pytorch/xla

Pull Requests Commits

Addressing comments from Mike

yaoshiang committed 199 days ago

3da81651

yaoshiang committed 200 days ago

d89f44cf

yaoshiang committed 200 days ago

18ce9114

Guide to debugging in PyTorch

yaoshiang committed 200 days ago

6f95d544

Set scoped vmem for paged attention (#8988)

zpcore committed 200 days ago

Verified 366f248a

Fix merge conflict (#8991)

zpcore committed 200 days ago

Verified 083934fd

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

qihqi committed 201 days ago

Verified cce7cc07

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

qihqi committed 201 days ago

Verified 78824755

Fix splash attention test (#8978)

zpcore committed 201 days ago

Verified 13155c9c

Fix call_jax hashing (#8981)

zpcore committed 201 days ago

Verified a48d6a97

Update to ubuntu latest (#8979)

qihqi committed 201 days ago

Verified a37a5852

@assume_pure (#8962)

tengyifei committed 201 days ago

Verified cf764ffa

Avoid re-computing computation hashes (#8976)

rpsilva-aws committed 202 days ago

Verified bcca5eba

Disable one splash attention test (#8970)

zpcore committed 202 days ago

Verified ba5cd4cb

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

yarongmu-google committed 205 days ago

Verified 4eea5e13

Adapt Splash Attention from TorchPrime (#8911)

zpcore committed 206 days ago

Verified 45830519

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

iwknow committed 206 days ago

Verified aad87e61

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

tengyifei committed 207 days ago

Verified 297f5de6

Remove deprecated typing module in ansible flow (#8955)

tengyifei committed 207 days ago

Verified e77bd2f6

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

iwknow committed 208 days ago

Verified d4d997dc

Update to pytorch 2.6 (#8944)

qihqi committed 208 days ago

Verified 085f8bc1

Pin update to 20250406 (#8945)

tengyifei committed 209 days ago

Verified 0837fac3

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

yaochengji committed 209 days ago

Verified 9ec626e6

Fix ragged_paged_attention op signature (#8943)

yaochengji committed 210 days ago

Verified 80929a91

add rc4 trigger (#8941)

zpcore committed 212 days ago

Verified ac9a39f4

Add block size table for ragged_paged_attention (#8942)

yaochengji committed 212 days ago

Verified cbff76f5

Add alternative dynamo backend (#8893)

qihqi committed 212 days ago

Verified fb6038de

Trigger build for 2.6.1 patch (#8938)

zpcore committed 213 days ago

Verified 379ebd5f

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

tengyifei committed 213 days ago

Verified 9e23ca85

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

iwknow committed 213 days ago

Verified 145c33ae

Older