A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 15, 21:00 UTC
Fix test_block_fp8.py test for MoE (#13915)
Signed-off-by: mgoin <mgoin64@gmail.com>
Michael Goin committed
Feb 27, 2025 at 05:00 UTC
788f284b5361d5192e6fc22856ea97e0e579530c
1 file changed
+2
-2
tests/kernels/test_block_fp8.py
+2
-2
@@ -30,8 +30,8 @@ M_moe = [1, 7, 83, 512, 2048]
30
N_moe = [4608] # [128, 4608, 13824]
31
K_moe = [7168] # [256, 7168, 13824]
32
BLOCK_SIZE = [[128, 128]]
33
-E = [256] # [8, 24, 128, 256]
34
-TOP_KS = [1] # [1, 2, 6]
33
+E = [8, 24] # [8, 24, 128, 256]
34
+TOP_KS = [2] # [1, 2, 6]
35
OUT_DTYPES = [torch.bfloat16] # [torch.float32, torch.half, torch.bfloat16]
36
SEEDS = [0]
37