pytorch/xla

Pull Requests Commits

remove the permute

yaochengji committed 203 days ago

b5d0cc17

yaochengji committed 207 days ago

8e316dc0

yaochengji committed 207 days ago

4a0a6106

yaochengji committed 207 days ago

d583dc88

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 208 days ago

Verified 2e4f073a

Enable default buffer donation for gradient accumulation (#8758)

rpsilva-aws committed 208 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 209 days ago

Verified b4ba17b5

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

bhavya01 committed 209 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 217 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 221 days ago

Verified 8a712495

update pin (#8677)

lsy323 committed 222 days ago

Verified 6f423d0b

Fix CUDA plugin CI. (#8593)

ysiraichi committed 223 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

Mention remote build in CONTRIBUTING.md (#8699)

tengyifei committed 224 days ago

Verified 45af5193

Add missing NeuronPlugin configure_single_process function (#8694)

jeffhataws committed 225 days ago

Verified 065cb5b1

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

zhangp365 committed 225 days ago

Verified cff9f4e0

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

tengyifei committed 228 days ago

Verified c0afda3a

Older