googleapis / googleapis/python-aiplatform

BatchPredictionJob.completion_stats should be a method and not a property

Abierto
#5,067 0 comentarios 0 reacciones 0 asignados Ver en GitHub
api: vertex-ai
Lenguaje dominante
Python
Estrellas
905
Forks
465
Merge medio
1 d 13 h
PR fusionados (30 d)
44

Descripción

The below bit of code I suspect incorrectly has the `@property` decorator. This is caching, so when you've created a BatchPredictionJob instance and called submit or create. the first time you call `.completion_stats` is the value that this instance will have forever. For obvious reasons this is extremely irritating and means you either have to keep creating fresh instances of BatchPredictionJob to get updated completion numbers. Or you have to copy the private code and call yourself, which isn't super great from a user maintenance PoV. It would be extremely useful to have the total number of input instances somewhere cached based on if its BQ or a GCS folder, in a similar way that this class provides an iterator to outputs. Then you can get a %age progress through the job rather than predictions output, again one has to write all this themselves on the outside of this class which is just kind of horrible aesthetics as a user.

https://github.com/googleapis/python-aiplatform/blob/96d2ecbb527cc52c71bbeddc2394eaf9c4f91e06/google/cloud/aiplatform/jobs.py#L385-L390

I'd happily raise an MR around these changes if they're seen as a good change that might get accepted. The change from being a property will be breaking, but I suspect *anyone* using this feature as is either has broken code that incorrectly states progress, or they are doing something horrible they would like to remove to work around it being a property.

Guía de contribución

Abrir la guía de contribución

Línea de trabajo

Start at google/cloud/aiplatform/jobs.py around lines 385-390 and inspect how BatchPredictionJob.completion_stats is implemented and used. Compare the property behavior with the issue's requested method semantics and review related batch prediction progress handling; done should include a clear, non-stale way to retrieve completion statistics without breaking existing behavior unnoticed.

Escrito por el modelo de indexación a partir del texto del issue.

Evaluación

Stack tecnológico
python
Área
machine-learning
Tipo de issue
Error
Dificultad
4/5
Tiempo estimado
3-5 días
Estado de actividad
Estancado
Claridad
Bastante claro
Aptitud para principiantes
35/100

Recibe los nuevos issues en tu correo

Un resumen breve de issues de GitHub para principiantes.