ARK v0.2.0 Release Plan (Released)

[chhwang]

Timeline

Released Date: Sep. 5th, 2023

Work Items

Model

1. • Interface: expose the underlying buffer info to Tensor (Update TensorBuf interfaces #79)

Scheduler

1. • [ ] Graph Optimization: enable this feature moved to the next version
2. • [ ] SimpleScheduler: fix broken features moved to the next version

Communication Stack

1. • Interface: hide GpuCommSw implementation from the interface (Update GpuCommSw interfaces #81)
2. • Interface: extend the current interface (Extend the number of SIDs #104)

Operators Support

1. • Operator: add more operators (ARK v0.2.0 New Operators #62)
2. • Operator: upgrade CUTLASS (CUTLASS v3.2.0 #105)

Python

1. • Interface: Update Runtime.run() interface #96

Examples

1. • [ ] Example: add Llama2 example (Add a new example that implements the LLaMa model #102) moved to the next version
2. • Example: parallel matmul example (Parallel matmul example #64)

Bug Fix

• Fix a bug in the python wrapper #65
• Fix a simple bug in the code generator #66
• Fix Python identity op #75
• An IB bug fix & update tutorial #77
• Remove pytorch dependency #90
• Fix a bug in the matmul operator transpose option #94
• Fix a bug in the interface from_numpy() function #98
• Fix an AllReduce bug #97
• Fix an init() bug #99
• Fix ops_sendrecv_test #100
• Fixed allgather #103
• Fix a bug in the broadcast.h #101
• ARK v0.1.0 Known Bugs & Issues #35

Documents

1. • Docs: update documents (Minor updates & docs #76, Update install.md #78, Update documents #87)

CI

1. • Code Coverage: add code coverage (Update CI #110)
2. • Unit Tests: add a unit test pipeline (Azure pipelines #88)
3. • Unit Tests: New unit test framework for operators #91

[chhwang]

Released: https://github.com/microsoft/ark/releases/tag/v0.2.0