fan yang

/assign

Hi @xiaofan-luan , I'm looking into the previous [PR](https://github.com/milvus-io/milvus/pull/16991/files). It's quite a huge PR. Since I'm still getting familiar with Milvus code, I'd like to know if it's better for...

> 1. is pebble fully compatible with rocksdb? which is saying we can upgrade without change any data format? > > 2. is there a possibility to make pebble store...

Hi @xiaofan-luan , regarding the questions: 1. There are some features not supported by Pebble, according to the [readme](https://github.com/cockroachdb/pebble/blob/master/README.md). But all the features we are using are included, like simply...

1. Typically, most features can be shared between retrieval and ranking models. But in most cases, the ranking model will have more complex features because of the limited candidate numbers...

>contains approximately 100k to 200k data points, Is this a total dataset or just samples from one day? If it's a daily number, a simple approach is just expand the...

> Do you think this is too small for the two-tower model even if it's not very deep? Aha, I didn't have experience with such a small dataset. We can...

> Is the two-tower model a suitable approach for item-to-item recommendations, like recommending similar items based on a queried item? Yes. It is. A good practice is to leverage NLP...

> Do you think cosine similarity will still be effective if we concatenate text embeddings and encoded categorical features? Yes, the cosine similarity is what the two-tower model learns. Concatenation...

Thanks for providing this fix. I also noticed this issue. 👍