pytorch/xla

Pull Requests Commits

Partially Codegen Baddbmm (#4222)

ManfeiBai committed 3 years ago

Verified 1251261c

Fix a corner case where view tensor update will be ignore by mark_step (#4232)

JackCaoG committed 3 years ago

Verified ba9f0df0

fix casting SymInt to XLASymNodeImpl (#4212)

Krovatkin committed 3 years ago

Verified d471d08e

[SPMD] do add instead of matmul in SPMD unit testing (#4237)

yeounoh committed 3 years ago

Verified 7a4ee34d

Companion PR for pytorch/pytorch#89326 (#4240)

Nikita Karetnikov committed 3 years ago

Verified 281f3696

Disable failing take python op tests on TPU (#4231)

wonjoo-wj committed 3 years ago

Verified 4fdee512

inplace operation within no_grad context for all_gather (#4235)

aws-rhsoln committed 3 years ago

Verified 0e955758

XLA Op Key Registration should only happen once. (#4228)

JackCaoG committed 3 years ago

Verified 98ecca30

model should be on xla device for Dynamo torchxla_trace_once backend (#4205)

JackCaoG committed 3 years ago

Verified 36e2acab

Add const override flag to getstaticValue (#4229)

JackCaoG committed 3 years ago

Verified 6a792ddc

move all patches to different files and add comments (#4225)

JackCaoG committed 3 years ago

Verified bbcea0f2

Add python test for sizeAdd node and implement `int_` for XLASymNodeImpl. (#4218)

vanbasten23 committed 3 years ago

Verified f790bc8a

Introduce virtual device class (#4091)

steventk-g committed 3 years ago

Verified b2bd721c

Replace XLA_COUNTER with TORCH_LAZY_COUNTER (#4208)

alanwaketan committed 3 years ago

Verified 788001c5

Improve XLASynNodeImpl str() format (#4216)

vanbasten23 committed 3 years ago

Verified a46c616e

Add a api to reset and extract RNG Seed (#4196)

JackCaoG committed 3 years ago

Verified 0daf6fe9

Implement `xm.rendezvous` with XLA collective communication for PJRT (#4181)

will-cromar committed 3 years ago

Verified a9eb3459

[SPMD] fix sharding util bugs & implement unit test cases for 5D tile sharding (#4192)

yeounoh committed 3 years ago

Verified 95fcf680

Run cached graph api should grab device lock first (#4195)

JackCaoG committed 3 years ago

Verified 08a9c5dd

Add the dynamic shape unit test the CI. (#4180)

vanbasten23 committed 3 years ago

Verified a843d8bf

Create new Cloud Build workflow for nightly images/packages (#4188)

will-cromar committed 3 years ago

Verified 3f102965

[LTC] Adopt getBackend()->GetComputationDataFromNode() (#4191)

alanwaketan committed 3 years ago

Verified dd9b67ff

Fix multi-gpu metrics report (#4187)

ymwangg committed 3 years ago

Verified 44362e6d

[LTC] Use getBackend()->ExecuteComputation() (#4186)

alanwaketan committed 3 years ago

Verified ee65ff15

[LTC] Use torch::lazy::Util::ComputePostOrder (#4178)

alanwaketan committed 3 years ago

Verified ae35e1e2

unblock dynamo integration, get autograd ops to work with python tensor subclasses (#4165)

bdhirsh committed 3 years ago

Verified 4a0aaefb

[WIP] Partially Codegen Atan2 (#4184)

ManfeiBai committed 3 years ago

Verified b5496122

Fix check_tensor_need_materialization (#4185)

JackCaoG committed 3 years ago

Verified b1f167e6

[LTC] Make some torch::lazy::Node* to const T (#4182)

alanwaketan committed 3 years ago

Verified 0115cc00

Additional dynamo test cases (#4175)

JackCaoG committed 3 years ago

Verified bc2694ff

Newer Older