Yi Zhu

Semantic segmentation training on Cityscapes dataset

1

Training semantic segmentation models on Cityscapes is fine using MXNet 1.4.1. Using version > 1.4.1 will cause training hang. Needs investigation.

*Issue #, if available:* NA *Description of changes:* This PR aims to enable object detection in MultimodalPredictor by integrating MMdetection. Will add data handling and model training in following PRs....

Difference between realtime res18 and non-realtime res18 model

1

Hi @ycszen , thank you for providing this great repo. I have a question about the difference between realtime res18 and non-realtime res18 model. As shown in the readme, realtime...

Does langchain support using Contriever as an embedding method?

4

It is great to see langchain already support [HyDE](https://langchain.readthedocs.io/en/latest/modules/utils/combine_docs_examples/hyde.html). But in its original paper, once the hypothetical documents are generated, the embedding is computed using [Contriever](https://github.com/facebookresearch/contriever) model as described in...

Too large minhashLSH index

10

Hi, I have a question about large-scale LSH index. If I have billions of documents, I suppose even 1T RAM is not enough to do in-memory LSH, is there any...

Liencese of this repo

1

Hi, thank you for the great repo. Could you add a license to it, like MIT or Apache 2.0? Thank you very much.

Hi @Breakend , can I ask has there any deduplication been done for Pile of Law? Either local deduplication from a single source, or global deduplication among all the data...

Hi guys, I was wondering if I want to download videos from Vine for some region, for example, New York city, how can I start? Is there a location search...