Skip to content

Fix up pin update #8935

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Apr 4, 2025
Merged

Fix up pin update #8935

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 8 additions & 0 deletions test/test_pallas.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -563,6 +563,8 @@ def test_flash_attention_backward(self):
for i in [(q, q_grad), (k, k_grad), (v, v_grad)]:
self.assertTrue(torch.allclose(i[0].grad.cpu(), i[1].cpu(), atol=1e-05))

@unittest.skipIf(xr.device_type() != 'TPU' or tpu.version() < 4,
"This test only works on TPUv4+.")
def test_paged_attention_wrapper(self):
from torch_xla.experimental.custom_kernel import paged_attention
from jax.experimental.pallas.ops.tpu.paged_attention.paged_attention_kernel import paged_attention as jax_paged_attention
Expand Down Expand Up @@ -1088,6 +1090,8 @@ def multi_queries_paged_attention_wrapper(q, k_pages, v_pages, kv_seq_lens,
torch.allclose(
output.cpu(), nonkernel_output.cpu(), atol=1e-2, rtol=1e-2))

@unittest.skipIf(xr.device_type() != 'TPU' or tpu.version() != 4,
"This test only works on TPUv4 and TPUv5p.")
def test_paged_attention_wrapper_with_megacore_modes(self):
# TODO: enable checking TPU accelerator types.
from torch_xla.experimental.custom_kernel import paged_attention
Expand Down Expand Up @@ -1156,6 +1160,8 @@ def test_paged_attention_wrapper_with_megacore_modes(self):
atol=1e-5,
rtol=1e-5))

@unittest.skipIf(xr.device_type() != 'TPU' or tpu.version() < 4,
"This test only works on TPUv4+.")
def test_paged_attention_wrapper_with_dynamo(self):
from jax.experimental.pallas.ops.tpu.paged_attention.paged_attention_kernel import paged_attention as jax_paged_attention

Expand Down Expand Up @@ -1232,6 +1238,8 @@ def paged_attention_wrapper(q, k, v, seq_lens, page_indices,
atol=1e-5,
rtol=1e-5))

@unittest.skipIf(xr.device_type() != 'TPU' or tpu.version() < 4,
"This test only works on TPUv4+.")
def test_paged_attention_wrapper_with_attn_logits_soft_cap(self):
# TODO: enable checking TPU accelerator types.
from torch_xla.experimental.custom_kernel import paged_attention
Expand Down
Loading