pytorch/xla

Pull Requests Commits

Disable CXX 11 ABI when building wheels (#2390)

jysohn23 committed 5 years ago

Verified 18fa011c

Issue notify in locked section to allow use from locals.

davidel committed 5 years ago

7a2f0493

Trim TF ops. (#2391)

davidel committed 5 years ago

Verified 354b2429

Move nll_loss2d_forward/backward to optional. (#2388)

ailzhang committed 5 years ago

Verified 63ff8fb3

Handling Tensor? to c10::optional<Tensor>. (#2372)

ailzhang committed 5 years ago

Verified 5aa8dbac

Remove unused registration for einsum. (#2385)

ailzhang committed 5 years ago

Verified 00f5d9a5

Remove resize_ in the special whitelist. (#2382)

ailzhang committed 5 years ago

Verified b5b36dfc

Remove all_to_all and collective_permute from the official API. (#2384)

taylanbil committed 5 years ago

Verified 046e304a

nll_loss2d/backward lowering (#2373)

JackCaoG committed 5 years ago

Verified c4f8873d

Disable test_discontiguous_out_cumsum which checks contiguity. (#2377)

ailzhang committed 5 years ago

Verified 7fc9c3a7

Call at::_convolution instead of at::native::_convolution (#2370)

zasdfgbnm committed 5 years ago

Verified 8cefb2a9

Disable test_masked_select_discontiguous. (#2367)

ailzhang committed 5 years ago

Verified cc3d52ab

Enable test_scatter_add_non_unique_index_xla on TPU

JackCaoG committed 5 years ago

57134f5b

Add celu/celu_ to aten_tensor_ops. (#2186)

ailzhang committed 5 years ago

Verified 9b55685a

Capture device by value.

dlibenzi committed 5 years ago

fc79ec0a

Pad the input on transposed conv forward with output_padding

JackCaoG committed 5 years ago

3be91400

Avoid adding HLO to XLA graph when there is no need.

dlibenzi committed 5 years ago

c5f440b5

Make cross replica reduce token a device data instead of XLA constant.

dlibenzi committed 5 years ago

b73615c0

Use python's environment when installing xla packages (#2344)

Mistobaan committed 5 years ago

Verified a42d78c3

Added MP test decorator.

dlibenzi committed 5 years ago

344ab71d

Fix typo for setup script on Colab notebook (#2336)

jysohn23 committed 5 years ago

Verified ba9876ad

Update ResNet18 colab to also enable XLA GPU (#2334)

jysohn23 committed 5 years ago

Verified 9137f6c7

Update script to also install CUDA wheels (#2331)

jysohn23 committed 5 years ago

Verified 5fc84157

Dsiable test_clamp_propagates_nans_xla

JackCaoG committed 5 years ago

51197a67

Update README.md (#2329)

jspisak committed 5 years ago

Verified eebfbabf

Do not use TLS for replication devices as calls into them from autograd are coming from "alien" threads.

dlibenzi committed 5 years ago

f43941b7

Minor refactoring.

dlibenzi committed 5 years ago

8504a130

Update docker tag to more recent one (#2324)

seemethere committed 5 years ago

Verified c4928c12

Add test case to ensure EmbeddingBag gradient computation works

gmagogsfm committed 5 years ago

d8d9b5ed

Added support for sets of detached devices, and MP for XLA CPU devices.

dlibenzi committed 5 years ago

26c28398

Newer Older