pytorch/xla

Pull Requests Commits

yaochengji committed 1 year ago

3a100d93

yaochengji committed 1 year ago

227fc0b4

yaochengji committed 1 year ago

90202cf2

[Kernel] add heuristic gmm block sizes choosing logic

yaochengji committed 1 year ago

41a5002e

Support fp8 lowering in scalar (#9283)

yaochengji committed 1 year ago

Verified d18fc759

[torchax] Make copy_ works on different device. (#9211)

qihqi committed 1 year ago

Verified c71ff48b

Remove GPU tests from test/run_tests.sh (#9280)

zhanyong-wan committed 1 year ago

Verified b177f3d3

[torchax] Fixes test for cat pytorch/xla#7398 (#9047)

qianminj123 committed 1 year ago

Verified 8b8c1a00

configurable number of thread for dcp (#9188)

aws-zhenguo committed 1 year ago

Verified d4c1be37

dynamo_bridge.py: convert tuple to list before none_remover is called (#9279)

wzhang313 committed 1 year ago

Verified ac4d411c

Add interop with flax (Part 1) (#9176)

qihqi committed 1 year ago

Verified 1f408008

Migrate runtime.xla_device in favor of core.xla_model.xla_device (#9200)

ghpvnist committed 1 year ago

Verified 022ecde6

Update OpenXLA to 2025/5/30. (#9271)

zhanyong-wan committed 1 year ago

Verified a159c9c8

Add instruction for exporting inlined constant (#8707)

qihqi committed 1 year ago

Verified 0b28030c

Remove CI jobs for GPU (#9277)

zhanyong-wan committed 1 year ago

Verified 9c21366a

Refactor GetNumDevices and create GetNumGlobalDevices (#9184)

pgmoka committed 1 year ago

Verified 1a82b709

Add scripts/update_deps.py to automate the updating the dependencies. (#9270)

zhanyong-wan committed 1 year ago

Verified fbc44603

[torchax] Support View in 'jax_view' (#9273)

lsy323 committed 1 year ago

Verified c2a9b32b

Tune CI job time-outs to be closer to the actual durations. (#9267)

zhanyong-wan committed 1 year ago

Verified cdd0202e

Update libtpu and jax to their latest nightly builds. (#9264)

zhanyong-wan committed 1 year ago

Verified 222dc4ab

In pjrt runtime client, raise a Python exception if XLA compilation fails. (#9138)

zhanyong-wan committed 1 year ago

Verified d1ca3c48

[torchax] Added support for bicubic and billinear resampling (#9222)

qihqi committed 1 year ago

Verified 2196d0cf

Run `test_collective_permute.py` on TPU CI. (#9257)

ysiraichi committed 1 year ago

Verified 06c5533d

[Kernel] support kv cache quantization in ragged attention kernel (#9249)

yaochengji committed 1 year ago

Verified 2edcd2e3

Improve `torch_xla.compile` documentation (#9194)

sdasgup3 committed 1 year ago

Verified 248a8b33

Document the difference between tracing time and execution time (#9133)

sdasgup3 committed 1 year ago

Verified 513ab7ca

Convert some XLA_CHECKs to fatal errors. (#9263)

zhanyong-wan committed 1 year ago

Verified d43142cb

Optimize build_developer.sh for the common workflow. (#9262)

zhanyong-wan committed 1 year ago

Verified 51a3bae2

Update README.md with supported Python version (#9239)

zzzwen committed 1 year ago

Verified 808773a9

[Kernel] add group_offset and transpose_rhs support in gmm kernel (#9251)

yaochengji committed 1 year ago

Verified 9871aed5

Older