Cengguang Zhang comments

Results 7 comments of


                                            Cengguang Zhang

Orca: Add 2 NCF PyTorch examples with data_loader or XShards as inputs.

Can we merge three train_*.py files?

[WIP]Orca: Add ncf_orca.py with spark dataframe

Do we need a README.md for each tutorial scripts?

Orca: add compile_args_creator in distributed training

YARN validation(cluster mode): ray backend: http://172.16.0.105:8088/cluster/app/application_1646614241228_2178 spark backend: http://172.16.0.105:8088/cluster/app/application_1646614241228_2181 YARN validation(client mode): ray backend: http://172.16.0.105:8088/cluster/app/application_1646614241228_2183 spark backend: http://172.16.0.105:8088/cluster/app/application_1646614241228_2182

Orca: add HDFS default port

Yes, you can just specify the host IP or specify both host IP and port.

RayOnSpark on K8s can not enable dashboard

The same problems also occur when running in yarn cluster, which may need further verification

Intel MPI support

When will it support TensorFlow 2.x? Could you please give me an approximate schedule?

2nd latency of llama3-8B-Instruct with int4 & all-in-one tool issue

Already reproduce the issue, and will fix it later. We recommend you use fp16 for non-linear layer, please refer to benchmark scripts [all-in-one](https://github.com/intel-analytics/ipex-llm/tree/main/python/llm/dev/benchmark/all-in-one), and select `transformer_int4_fp16_gpu` API.