pytorch/xla

Pull Requests Commits

Keep the command consistent

chandrasekhard2 committed 2 years ago

40f55a2b

Add flexibility to define multiple configs based on TPU versions

chandrasekhard2 committed 2 years ago

e48ea2a9

Modify name to more generalized way to keep it open for v5 config as well

chandrasekhard2 committed 2 years ago

66cbf693

Add flag to switch v4 optimized config

chandrasekhard2 committed 2 years ago

2e4c7b7e

remove extra line

chandrasekhard2 committed 2 years ago

bb513b74

Change description to v4 instead of just v4-8 as this config imporves resnet performance even on v4 slices and pods

chandrasekhard2 committed 2 years ago

3cde6498

Rename the config name to OPTIMIZED_KWARGS_v4

chandrasekhard2 committed 2 years ago

ebfc74c3

Revert test set batch_size to 64

chandrasekhard2 committed 2 years ago

feeea8e1

Add feature to increase the number of host to device transfer threads

chandrasekhard2 committed 2 years ago

b0c748f4

fix linter (#4725)

JackCaoG committed 2 years ago

Verified 18674daf

Manually track type of SymNode (#4722)

ezyang committed 2 years ago

Verified add720fc

Update API GUIDE to include multi host training and add some colors (#4706)

JackCaoG committed 2 years ago

Verified a6d186f7

Move dependencies on Tensorflow under third_party/xla_client and torch_xla/csrc to use TSL where possible (#4717)

ddunl committed 2 years ago

Verified c939de06

Enable PjRt multi-GPU tests in CI (#4718)

jonb377 committed 2 years ago

Verified 6d7cc081

[PJRT] Detect number of local TPU chips (#4720)

will-cromar committed 2 years ago

Verified 46d47d35

Revert "Support tf tf repo macros in external repos. (#4708)" (#4719)

alanwaketan committed 2 years ago

Verified 9e4b01af

Update assert to self.assertTrue (#4702)

JackCaoG committed 2 years ago

Verified fb092d27

Remove .torch_pin from the Functionalization PR (#4715)

alanwaketan committed 2 years ago

Verified 0466784d

Functionalization integration (#4158)

bdhirsh committed 2 years ago

Verified e49df835

Support tf tf repo macros in external repos. (#4708)

stgpetrovic committed 2 years ago

Verified 5ca5686a

Lower UpsampleBilinear/Nearest2DBackward with scale factor on TPU (#4710)

lsy323 committed 2 years ago

Verified f594c29a

[PJRT] Remove start method from _run_singleprocess (#4709)

will-cromar committed 2 years ago

Verified d86b323d

Update GPU wheel link to also include CUDA 117 and 118 (#4705)

JackCaoG committed 2 years ago

Verified 5b5dd27f

Add test for dynamo and sgd optimizer (#4698)

JackCaoG committed 2 years ago

Verified ff15a8ac

remove windows related diff in patch (#4694)

lsy323 committed 2 years ago

Verified 89a9d7bc

Update env-setup.py (#4695)

kumar-selvakumaran committed 2 years ago

Verified 388135e6

Fix dynamo test no op (#4697)

JackCaoG committed 2 years ago

Verified 5f6e05c5

[PJRT] Use the new local rank env var to calculate local_ordinal (#4684)

will-cromar committed 2 years ago

Verified c5d4103a

fix starting profiler server (#4689)

lsy323 committed 2 years ago

Verified 3b106635

[FSDP] Replace all .data usages with modern approaches (#4485)

alanwaketan committed 2 years ago

Verified 5e6a32d0

Older