Skip to content
This repository was archived by the owner on Sep 9, 2025. It is now read-only.

Conversation

@oindrillac
Copy link
Contributor

No description provided.

@oindrillac oindrillac requested a review from russellb June 17, 2024 21:35
Copy link
Member

@russellb russellb left a comment

Choose a reason for hiding this comment

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

Some tips for the linter issues

  • make md-lint-fix will fix most of the md-lint issues. I think there is one type of issue to fix manually after that.
  • and then for make spellcheck you have to add entries to .spellcheck-en-custom.txt

Comment on lines 61 to 64
# model parameters for generation
model_name: mistralai/Mixtral-8x7B-Instruct-v0.1
Copy link
Member

Choose a reason for hiding this comment

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

I think there's a need to configure the model name on a per-stage basis. We may have an endpoint configured with more than one model and we want to hit different ones depending on the stage.

Copy link
Member

Choose a reason for hiding this comment

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

This was marked as resolved, but I don't think it got moved.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

@russellb made updates to this to accept a client, model-name and default model prompt

Copy link
Member

Choose a reason for hiding this comment

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

@oindrillac but it's still not specified on a per-stage basis?

Copy link
Contributor Author

Choose a reason for hiding this comment

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

the value will be a class variable that gets passed on when an object is initialized. we have also decided to keep the run config in a .py file as per your suggestion let me update the doc

Copy link
Contributor Author

Choose a reason for hiding this comment

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

PTAL @russellb

Copy link
Contributor Author

Choose a reason for hiding this comment

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

@oindrillac oindrillac force-pushed the sdg-api branch 4 times, most recently from 099d0b4 to abc3b2b Compare June 18, 2024 15:54
@oindrillac oindrillac requested a review from russellb June 18, 2024 16:32
Copy link
Member

@russellb russellb left a comment

Choose a reason for hiding this comment

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

Thanks for the updates! There are just two minor comments left to resolve.

  • moving model_name to be per-step
  • noting that model_template and stop_token can be optional and defaulted to the right thing for the subset of models ilab supports right now.
  • I suggested an edit to describe two stages of CLI integration:
    1. changing generate_data() without impact to the CLI code
    2. adding new options to the CLI to allow custom pipeline extensions

Comment on lines 61 to 64
# model parameters for generation
model_name: mistralai/Mixtral-8x7B-Instruct-v0.1
Copy link
Member

Choose a reason for hiding this comment

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

This was marked as resolved, but I don't think it got moved.

@russellb
Copy link
Member

just checking back in on this -- there's still at least one unresolved comment.

You should also squash all the commits back down to one. That will resolve the DCO error in CI. It looks like something else needs to get added to the spellcheck dictionary, as well.

@oindrillac
Copy link
Contributor Author

its breaking on customizations @russellb how can I update the spellcheck library?

@aakankshaduggal
Copy link
Member

its breaking on customizations @russellb how can I update the spellcheck library?
You might want to update here --
https://github.com/instructlab/dev-docs/blob/main/.spellcheck-en-custom.txt

Signed-off-by: Oindrilla Chatterjee <oc@bu.edu>
Co-authored-by: Aakanksha Duggal <aduggal@redhat.com>
@mergify mergify bot added the backend InstructLab Backend Services label Jun 24, 2024
@oindrillac oindrillac force-pushed the sdg-api branch 2 times, most recently from 79fe0fd to 5517da7 Compare June 24, 2024 16:37
@oindrillac
Copy link
Contributor Author

#103 <- so that Spell Check passes on this PR

@oindrillac oindrillac requested review from leseb and russellb June 24, 2024 16:39
@russellb
Copy link
Member

#103 <- so that Spell Check passes on this PR

for this PR to pass CI, that change needs to be included in this PR instead of a separate one.

Co-authored-by: Sébastien Han <seb@redhat.com>
Co-authored-by: Russell Bryant <rbryant@redhat.com>
Signed-off-by: Oindrilla Chatterjee <oc@bu.edu>
@oindrillac
Copy link
Contributor Author

done, added this here

Copy link
Member

@russellb russellb left a comment

Choose a reason for hiding this comment

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

please see the open comment thread about the config format and model name

@oindrillac oindrillac requested a review from russellb June 24, 2024 19:02
Signed-off-by: Oindrilla Chatterjee <oc@bu.edu>
Copy link
Member

@russellb russellb left a comment

Choose a reason for hiding this comment

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

thank you!

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.

Labels

backend InstructLab Backend Services

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants