pytorch/xla

Pull Requests Commits

fix batch_size insistent in report (#7820)

zpcore committed 1 year ago

Verified f0bbaebd

Fix input output alias for custom inplace ops (#7822)

yitongh committed 1 year ago

Verified 8f794884

Use local TPU log dir in CI (#7816)

will-cromar committed 1 year ago

Verified ae2a3a4b

Add user guide for generating stablehlo composite op (#7826)

lsy323 committed 1 year ago

Verified 934771e7

Update fixing_op_info_test.md with addition for decomposed ops (#7825)

ManfeiBai committed 1 year ago

Verified 2dcdd455

Upgrade torch dependency to 2.4.0 plus misc changes: (#7821)

qihqi committed 1 year ago

Verified 60b9dfe7

Add ops `nn.functional.celu ... nn.functional.silu` (#7818)

ManfeiBai committed 1 year ago

Verified ec994104

Use torch export in test_composite (#7819)

qihqi committed 1 year ago

Verified 2eb4a60e

[benchmarks] Guard empty metrics on verifier runs. (#7809)

ysiraichi committed 1 year ago

Verified 9fbd64ac

Revert "make StepMarker's wait block for async device execution" (#7815)

JackCaoG committed 1 year ago

Verified b4335f8e

Support mark_dynamic (#7812)

JackCaoG committed 1 year ago

Verified 08d9595c

torch::lazy::MakeNode -> torch_xla::MakeNode (#7807)

JackCaoG committed 1 year ago

Verified c26b19eb

Remove stale `cloud-tpu-client` dependency (#7814)

will-cromar committed 1 year ago

Verified 52b81e41

remove openxla_eval in test (#7811)

zpcore committed 1 year ago

Verified c7d9755d

Support fp8e5m2 dtype (#7740)

lsy323 committed 1 year ago

Verified 1ed26267

[torch_xla2] Support `aten.avg_pool2d` and `aten.avg_pool3d` with `ceil_mode=True` (#7806)

chunnienc committed 1 year ago

Verified dd04c588

Allow user to pass a name to torch_xla.compile (#7802)

JackCaoG committed 1 year ago

Verified 6250c02d

Attach callback to a Tensor's underlying buffer (#7793)

will-cromar committed 1 year ago

Verified 02d83221

[benchmarks] Fix verifier error handling and OOM. (#7777)

ysiraichi committed 1 year ago

Verified 5bbc4b3d

Move `_xla_register_custom_call_target` implementation into `PjRtComputationClient` (#7801)

will-cromar committed 1 year ago

Verified 7135a7e9

[torch/xla2] Lower aten_copysign and corrcoef (#7713)

wenxindongwork committed 1 year ago

Verified a3dd4d81

Refresh xla2 doc (#7804)

qihqi committed 1 year ago

Verified 8969043e

Update XLA Pin (#7797)

lsy323 committed 1 year ago

Verified dd3b00ca

make StepMarker's wait block for async device execution (#7794)

JackCaoG committed 1 year ago

Verified dfd7b00d

Use reduce-scatter coalescing for FSDP (#6024)

jeffhataws committed 1 year ago

Verified 7fe070a6

Update function doc strings (#7757)

JackCaoG committed 1 year ago

Verified d29b7611

disable sdpa amp (#7800)

lsy323 committed 1 year ago

Verified 225ec7b3

Add 1-layer gradient accumulation test to check aliasing (#7692)

jeffhataws committed 1 year ago

Verified a5520e5d

Add .rst for quantized op doc (#7760)

lsy323 committed 1 year ago

Verified bc9b6061

Handle SPMD case inside of `ComputationClient::WaitDeviceOps` (#7796)

will-cromar committed 1 year ago

Verified 2766f9f7

Newer Older