SONG Ge

## Description ### 1. Why the change? * Current `save` and `load` do not support `h5` and `savemodel` format, since they mainly retrieve model states. * Users need the API...

## Description When users running Orca pyspark backend programs on Yarn, they must specify the `ip` and `port` of hdfs when saving model states, which could cause errors and difficult...

[Orca] Code Design for Distributed PyTorch DataLoader on Ray

2

## Description This issue mainly describes a new distributed data input, which could be used the same as the standard Pytorch Dataloader. ## Motivation * Pure Ray pipeline needs a...

[Orca] Tutorials for Submit and Run Programs on Yarn

5

Add tutorials for submitting and running programs on yarn. - [x] rename `spark-submit-with-bigdl` to `bigdl-submit` - [x] add a multiple python files demo - [x] built-in - [x] bigdl-submit -...

[Orca] TF2 Ray Estimator Supports Callback to Save Model

3

Add ModelCheckpoint callbacks with training using `estimator.fit( )` to save a model in a checkpoint file. - [x] Save model and weights only. - [x] Save and load model weights.

[Orca Draft] Tutorial for Submit and Run Programs on K8s

1

## Description This PR will release a tutorial for running Orca PyTorch Pyspark program on K8s cluster, it provides a detailed step-by-step guide and a simple FashionMNIST example that allows...

## Description Apply resource_quota to limit the total amount of compute resources that may be consumed by resources in that namespace. ### 1. Why the change? The current deployment scheme...

[LLM] Add quantize_kv optimization for yuan2 model

1

## Description Add quantize_kv to optimize the performance of yuan2 model on mtl. ### 1. Why the change? ### 2. User API changes ### 3. Summary of the change ###...

This PR is going to support ray runtime in Chronos examples which call `init_orca_context` and with `data_creator` input. - [x] Update network_traffic.py - [x] Update base-forecaster.py - [x] Add Test...

[LLM] Add attention_sinks for CodeShell example to optimize multi-turn chat

1

## Description This PR applies `attention_sinks` to optimize multi-turn chat, enabling multi-turn chat without requiring inputs history. ### 1. Why the change? The previous multi-turn chat required input history, which...