[BUG]V3.2:Dense matrix type can not use hash partitioner

[ghost]

---

name: Bug report/Feature request/Question about: Create a report to help us improve title: '' label: bug/enhancement/question assignees: ''

Environment:

• Java version:
• Scala version:
• Spark version:
• PyTorch and Python version:
• OS and version:

Checklist:

• Did you check if your bug/feature/question is answered in the FQA ?
• Did you search issues to find if somebody discuss your bug/feature/question before?
• If your bug/question is about install, did you read this doc?
• If your bug/question is about parameter setting, did you read this doc?

Your Bug/Feature request/Question: Please describe bug/enhancement/question in detail.

• For bugs, please post the submit commands, error report logs and related code snippet
• For feature requests, please describe what's you scenario and why you need this feature. if you required feature is big please connect us email list.

[LuiCB]

Hi,

I also notice this issue.

I tried to run the Spark-on-Angel examples like LR and PSVector and when creating a vector or a matrix, there is always an exception

com.tencent.angel.exception.AngelException: Dense matrix type can not use hash partitioner

I notice that in the MatrixContext.java , the partitionerClass variable by default is a HashPartitioner (check MatrixContext.java#L107). Also, in the check() function (MatrixContext.java#L723), the dense matrix by default will throw an exception if a dense vector/matrix needs to be created with the default partitioner.

I use version 3.2.0.

thanks!

[jinqinn]

i also got this exception