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

[Kernels] Clean up FusedMoeMethodBase and modular kernel setup. Remove extra arguments from modular kernel methods. (#22035)

Browse Source 
Signed-off-by: Bill Nell <bnell@redhat.com>
Co-authored-by: Michael Goin <mgoin64@gmail.com>
This commit is contained in:
bnellnm
2025-08-15 14:46:00 -04:00
committed by GitHub
parent 48b01fd4d4
commit 8ad7285ea2
54 changed files with 2010 additions and 1293 deletions
Download Patch File Download Diff File Expand all files Collapse all files

31
tests/kernels/moe/test_block_fp8.py
View File

@@ -161,18 +161,20 @@ def test_w8a8_block_fp8_fused_moe(M, N, K, E, topk, block_size, dtype, seed,
a = torch.randn((M, K), dtype=dtype) / 10
score = torch.randn((M, E), dtype=dtype)
_, w1, w1_s, _, w2, w2_s = make_test_weights(E,
N,
K,
dtype,
torch.float8_e4m3fn,
per_act_token_quant=False,
block_shape=block_size)
(_, w1, w1_s, _), (_, w2, w2_s,
_) = make_test_weights(E,
N,
K,
dtype,
torch.float8_e4m3fn,
per_act_token_quant=False,
block_shape=block_size)
m_fused_moe = modular_triton_fused_moe(use_fp8_w8a8=True,
use_int8_w8a8=False,
use_int8_w8a16=False,
use_int4_w4a16=False,
use_mxfp4_w4a4=False,
per_act_token_quant=False,
block_shape=block_size)
@@ -247,13 +249,14 @@ def test_w8a8_block_fp8_deep_gemm_fused_moe(M, N, K, E, topk, seed,
a = torch.randn((M, K), dtype=dtype) / 10
score = torch.randn((M, E), dtype=dtype)
_, w1, w1_s, _, w2, w2_s = make_test_weights(E,
N,
K,
dtype,
torch.float8_e4m3fn,
per_act_token_quant=False,
block_shape=block_size)
(_, w1, w1_s, _), (_, w2, w2_s,
_) = make_test_weights(E,
N,
K,
dtype,
torch.float8_e4m3fn,
per_act_token_quant=False,
block_shape=block_size)
# Note: for now use_compile will error out if the problem size is
# large enough to trigger chunking. I'm leaving the flag and