

@seanshi-scale seanshi-scale marked this pull request as ready for review October 13, 2023 00:45
@seanshi-scale seanshi-scale self-assigned this Oct 13, 2023
@seanshi-scale seanshi-scale requested a review from yixu34 October 13, 2023 00:45
@seanshi-scale seanshi-scale changed the title Clarify get-events API usage [Docs] Clarify get-events API usage Oct 13, 2023
@seanshi-scale seanshi-scale enabled auto-merge (squash) October 13, 2023 17:06
@yixu34
Copy link
Member

yixu34 commented Oct 13, 2023

Don't merge yet, was going to make some comments

@seanshi-scale seanshi-scale requested a review from yixu34 October 18, 2023 01:19
as well as any errors that may have occurred. If you encounter any errors with your fine-tune,
the events are a good place to start debugging. For example, if you see `Unable to read training or validation dataset`,
you may need to make your files accessible to LLM Engine. If you see `Invalid value received for lora parameter 'lora_alpha'!`,
you should check that your hyperparameters are valid.
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Might be good to add a reminder link to the canonical definition of valid hyperparameters?

The status of your fine-tune will give a high-level overview of the fine-tune's progress.
The events of your fine-tune will give more detail, such as the training loss and validation loss at each epoch,
as well as any errors that may have occurred. If you encounter any errors with your fine-tune,
the events are a good place to start debugging. For example, if you see `Unable to read training or validation dataset`,
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Do we have a separate error for malformed files, e.g. files that don't have prompt,response headers?

Would it make sense to put these in a table?

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Discussed offline: prompt,response header validation would happen earlier, at FineTune.create time.

@seanshi-scale seanshi-scale requested a review from yixu34 October 18, 2023 01:41
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants