Try to support float16 for flat.cc

[jjyaoao]

related to #877

[sre-ci-robot]

[APPROVALNOTIFIER] This PR is NOT APPROVED

This pull-request has been approved by: jjyaoao
To complete the pull request process, please assign presburger after the PR has been reviewed.
You can assign the PR to them by writing /assign @presburger in a comment when ready.

The full list of commands accepted by this bot can be found here.

Needs approval from an approver in each of these files:

• OWNERS

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[sre-ci-robot]

Welcome @jjyaoao! It looks like this is your first PR to milvus-io/knowhere 🎉

[mergify]

@jjyaoao Please associate the related issue to the body of your Pull Request. (eg. “issue: #”)

[liliu-z]

Hi @jjyaoao Thanks for contributing. We only allow one commit for one PR, can you help squeeze you commit and pass the test first?

[jjyaoao]

Hi @jjyaoao Thanks for contributing. We only allow one commit for one PR, can you help squeeze you commit and pass the test first?

Ok thank you, I will sort out the commit as 1 after passing the test locally. I would like to ask whether the idea of converting the incoming float16 to float32 is correct?

[liliu-z]

Hi @jjyaoao Thanks for contributing. We only allow one commit for one PR, can you help squeeze you commit and pass the test first?

Ok thank you, I will sort out the commit as 1 after passing the test locally. I would like to ask whether the idea of converting the incoming float16 to float32 is correct?

For now Knowhere's input vector is a void* and by default will be transfer to a float32. Looks like this PR is trying to support using fp16 as Knowhere's input type, and can I ask if you plan to use it with Milvus? If so this will never work since Milvus doesn't support fp16 yet.

[jjyaoao]

Hi @jjyaoao Thanks for contributing. We only allow one commit for one PR, can you help squeeze you commit and pass the test first?

Ok thank you, I will sort out the commit as 1 after passing the test locally. I would like to ask whether the idea of converting the incoming float16 to float32 is correct?

For now Knowhere's input vector is a void* and by default will be transfer to a float32. Looks like this PR is trying to support using fp16 as Knowhere's input type, and can I ask if you plan to use it with Milvus? If so this will never work since Milvus doesn't support fp16 yet.

Thank you for your explanation. I want to undertake OSPP's Milvus support FP16 type vector competition questions, so I am doing some experiments now.

[liliu-z]

Aha, plz let me know if I can do any help. Milvus support FP16 is an ambiguous topic:

1. It can mean support using FP16 as input type of Milvus
2. Or it can mean do calculation in FP16 in Knowhere, and still FP32 as input of Milvus.

For the first one, I have to say, it is a little bit complicated since we need to define how to accept FP16 as input from end to end. (Pymilvus->Milvus->Knowhere)
For the second one, we need to modify the 3rdparty lib to support FP16.

[jjyaoao]

Aha, plz let me know if I can do any help. Milvus support FP16 is an ambiguous topic:

1. It can mean support using FP16 as input type of Milvus
2. Or it can mean do calculation in FP16 in Knowhere, and still FP32 as input of Milvus.

For the first one, I have to say, it is a little bit complicated since we need to define how to accept FP16 as input from end to end. (Pymilvus->Milvus->Knowhere) For the second one, we need to modify the 3rdparty lib to support FP16.

Thank you, I think it should be the second meaning (because the difficulty of this question is the basis), if I want to modify the 3rdparty lib, what should I do? The instructor Jiao of this question told me that I should investigate Knowhere Index IVF HNSW, etc., choose a simple one to try to support Float16