pytorch/xla

Pull Requests Commits

add missing file

JackCaoG committed 1 year ago

ecc020f5

handle all_reduce in place for eager mode

JackCaoG committed 1 year ago

41d3e781

JackCaoG committed 1 year ago

34b04b8e

Handle in plac eops

JackCaoG committed 1 year ago

13c642c0

add basic eager tests

JackCaoG committed 1 year ago

6313c8ab

Type annotation for benchmarks/ (#7289)

zpcore committed 1 year ago

Verified 98dd99e7

Update the instruction for how to pin upstream pr (#7314)

JackCaoG committed 1 year ago

Verified cb6549aa

Update XLA Pin (#7308)

JackCaoG committed 1 year ago

Verified 2cfa13ca

Wrap adding logging options in try / catch. (#7307)

qihqi committed 1 year ago

Verified c2753715

[Fori_loop|While_loop] Enable while_loop/fori_loop, add test case (#7157)

ManfeiBai committed 1 year ago

Verified 91389c7c

[benchmarks] Initialize `moco` and TPU experiments on XLA. (#7257)

ysiraichi committed 1 year ago

Verified abf5c31b

reduce dcoder model size to make eager runable on v2-8 (#7288)

JackCaoG committed 1 year ago

Verified a0b76fce

use the eager_mode_context for some of the tests (#7276)

JackCaoG committed 1 year ago

Verified 1492eba8

Add tpu device id for v6e (#7269)

bhavya01 committed 1 year ago

Verified e8601ea3

Use np.prod instead of np.product (#7301)

JackCaoG committed 1 year ago

Verified 5fa23c1a

`select` op support in torchxla2 (#7293)

miladm committed 1 year ago

Verified f50e9519

Fix a few tig ops and `abs` in torchxla2 (#7294)

miladm committed 1 year ago

Verified 61c794d8

added argsort support (#7290)

miladm committed 1 year ago

Verified 8b5c7ef6

Use default github token for docs preview (#7279)

will-cromar committed 1 year ago

Verified 5fd6d0fc

Fix recv docstring (#7292)

BitPhinix committed 1 year ago

Verified 8c669580

Add vim to release wheel (#6604)

vanbasten23 committed 1 year ago

Verified 778eb03f

Add a guide to support new models (#7298)

qihqi committed 1 year ago

Verified 1c748568

Remove use_dynamo_custom_op flag from unit tests (#7287)

wonjoo-wj committed 1 year ago

Verified 3f22daa9

Adding a bunch of ops to make torchbench's BERT_pytorch pass (#7129)

qihqi committed 1 year ago

Verified cc55cc4f

Fix benchmark not to assume the input is a list of tensor. (#7280)

vanbasten23 committed 1 year ago

Verified afd31c3a

Enable bucketized all-reduce for gradients (#7216)

amithrm committed 1 year ago

Verified 28f9887b

Refactor SPMD docs (#7283)

JackCaoG committed 1 year ago

Verified f075bafe

Clean up and add doc for dynamo_mark_sharding (#7281)

wonjoo-wj committed 1 year ago

Verified 61e03890

[Doc] Add an instruction to install nightly wheels. (#7275)

vanbasten23 committed 1 year ago

Verified 62015ee1

Disable remote cache for PRs from forks (#7282)

will-cromar committed 1 year ago

Verified a7976d07

Older