pytorch/xla

Pull Requests Commits

Use pages_per_seq * page_size instead of directly passing max_model_len

yaochengji committed 273 days ago

25becbf2

Fix ragged_paged_attention op signature (#8943)

yaochengji committed 273 days ago

Verified 80929a91

add rc4 trigger (#8941)

zpcore committed 275 days ago

Verified ac9a39f4

Add block size table for ragged_paged_attention (#8942)

yaochengji committed 275 days ago

Verified cbff76f5

Add alternative dynamo backend (#8893)

qihqi committed 276 days ago

Verified fb6038de

Trigger build for 2.6.1 patch (#8938)

zpcore committed 276 days ago

Verified 379ebd5f

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

tengyifei committed 276 days ago

Verified 9e23ca85

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

iwknow committed 276 days ago

Verified 145c33ae

Fix up pin update (#8935)

tengyifei committed 276 days ago

Verified 980ead54

Include torchax in torch_xla (#8895)

tengyifei committed 276 days ago

Verified bb5c6a0b

pin update (#8908)

lsy323 committed 277 days ago

Verified 6400e169

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

bythew3i committed 277 days ago

Verified c044c691

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

yaochengji committed 277 days ago

Verified 6c3f2313

Clean up deprecated APIs (#8927)

zpcore committed 278 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 278 days ago

Verified f0881b5a

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

dvhg committed 279 days ago

Verified 2d6f57a3

Fix cuda dependency not found (#8903)

zpcore committed 279 days ago

Verified dbb03aa1

Lower `isneginf()`. (#8912)

ysiraichi committed 279 days ago

Verified cba9ff9f

Fix fp8 test (#8909)

lsy323 committed 279 days ago

Verified 985fa188

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

vanbasten23 committed 280 days ago

Verified 450b4dce

point build to rc3 (#8907)

zpcore committed 280 days ago

Verified 2999e72b

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

tengyifei committed 283 days ago

Verified 6d88c089

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

lsy323 committed 283 days ago

Verified fe3bb7f6

Remove dynamic grid (#8896)

bythew3i committed 284 days ago

Verified b24e6e95

Update cudnn dependency for newer cuda versions (#8894)

zpcore committed 284 days ago

Verified 92873377

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

pgmoka committed 284 days ago

Verified e341ff09

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

bythew3i committed 284 days ago

Verified 7a3c0518

Encapsulate Mesh invariants (#8882)

rpsilva-aws committed 285 days ago

Verified 96ad8f5f

Avoid unnecessary copy in TensorSource (#8849)

lsy323 committed 285 days ago

Verified 8dc5b496

update Dockerfile config for CI (#8886)

zpcore committed 285 days ago

Verified 3066b034

Older