pytorch/xla

Pull Requests Commits

yaochengji committed 203 days ago

4584a213

yaochengji committed 203 days ago

b6951ed3

yaochengji committed 204 days ago

703953ca

write _shard_map; refactor flash attention to support 5d inputs. (#8730)

qihqi committed 204 days ago

Verified cee0820e

Introduce a GRU module implemented with scan (#8777)

tengyifei committed 204 days ago

Verified 0017ff16

Fix build on Python 3.9 (#8759)

tengyifei committed 206 days ago

Verified c76e9497

Misc changes to make torchax runnable on GPU. (#8756)

qihqi committed 206 days ago

Verified 60001c74

[scan] Make sure inputs into `fn` are not `device_data` IR nodes (#8769)

tengyifei committed 206 days ago

Verified 2feb0ac7

Add sm_scale in ragged attention kernel (#8771)

yaochengji committed 207 days ago

Verified 2675e689

Revert "Change num_seqs type from int to torch.Tensor" (#8767)

yaochengji committed 207 days ago

Verified 2e4f073a

Enable default buffer donation for gradient accumulation (#8758)

rpsilva-aws committed 207 days ago

Verified 6712eb91

Change num_seqs type from int to torch.Tensor (#8736)

vanbasten23 committed 208 days ago

Verified b18a65fa

Lower `as_strided_copy` use fast path with `slice` (#8734)

zpcore committed 208 days ago

Verified 1ab8216a

Add start_trace and stop_trace API in profiler (#8743)

lsy323 committed 208 days ago

Verified b4ba17b5

Add --network=host to TPU docker build command (#8735)

bhavya01 committed 208 days ago

Verified 71d0ce65

Increase tolerance in test_scan_xla_patched_linear (#8749)

tengyifei committed 209 days ago

Verified ac430abf

Extend buffer donation aliasing APIs (#8721)

rpsilva-aws committed 209 days ago

Verified 07976424

Introduce apply_xla_patch_to_nn_linear and test that in a scan (#8739)

tengyifei committed 210 days ago

Verified 6f020aad

Add padding ragged paged attention test (#8741)

vanbasten23 committed 210 days ago

Verified 5e1478d6

Improve the ragged kernel benchmarking script. (#8733)

vanbasten23 committed 211 days ago

Verified 79d1e86b

Add autograd function for mark_sharding (#8723)

bhavya01 committed 215 days ago

Verified 1acc987a

Add information to xla.launch (#8724)

pgmoka committed 215 days ago

Verified a08b7f86

Split page indices in the ragged paged attention. (#8688)

vanbasten23 committed 215 days ago

Verified 0c4e82d6

Bump CUDNN version in docker image. (#8718)

ysiraichi committed 216 days ago

Verified 60160233

Bump XLA commit pin. (#8714)

ysiraichi committed 217 days ago

Verified ca248c7e

Support splitting physical axis in SPMD mesh (#8698)

tengyifei committed 220 days ago

Verified 8a712495

update pin (#8677)

lsy323 committed 222 days ago

Verified 6f423d0b

Fix CUDA plugin CI. (#8593)

ysiraichi committed 222 days ago

Verified 42edbe18

Include the alias ID in GetXLATensorDebugInfo (#8701)

rpsilva-aws committed 223 days ago

Verified 3a4ad6f3

Integrate ragged paged attn to pytorch/xla (#8692)

bythew3i committed 223 days ago

Verified 06e1b591

Older