pytorch/xla

Pull Requests Commits

Update the amp.md according to the latest changes

yeounoh committed 2 years ago

a36265f5

autocat test helper dtype default needs to be unset (#5622)

yeounoh committed 2 years ago

Verified 63a90204

Fix log spam when libtpu is loaded (#5619)

will-cromar committed 2 years ago

Verified eee1f2dd

Add cuda==12.1 trigger (#5617)

vanbasten23 committed 2 years ago

Verified 45bdf70b

Remove `on_device_shape` and `logical_on_device_shape` (#5546)

will-cromar committed 2 years ago

Verified dbb92c96

Clean up the redundant arg_min/max releated file (#5605)

zpcore committed 2 years ago

Verified a4874e2d

Allow running with bfloat16 on XLA:GPU autocast (#5598)

yeounoh committed 2 years ago

Verified b214c104

increase the GPU coverge timeout to be same as the GPU test (#5610)

JackCaoG committed 2 years ago

Verified e9c38170

Compare the Dynamo vs Non-Dynamo res instead of cpu res (#5609)

JackCaoG committed 2 years ago

Verified 95045e03

Print the actual opsharding on PJRTShardedData::ToString (#5596)

JackCaoG committed 2 years ago

Verified bdb075b2

Update torch.compile usage int est (#5584)

JackCaoG committed 2 years ago

Verified 83692be6

Workaround that einsum with FakeTensor input (#5597)

JackCaoG committed 2 years ago

Verified 786722c5

Build `xrt` branch against PyTorch 2.1 (#5601)

will-cromar committed 2 years ago

Verified 2ae3fa7b

XLA pin downgrade to 20230825 (#5592)

alanwaketan committed 2 years ago

Verified 9e1f87d9

custom autograd op redispatching fix (#5600)

bdhirsh committed 2 years ago

Verified 5dd9108a

Update op argmin and argmax from manual lowering to codegen. (#5586)

zpcore committed 2 years ago

Verified a008de35

Don't trigger CI build on release tag push (#5594)

will-cromar committed 2 years ago

Verified ed7879df

Set PyTorch git revision to v2.1.0-rc5 for 2.1 releases (#5572)

ManfeiBai committed 2 years ago

Verified 6b380c78

Support kwargs in export and converting to stablehlo (#5563)

qihqi committed 2 years ago

Verified fc48cfe2

Enable xla:gpu autocast for bfloat16 if not restricted (#5570)

yeounoh committed 2 years ago

Verified 64bbf151

Fix release branch and tag patterns for GitHub Actions (#5587)

will-cromar committed 2 years ago

Verified 6dd13a10

Improve bernoulli rng-bit-generation memory footprint (#5581)

yeounoh committed 2 years ago

Verified 26a81a1e

fixing num_local_processes typo (#5573)

aws-kingrj committed 2 years ago

Verified c77e2499

Add bundled libtpu build for Python 3.10 (#5564)

will-cromar committed 2 years ago

Verified 2471e4da

Update gpu.md to use CUDA 11.8 image since, CUDA 11.7 image is stale (#5578)

yeounoh committed 2 years ago

Verified 2c11351b

Move where clear pending IR is called to avoid crash (#5552)

JackCaoG committed 2 years ago

Verified 9b120096

[author: jluntamazon] Adding more explicit HLO lowering control by exposing LoweringContext… (#5431)

aws-kingrj committed 2 years ago

Verified 41b38f55

Handle dynamo function without input (#5565)

JackCaoG committed 2 years ago

Verified 0aa9e82e

XLA bindings for torch_neuronx.trace and profile bindings (#5502)

mrnikwaws committed 2 years ago

Verified 547ea8b5

Update PJRT documentation for the 2.1 release (#5557)

will-cromar committed 2 years ago

Verified 6a14f6d9

Older