Yingcan Wei

Thanks for your feedback! Support for dlpack will be included in a future release. But the limitations of the python interface need to be clarified. - The python interface provided...

@nv-dlasalle Thanks for your detailed background info. If my understanding is correct, then I think you will basically not use complex HPS deployment scenarios (just 1 python process per GPU...

@jperez999 There are 9 errors in your configuration files and test scripts as follows: - Configuration errors in ps.json file: 1. The value of `"model": "model"` should be "huegctr" in...

@jperez999 > HCTR][14:04:48.616][INFO][RK0][main]: Table: hps_et.0_hugectr.sparse_embedding1; **cached 1 / 1 embeddings in volatile database (PreallocatedHashMapBackend); load: 1 / 18446744073709551615 (0.00%).** > [HCTR][14:04:48.616][DEBUG][RK0][main]: Real-time subscribers created! > [HCTR][14:04:48.616][INFO][RK0][main]: Create embedding cache in...

> from the above log we can see each gpu cache store the independently embedding vector, not see any cross gpu embedding look up. When I set "gpucacheper":"0.5" in ps.json,...

> Thank you so much! So the hugectr-backend-framework does not support model parallelism like the hugeCTR training framework? If utilizing Redis cluster deployments for Distributed Database, the hugectr-backend-framework supports model...

Although **embedding cache sharing across GPUs** can expand the capacity of GPU cache, it also has many disadvantages: - Inference services deployed in the production environment must be highly available,...

@albert17 - `simple_inference_config.json` fils exist in hugectr repo https://github.com/NVIDIA-Merlin/HugeCTR/blob/master/test/utest/simple_inference_config.json. So you don't need to create this file. just make sure your `/workdir/` is hugectr repo root folder - Make sure...

The DLpack interface has supported in version 22.07

@sezhiyanhari We usually divide the structure of the recommendation model into embedding parts and dense parts, so in the multiple GPU deployment setup, each gpu will have its own independent...