pytorch/xla

Pull Requests Commits

Add cache to `value_and_grad_partitioned` (#9163)

iwknow committed 112 days ago

Verified d1ddeed6

Introduce annotate_custom_sharding binding (#9203)

rpsilva-aws committed 112 days ago

Verified 402612d1

Deprecate `devkind` field from `xla_model.xla_device` (#9284)

ghpvnist committed 112 days ago

Verified 5771cd31

Add w8a8 quantized matmul torchxla wrapper (#9290)

vanbasten23 committed 112 days ago

Verified a6f2b27d

Add a script to update local/forked repos to match upstream (#9288)

zhanyong-wan committed 112 days ago

Verified bc8194e0

Add w8a8 quantized matmul kernel (#9278)

vanbasten23 committed 112 days ago

Verified 25048887

[Kernel] add heuristic gmm block sizes choosing logic (#9289)

yaochengji committed 112 days ago

Verified e51af253

Disable AMP by default on CPU (#9218)

haifeng-jin committed 112 days ago

Verified 275f6e92

Split training tests to separate test shard (#9281)

ghpvnist committed 113 days ago

Verified 3e556dc8

Remove a few CUDA tests in preparation for #9202 (#9286)

ghpvnist committed 113 days ago

Verified 97b75eab

Support fp8 lowering in scalar (#9283)

yaochengji committed 113 days ago

Verified d18fc759

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

qihqi committed 114 days ago

Verified c71ff48b

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

zhanyong-wan committed 114 days ago

Verified b177f3d3

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

qianminj123 committed 114 days ago

Verified 8b8c1a00

configurable number of thread for dcp (#9188)

aws-zhenguo committed 114 days ago

Verified d4c1be37

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

wzhang313 committed 114 days ago

Verified ac4d411c

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

qihqi committed 114 days ago

Verified 1f408008

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

ghpvnist committed 114 days ago

Verified 022ecde6

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

zhanyong-wan committed 114 days ago

Verified a159c9c8

Add instruction for exporting inlined constant (#8707)

qihqi committed 114 days ago

Verified 0b28030c

Remove CI jobs for GPU (#9277)

zhanyong-wan committed 114 days ago

Verified 9c21366a

Refactor GetNumDevices and create GetNumGlobalDevices (#9184)

pgmoka committed 115 days ago

Verified 1a82b709

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

zhanyong-wan committed 115 days ago

Verified fbc44603

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

lsy323 committed 117 days ago

Verified c2a9b32b

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

zhanyong-wan committed 117 days ago

Verified cdd0202e

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

zhanyong-wan committed 117 days ago

Verified 222dc4ab

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

zhanyong-wan committed 118 days ago

Verified d1ca3c48

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

qihqi committed 118 days ago

Verified 2196d0cf

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

ysiraichi committed 118 days ago

Verified 06c5533d

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

yaochengji committed 118 days ago

Verified 2edcd2e3

Newer Older