biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[CI Perf] Prune tests in tests/kernels/attention/ (#22936)

Browse Source 
Signed-off-by: mgoin <mgoin64@gmail.com>
This commit is contained in:
Michael Goin
2025-08-14 23:34:53 -04:00
committed by GitHub
parent d2b0e97ea6
commit 0fe85087a9
8 changed files with 39 additions and 38 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
tests/kernels/attention/test_cache.py
View File

@@ -11,11 +11,11 @@ from vllm import _custom_ops as ops
from vllm.platforms import current_platform
COPYING_DIRECTION = [('cuda', 'cpu'), ('cuda', 'cuda'), ('cpu', 'cuda')]
DTYPES = [torch.half, torch.bfloat16, torch.float]
DTYPES = [torch.bfloat16, torch.float]
NUM_TOKENS = [42] # Arbitrary values for testing
NUM_LAYERS = [1] # Arbitrary values for testing
NUM_HEADS = [8] # Arbitrary values for testing
HEAD_SIZES = [64, 80, 120, 256]
HEAD_SIZES = [64, 80, 256]
BLOCK_SIZES = [8, 16, 32]
CACHE_LAYOUTS = ["NHD", "HND"]