pytorch/xla

Pull Requests Commits

Add sm_scale in ragged attention kernel (#8771)

yaochengji committed 246 days ago

Verified 2675e689

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

yaochengji committed 247 days ago

Verified 2e4f073a

Enable default buffer donation for gradient accumulation (#8758)

rpsilva-aws committed 247 days ago

Verified 6712eb91

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

vanbasten23 committed 247 days ago

Verified b18a65fa

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

zpcore committed 247 days ago

Verified 1ab8216a

Add start_trace and stop_trace API in profiler (#8743)

lsy323 committed 248 days ago

Verified b4ba17b5

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

bhavya01 committed 248 days ago

Verified 71d0ce65

Increase tolerance in test_scan_xla_patched_linear (#8749)

tengyifei committed 248 days ago

Verified ac430abf

Extend buffer donation aliasing APIs (#8721)

rpsilva-aws committed 248 days ago

Verified 07976424

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

tengyifei committed 249 days ago

Verified 6f020aad

Add padding ragged paged attention test (#8741)

vanbasten23 committed 249 days ago

Verified 5e1478d6

Improve the ragged kernel benchmarking script. (#8733)

vanbasten23 committed 250 days ago

Verified 79d1e86b

Add autograd function for mark_sharding (#8723)

bhavya01 committed 254 days ago

Verified 1acc987a

Add information to xla.launch (#8724)

pgmoka committed 254 days ago

Verified a08b7f86

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

vanbasten23 committed 254 days ago

Verified 0c4e82d6

Bump CUDNN version in docker image. (#8718)

ysiraichi committed 256 days ago

Verified 60160233

Bump XLA commit pin. (#8714)

ysiraichi committed 256 days ago

Verified ca248c7e

Support splitting physical axis in SPMD mesh (#8698)

tengyifei committed 260 days ago

Verified 8a712495

update pin (#8677)

lsy323 committed 261 days ago

Verified 6f423d0b

Fix CUDA plugin CI. (#8593)

ysiraichi committed 262 days ago

Verified 42edbe18

Include the alias ID in GetXLATensorDebugInfo (#8701)

rpsilva-aws committed 262 days ago

Verified 3a4ad6f3

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

bythew3i committed 262 days ago

Verified 06e1b591

Mention remote build in CONTRIBUTING.md (#8699)

tengyifei committed 263 days ago

Verified 45af5193

Add missing NeuronPlugin configure_single_process function (#8694)

jeffhataws committed 264 days ago

Verified 065cb5b1

Fix a bug in flash attention where kv_seq_len should divide block_k_major. (#8671)

zhangp365 committed 264 days ago

Verified cff9f4e0

Add another C++11 ABI docker reference (#8649)

tengyifei committed 267 days ago

Verified c0afda3a

[torchax] Update installation guide in readme (#8664)

lsy323 committed 267 days ago

Verified ace5e24b

Fix dk/dv autograd error on TPU flash attention (#8685)

zmelumian972 committed 267 days ago

Verified 0cd1fc2a

Added pow to autocast policy and unit test (#8684)

avizon-aws committed 267 days ago

Verified afaf0d05

Misc fixes (#8679)

qihqi committed 268 days ago

Verified 8b45e599

Newer Older