pytorch/xla

Pull Requests Commits

alanwaketan committed 1 year ago

7098b3da

Make kSharding as default

alanwaketan committed 1 year ago

c70900b2

address comments

alanwaketan committed 1 year ago

52bf2901

Fix rebase error

alanwaketan committed 1 year ago

d9c3e24c

alanwaketan committed 1 year ago

ef834bf4

alanwaketan committed 1 year ago

d77f85c9

Make test case and fix it

alanwaketan committed 1 year ago

3ececd34

Make it even more extensible

alanwaketan committed 1 year ago

a6d1be2b

expand custom sharding

alanwaketan committed 1 year ago

cfa9c560

initial commit for pause

alanwaketan committed 1 year ago

b3ee21a1

Disallow mark manual sharding on data tensors

alanwaketan committed 1 year ago

a1120519

alanwaketan committed 1 year ago

0e0ad181

alanwaketan committed 1 year ago

fa34ef68

alanwaketan committed 1 year ago

74b494df

Add dynamo expand test. (#6659)

ysiraichi committed 1 year ago

Verified 58a412cb

Build plugin during nightly build (#6917)

will-cromar committed 1 year ago

Verified 5a269d7b

Install libtpu for release dockers (#6919)

alanwaketan committed 1 year ago

Verified ec5615c9

[SPMD] Support manual sharding (#6915)

alanwaketan committed 1 year ago

Verified e5513ff4

fix pixel_shuffle return empty (#6907)

zpcore committed 1 year ago

Verified 43f10715

Build TPU package in CI (#6918)

will-cromar committed 1 year ago

Verified 7375a8ad

Add an option to not generate weights. (#6909)

qihqi committed 1 year ago

Verified fc11bbbf

Add `xla.sync` (#6914)

will-cromar committed 1 year ago

Verified 42452717

[Fori_loop|While_loop] Fori loop wrapped from while_loop (#6850)

ManfeiBai committed 1 year ago

Verified f0354ecc

[torch_xla2] Combined tensor.XLADispatchMode() and functions.XLAFunctionMode() (#6902)

will-cromar committed 1 year ago

Verified cfc70e68

Fix runtime error when run dynamo with a profiler scope (#6913)

JackCaoG committed 1 year ago

Verified 7dfbef02

Add ansible stage for building CUDA plugin (#6910)

will-cromar committed 1 year ago

Verified a170ffe6

[Distributed] Switch all_reduce to use the new functional collective op (#6887)

yifuwang committed 1 year ago

Verified a816c42f

fix wheel link (#6908)

zpcore committed 1 year ago

Verified 756b0eca

Update XLA pin to 04/08/2024 (#6904)

JackCaoG committed 1 year ago

Verified fd04a20c

[torch_xla2] Implement some common tensor constructors (#6874)

will-cromar committed 1 year ago

Verified 786484d2

Older