[AIR][Train][Doc] Restructure API reference - #32360
Conversation
Signed-off-by: Justin Yu <justinvyu@berkeley.edu>
Signed-off-by: Justin Yu <justinvyu@berkeley.edu>
Signed-off-by: Justin Yu <justinvyu@berkeley.edu>
c21
left a comment
There was a problem hiding this comment.
Looks good by spot-checking several doc page. cc @matthewdeng for approval.
Signed-off-by: Justin Yu <justinvyu@berkeley.edu>
Signed-off-by: Justin Yu <justinvyu@berkeley.edu>
| .. toctree:: | ||
| :maxdepth: 2 |
There was a problem hiding this comment.
Love this structure so much!
| Dataset Ingest (Ray Data)<dataset-ingest.rst> | ||
| Trainers (Ray Train)<../../train/api/api> | ||
| Tuner (Ray Tune)<../../tune/api_docs/execution> | ||
| Results (Ray Train + Ray Tune)<../../tune/api_docs/result_grid> |
There was a problem hiding this comment.
This will be fixed by merging in Tune PR.
| Ray Train Built-in Predictors | ||
| ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ | ||
|
|
||
| .. autosummary:: | ||
| :toctree: doc/ | ||
|
|
||
| ~xgboost.XGBoostPredictor | ||
| ~lightgbm.LightGBMPredictor | ||
| ~tensorflow.TensorflowPredictor | ||
| ~torch.TorchPredictor | ||
| ~huggingface.HuggingFacePredictor | ||
| ~sklearn.SklearnPredictor | ||
| ~rl.RLPredictor | ||
|
|
||
| Ray Train Framework-specific Checkpoints | ||
| ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ | ||
|
|
||
| .. autosummary:: | ||
| :toctree: doc/ | ||
|
|
||
| ~xgboost.XGBoostCheckpoint | ||
| ~lightgbm.LightGBMCheckpoint | ||
| ~tensorflow.TensorflowCheckpoint | ||
| ~torch.TorchCheckpoint | ||
| ~huggingface.HuggingFaceCheckpoint | ||
| ~sklearn.SklearnCheckpoint | ||
| ~rl.RLCheckpoint |
There was a problem hiding this comment.
Personally feel like we should move these to the predictor/checkpoint pages even though they're in the train package...
There was a problem hiding this comment.
I think checkpoints should stay in train, since they are still needed by users of Train only (w/o prediction). I can include another reference in the air checkpoint page for framework specific ckpts, but without toctree.
Predictors I agree should be moved out.
| @@ -0,0 +1,18 @@ | |||
| .. _air-serve-integration: | |||
|
|
|||
| Model Serving in AIR | |||
There was a problem hiding this comment.
Add ray.serve.air_integrations.PredictorWrapper?
There was a problem hiding this comment.
PredictorWrapper should be on this page already.
There was a problem hiding this comment.
Yeah, to be fair this page is kinda a mess. The see also tooltip + autosummary + autofunction is a bit much. But I had to do switch predictor deployment to an autofunction because it's an empty class that was erroring if I put it in autosummary.
| AIR Checkpoint (All Libraries)<checkpoint.rst> | ||
| Predictors (Ray Data + Ray Train)<predictor.rst> | ||
| Model Serving in AIR (Ray Serve)<serve.rst> | ||
| External Library Integrations<../../tune/api_docs/integration> |
There was a problem hiding this comment.
Just double checking that we're intentionally keeping these unchanged?
There was a problem hiding this comment.
The external libraries one will be updated with the Tune PR. The other ones are updated.
β¦n/api_restructure Signed-off-by: Justin Yu <justinvyu@berkeley.edu>
β¦in air Signed-off-by: Justin Yu <justinvyu@berkeley.edu>
β¦n/api_restructure
Signed-off-by: Justin Yu <justinvyu@berkeley.edu>
β¦n/api_restructure
Signed-off-by: Justin Yu <justinvyu@berkeley.edu>
c21
left a comment
There was a problem hiding this comment.
cc @richardliaw - can you merge this?
Signed-off-by: Justin Yu <justinvyu@berkeley.edu>
Signed-off-by: Justin Yu <justinvyu@berkeley.edu>
β¦n/api_restructure
|
@richardliaw Let's wait to merge after tests pass. |
|
ping me when tests pass
β¦On Mon, Feb 13, 2023 at 12:17 PM Justin Yu ***@***.***> wrote:
@richardliaw <https://github.com/richardliaw> Let's wait to merge after
tests pass.
β
Reply to this email directly, view it on GitHub
<#32360 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/ABCRZZNLX3C66YPR277ZZEDWXKJGJANCNFSM6AAAAAAUWMM5MI>
.
You are receiving this because you were mentioned.Message ID:
***@***.***>
|
Signed-off-by: Justin Yu <justinvyu@berkeley.edu>
β¦n/api_restructure
Signed-off-by: Justin Yu <justinvyu@berkeley.edu>
Signed-off-by: Justin Yu <justinvyu@berkeley.edu>
Signed-off-by: Justin Yu <justinvyu@berkeley.edu>
β¦n/api_restructure
This PR splits up long API refs in AIR and Train into individual pages, one dedicated to each method/class. This PR is a followup to ray-project#31204 and ray-project#32311, which made the same changes for Ray Data/Tune docs. Signed-off-by: Justin Yu <justinvyu@berkeley.edu>
This PR splits up long API refs in AIR and Train into individual pages, one dedicated to each method/class. This PR is a followup to ray-project#31204 and ray-project#32311, which made the same changes for Ray Data/Tune docs. Signed-off-by: Justin Yu <justinvyu@berkeley.edu> Signed-off-by: Edward Oakes <ed.nmi.oakes@gmail.com>
This PR splits up long API refs in AIR and Train into individual pages, one dedicated to each method/class. This PR is a followup to ray-project#31204 and ray-project#32311, which made the same changes for Ray Data/Tune docs. Signed-off-by: Justin Yu <justinvyu@berkeley.edu> Signed-off-by: elliottower <elliot@elliottower.com>


Why are these changes needed?
This PR splits up long API refs in AIR and Train into individual pages, one dedicated to each method/class.
This PR is a followup to #31204 and #32311, which made the same changes for Ray Data/Tune docs.
Related issue number
Checks
git commit -s) in this PR.scripts/format.shto lint the changes in this PR.