pytorch/xla

Pull Requests Commits

Fix some more core aten ops (#6519)

wonjoo-wj committed 1 year ago

Verified b198a48b

[benchmarks] Fix execution with AMP precision. (#6512)

ysiraichi committed 1 year ago

Verified 408b3766

[SPMD] input sharding for both train & test data loading in imagenet example (#6515)

yeounoh committed 1 year ago

Verified a5692c20

Read from constants if it exists (#6510)

qihqi committed 1 year ago

Verified 0fa24a13

[benchmarks] Increase compilation cache. (#6509)

ysiraichi committed 1 year ago

Verified 7a1ee8f8

add unbounded dynamism test for some aten ops, support dynamism on add (#6443)

lsy323 committed 1 year ago

Verified 8d91ff58

Correctly set the exit code when an exception is raised in the atexit callback _prepare_to_exit. (#6383)

vanbasten23 committed 1 year ago

Verified 1cd3f460

Reapply lower randperm (#6482)

changm committed 1 year ago

Verified 157e06e4

Fix a bug in convolution for torch_xla2. Enable tests (#6496)

qihqi committed 1 year ago

Verified b9352685

[benchmarks] Remove `--strict-compatible` argument. (#6491)

ysiraichi committed 1 year ago

Verified 6f4409fa

[FSDPv2] Use the global mesh API (#6500)

alanwaketan committed 1 year ago

Verified ef8f6c3f

[FSDPv2] Enable auto-wrapping (#6499)

alanwaketan committed 1 year ago

Verified cbe0ca84

[SPMD] Introduce global mesh (#6498)

alanwaketan committed 1 year ago

Verified 9b123349

CI: increase build timeout from 180 to 240 minutes (#6492)

cota committed 1 year ago

Verified 6ae122d3

[export] Remove torch._export.export uses (#6486)

angelayi committed 1 year ago

Verified fba464b1

Remove random tests in test_core_aten_ops (#6436)

qihqi committed 1 year ago

Verified acf12a38

[benchmarks] Add `no-skip` argument. (#6484)

ysiraichi committed 1 year ago

Verified e8a5f007

torchxla2 add more ops (#6490)

qihqi committed 1 year ago

Verified a4dc6cf4

Add docs on how to integrate with dynamo (#6459)

changm committed 1 year ago

Verified d0dd1a7f

[fix-up] nightly.sh: define BM_DIR earlier in the file (#6487)

cota committed 1 year ago

Verified 14b1ee7d

nightly.sh: run llama2 benchmarks (#6485)

cota committed 1 year ago

Verified c9d38348

Get stablehlo version programmatically (#6476)

lsy323 committed 1 year ago

Verified 99b56c43

Remove assert on zero_point dtype of quantize/dequantize op (#6475)

lsy323 committed 1 year ago

Verified 81284808

benchmarks: add script to run llama2 inference benchmarks (#6481)

cota committed 1 year ago

Verified a9dc9f12

aggregate.py: add format=latest_grouped (#6472)

cota committed 1 year ago

Verified f37c4542

Infer output dtype for rsub using generic template (#6473)

lsy323 committed 1 year ago

Verified 9884e05a

Fix some more core aten ops (#6462)

wonjoo-wj committed 1 year ago

Verified e8dfc865

Add more ops to torch_xla2 (#6474)

qihqi committed 1 year ago

Verified 6a0cb712

Wrap constant arg in XLAExportInterpreter (#6460)

lsy323 committed 1 year ago

Verified 5dbbb284

bring TPU CI bakc to green by reinstall torch/torchvision (#6458)

ManfeiBai committed 1 year ago

Verified e6baedd5

Newer Older