pytorch/xla

Pull Requests Commits

typo fix (#8990)

haifeng-jin committed 243 days ago

Verified 428c86d2

scan-based GRU falls back to nn.GRU when `bidirectional` is true. (#8984)

iwknow committed 243 days ago

Verified 0239d8c5

Fix the source param sharding for GradAcc API (#8999)

rpsilva-aws committed 243 days ago

Verified db49eddb

Set scoped vmem for paged attention (#8988)

zpcore committed 243 days ago

Verified 366f248a

Fix merge conflict (#8991)

zpcore committed 244 days ago

Verified 083934fd

Add an option for JittableModule to dedup parameters. (#8965)

qihqi committed 244 days ago

Verified cce7cc07

Add a helper class to handle mesh and sharding (#8967)

qihqi committed 244 days ago

Verified 78824755

Fix splash attention test (#8978)

zpcore committed 244 days ago

Verified 13155c9c

Fix call_jax hashing (#8981)

zpcore committed 244 days ago

Verified a48d6a97

Update to ubuntu latest (#8979)

qihqi committed 244 days ago

Verified a37a5852

@assume_pure (#8962)

tengyifei committed 245 days ago

Verified cf764ffa

Avoid re-computing computation hashes (#8976)

rpsilva-aws committed 245 days ago

Verified bcca5eba

Disable one splash attention test (#8970)

zpcore committed 246 days ago

Verified ba5cd4cb

Add tuned parameters for Qwen/Qwen2.5-32B (#8966)

yarongmu-google committed 249 days ago

Verified 4eea5e13

Adapt Splash Attention from TorchPrime (#8911)

zpcore committed 249 days ago

Verified 45830519

Make scan-based GRU support `batch_first` parameter. (#8964)

iwknow committed 250 days ago

Verified aad87e61

[call_jax] support returning PyTree from the JAX function (#8957)

tengyifei committed 250 days ago

Verified 297f5de6

Remove deprecated typing module in ansible flow (#8955)

tengyifei committed 250 days ago

Verified e77bd2f6

Add test to ensure scan-based and the standard GRU are interchangeable. (#8949)

iwknow committed 251 days ago

Verified d4d997dc

Update to pytorch 2.6 (#8944)

qihqi committed 252 days ago

Verified 085f8bc1

Pin update to 20250406 (#8945)

tengyifei committed 252 days ago

Verified 0837fac3

Use pages_per_seq * page_size instead of directly passing max_model_len (#8950)

yaochengji committed 253 days ago

Verified 9ec626e6

Fix ragged_paged_attention op signature (#8943)

yaochengji committed 253 days ago

Verified 80929a91

add rc4 trigger (#8941)

zpcore committed 255 days ago

Verified ac9a39f4

Add block size table for ragged_paged_attention (#8942)

yaochengji committed 255 days ago

Verified cbff76f5

Add alternative dynamo backend (#8893)

qihqi committed 256 days ago

Verified fb6038de

Trigger build for 2.6.1 patch (#8938)

zpcore committed 256 days ago

Verified 379ebd5f

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

tengyifei committed 256 days ago

Verified 9e23ca85

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

iwknow committed 256 days ago

Verified 145c33ae

Fix up pin update (#8935)

tengyifei committed 256 days ago

Verified 980ead54

Newer Older