pytorch/xla

Pull Requests Commits

yaochengji committed 73 days ago

ef162a26

yaochengji committed 75 days ago

eabe16bc

yaochengji committed 75 days ago

c7c72bc6

yaochengji committed 75 days ago

9caa2cf8

yaochengji committed 75 days ago

baec999b

fix ragged_paged_attention op signature

yaochengji committed 75 days ago

1054c14e

Add block size table for ragged_paged_attention (#8942)

yaochengji committed 75 days ago

Verified cbff76f5

Add alternative dynamo backend (#8893)

qihqi committed 75 days ago

Verified fb6038de

Trigger build for 2.6.1 patch (#8938)

zpcore committed 76 days ago

Verified 379ebd5f

Update and rename torch_xla2.yml to torchax.yml (#8936)

tengyifei committed 76 days ago

Verified 9e23ca85

Rewrite scan-based GRU based on nn.GRU (#8914)

iwknow committed 76 days ago

Verified 145c33ae

Fix up pin update (#8935)

tengyifei committed 76 days ago

Verified 980ead54

Include torchax in torch_xla (#8895)

tengyifei committed 76 days ago

Verified bb5c6a0b

pin update (#8908)

lsy323 committed 77 days ago

Verified 6400e169

[ragged-paged-attn] Unify kv strided load to one. (#8929)

bythew3i committed 77 days ago

Verified c044c691

Add heuristic default block sizes for different cases in ragged attention kernel (#8922)

yaochengji committed 77 days ago

Verified 6c3f2313

Clean up deprecated APIs (#8927)

zpcore committed 77 days ago

Verified f1fe8bc0

Use f32 scratch for output so we only need to transfer output with desired dtype back to HBM. (#8924)

vanbasten23 committed 77 days ago

Verified f0881b5a

[torchax] Support linalg.det and linalg.lu_solve (#8872)

dvhg committed 79 days ago

Verified 2d6f57a3

Fix cuda dependency not found (#8903)

zpcore committed 79 days ago

Verified dbb03aa1

Lower `isneginf()`. (#8912)

ysiraichi committed 79 days ago

Verified cba9ff9f

Fix fp8 test (#8909)

lsy323 committed 79 days ago

Verified 985fa188

Add Pallas tutorials to the Pallas guide at torch_xla. (#8904)

vanbasten23 committed 80 days ago

Verified 450b4dce

point build to rc3 (#8907)

zpcore committed 80 days ago

Verified 2999e72b

Use shard_as in scan to ensure that inputs and their gradients have the same sharding (#8879)

tengyifei committed 82 days ago

Verified 6d88c089

[Build] Pin cmake to build PyTorch (#8902)

lsy323 committed 83 days ago

Verified fe3bb7f6

Remove dynamic grid (#8896)

bythew3i committed 84 days ago

Verified b24e6e95

Update cudnn dependency for newer cuda versions (#8894)

zpcore committed 84 days ago

Verified 92873377

8740 add single processing to getting started instructions (#8875)

pgmoka committed 84 days ago

Verified e341ff09

[ragged-paged-attn] Combine k_pages and v_pages on num_kv_head (#8892)

bythew3i committed 84 days ago

Verified 7a3c0518

Older