pytorch/xla

Pull Requests Commits

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

Change `pjrt://` init method to `xla://` (#5560)

will-cromar committed 2 years ago

Verified e2a15714

Add resnet50-weight-quant colab notebook (#5407)

lsy323 committed 2 years ago

Verified a1b92796

[Pin Update] Version 20230826 (#5527)

alanwaketan committed 2 years ago

Verified 9732daf6

fix: missing import numpy (#5533)

wbmc committed 2 years ago

Verified b9d567ad

Make cpu tensor on XLA dynamo backend a warning instead of error (#5549)

JackCaoG committed 2 years ago

Verified e51d28b6

Add nightly wheel for py3.10 and cuda 11.8 (#5547)

vanbasten23 committed 2 years ago

Verified f38e4a50

Fix pip install command typo in GPU doc. (#5545)

vanbasten23 committed 2 years ago

Verified 2c8ebd14

Update nightly to 2.2 (#5542)

JackCaoG committed 2 years ago

Verified 04118d5e

Allow to override PyTorch git revision and set it to v2.1.0-rc2 for 2.1 releases (#5543)

mateuszlewko committed 2 years ago

Verified 3f978dfb

Add option to unbundle libtpu (#5534)

will-cromar committed 2 years ago

Verified d757e85e

Fix FSDP for Models with Frozen Weights (#5484)

Liyang90 committed 2 years ago

Verified 3bdb4b10

Newer Older