Conversation
✱ Stainless preview buildsThis PR will update the Edit this comment to update it. It will appear in the SDK's changelogs. ✅ togetherai-openapi studio · code · diff
✅ togetherai-python studio · code · diff
✅ togetherai-go studio · code · diff
✅ togetherai-typescript studio · code · diff
✅ togetherai-terraform studio · code · diff
This comment is auto-generated by GitHub Actions and is automatically kept up to date as you push. |
openapi.yaml
Outdated
| description: Weight decay. Regularization parameter for the optimizer. | ||
| random_seed: | ||
| type: integer | ||
| description: Random seed for reproducible training. When set, the same seed produces the same run (e.g. data shuffle, init). If not provided, a default seed `42` is used. |
There was a problem hiding this comment.
- should it contain
default: nullas well (since optional)? - Do we want to explicitly mention
42here? - Do we want to leave an option for non-deterministic (as it was before), e.g. when "-1" is passed (but it'll require some tweaks on the backend)? Imho not, a user can do
random_seed=randint()themselves if they really want it, but it's not a strong opinion
There was a problem hiding this comment.
- yes will update.
- Maybe we should? If user wants to run a different random value besides of the default one they need to know it's 42.
- I think user can do
random_seed=randint()if they want to randomize the experiments.
openapi.yaml
Outdated
| default: 0.0 | ||
| description: Weight decay. Regularization parameter for the optimizer. | ||
| random_seed: | ||
| type: integer |
There was a problem hiding this comment.
Hm, we still have string type in the backend, so iiuc it won't work if a client send int value
There was a problem hiding this comment.
You are right after I checked again the backend does expect string type. Not related to this PR but do we consider to update backend to support int type instead for random seed param?
| description: Weight decay value used | ||
| random_seed: | ||
| type: integer | ||
| nullable: true |
There was a problem hiding this comment.
are you sure it's nullable? I thought we always explicitly set some random_seed from now on
There was a problem hiding this comment.
After double check, the API model allows empty value for this. It's the backend who sets the default.
| type: number | ||
| format: float | ||
| description: Weight decay value used | ||
| random_seed: |
There was a problem hiding this comment.
btw it's not shown in the UI, probably we need to create a dependency for FE. But not 100% sure we want to show it there
There was a problem hiding this comment.
Sorry what does FE stand for?
There was a problem hiding this comment.
I think it's nice to have if you meant frontend.
openapi.yaml
Outdated
| random_seed: | ||
| type: integer | ||
| nullable: true | ||
| description: Random seed used for reproducible training, if set |
There was a problem hiding this comment.
, if set - again, we always set it a specific value now, even if not passed explicitly?
There was a problem hiding this comment.
Yes, I thought the nullable here is for the client who called this API. They don't need to set any value. It will be default to 42 behind though.
openapi.yaml
Outdated
| format: float | ||
| description: Weight decay value used | ||
| random_seed: | ||
| type: integer |
There was a problem hiding this comment.
also a type mismatch, currently we return string, not integer
3 participants
Note
Low Risk
Low risk documentation-only change in
openapi.yamladding a new optional field to the published API schema.Overview
Documents a new
random_seedparameter inopenapi.yamlfor fine-tuning: it can be provided on job creation and is also surfaced in job details as nullable.The schema describes
random_seedas a numeric string (e.g."42") used to make training runs reproducible.Written by Cursor Bugbot for commit ce8d086. This will update automatically on new commits. Configure here.