Skip to content

test: Mark test_fp8_prefill.py as xfail on SM90 #2038

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
yzh119 merged 2 commits into from
Nov 4, 2025
Merged

test: Mark test_fp8_prefill.py as xfail on SM90 #2038

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
9aedff3
xfail fp8 prefill test on SM90
bkryu Nov 4, 2025
d5cea24
Remove xfails, explicitly specify fa2
bkryu Nov 4, 2025
File filter

Filter by extension

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: 4 additions & 4 deletions tests/attention/test_fp8_prefill.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -66,7 +66,7 @@ def test_batch_prefill_with_paged_kv_cache_fp8_calibration_scale(

workspace_buffer = torch.empty(32 * 1024 * 1024, dtype=torch.int8).to(0)
wrapper_f16 = flashinfer.BatchPrefillWithPagedKVCacheWrapper(
workspace_buffer, kv_layout
workspace_buffer, kv_layout, backend="fa2"
)
wrapper_f16.plan(
qo_indptr,
Expand All @@ -90,7 +90,7 @@ def test_batch_prefill_with_paged_kv_cache_fp8_calibration_scale(
kv_data_fp8 = torch.cat([k_fp8, v_fp8], dim=1)

wrapper_f8 = flashinfer.BatchPrefillWithPagedKVCacheWrapper(
workspace_buffer, kv_layout
workspace_buffer, kv_layout, backend="fa2"
)
wrapper_f8.plan(
qo_indptr,
Expand Down Expand Up @@ -156,7 +156,7 @@ def test_batch_decode_with_prefill_with_paged_kv_cache(

workspace_buffer = torch.empty(32 * 1024 * 1024, dtype=torch.int8).to(0)
wrapper = flashinfer.BatchPrefillWithPagedKVCacheWrapper(
workspace_buffer, kv_layout
workspace_buffer, kv_layout, backend="fa2"
)
wrapper.plan(
qo_indptr,
Expand All @@ -173,7 +173,7 @@ def test_batch_decode_with_prefill_with_paged_kv_cache(
o_fp8 = wrapper.run(q, kv_data)

decode_wrapper = flashinfer.BatchDecodeWithPagedKVCacheWrapper(
workspace_buffer, kv_layout
workspace_buffer, kv_layout, backend="fa2"
)
decode_wrapper.plan(
kv_indptr,
Expand Down