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

[CI Perf] Prune tests in tests/kernels/quantization/ (#22942)

Browse Source 
Signed-off-by: mgoin <mgoin64@gmail.com>
This commit is contained in:
Michael Goin
2025-08-14 23:25:34 -04:00
committed by GitHub
parent ae05a6d83d
commit 590bddbfc5
6 changed files with 66 additions and 33 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
tests/kernels/quantization/test_int8_quant.py
View File

@@ -9,10 +9,9 @@ from tests.kernels.utils import opcheck
from vllm._custom_ops import scaled_int8_quant
from vllm.platforms import current_platform
DTYPES = [torch.half, torch.bfloat16, torch.float]
HIDDEN_SIZES = [16, 67, 768, 5137, 8193] # Arbitrary values for testing
HIDDEN_SIZES += list(range(1024, 1033)) # vectorized conversion edge cases
NUM_TOKENS = [1, 7, 83, 4096] # Arbitrary values for testing
DTYPES = [torch.bfloat16, torch.float]
HIDDEN_SIZES = [17, 1024, 1025, 1026, 5137, 8193]
NUM_TOKENS = [1, 7, 4096]
SEEDS = [0]
SCALE = [0.1, 2.1]