Simon_CQK comments

Results 29 comments of


                                            Simon_CQK

pytorch job seems to succeed but data is not completed

hi @WencongXiao , the issue you've mentioned will be fixed soon, we'll introduce `successPolicy` to `PyTorchJob` to address this, hence `PyTorchJob` will be marked as Succeed when all workers exit...

[feature request] enrich the serving case in the example

@ZinuoCai hi nuocai, here we provide a simple tutorial of training and deploying a tiny model by kubedl, and example manifests are under `/example` : https://kubedl.io/docs/tutorial/video/ as for detailed documentation,...

feat: enable pytorch elastic training fashion based on torch elastic

@wanziyu hi wanziyu, thanks for your contribution! before merge your PR to master branch, you should sign-off your commit first.

[feature request] refactor CacheBackend to Dataset

@zjchenn

Set user customized wait-timeout-seconds for PodGroup based gang scheduling protocol.

@Thor-wl hi, thanks for replying, it seems that `SLA` plugin implements the semantics I present above, what make me curious is it coupled with `Batch Volcano Job` api ? What...

Feature Request: remove items from the B+ Tree

Well done dude, I'm working on my personal database engined by B Tree (B+ Tree firstly, but edge cases drived me crazy...), I gonna adapt some of your designes !!

[feature request] job level scheduling and orchestration

@jian-he I have updated the description

[BUG] JindoMaster mount point did not recovered from pod migration.

> Maybe we can refer to #1322 for similar solution. Yes it is, and it seems that master still corrupted when recreated or restarted with JindoRuntime, anyway, I'll open an...

[feature request] Delete the completed tfjob created by cron when tfjob number reach historyLimit

> well, it has not implemented yet, it looks good if you'd like to contribute!

feat: re-setup master when master role restarted or recreated and recover dataset

> @SimonCqk please fix the source code check, thanks. done :)