Skip to content

Guard fused Marlin MoE atomic add usage to prevent crashes #13262

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

Closed
0xymoro wants to merge 1 commit into from
Closed

Guard fused Marlin MoE atomic add usage to prevent crashes #13262

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
23 changes: 17 additions & 6 deletions sgl-kernel/python/sgl_kernel/fused_moe.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -71,6 +71,9 @@ def fused_marlin_moe(
moe_align_block_size,
try_get_optimal_moe_config,
)
from sglang.srt.layers.quantization.marlin_utils import (
should_use_atomic_add_reduce,
)

# Check constraints.
assert hidden_states.shape[0] == gating_output.shape[0], "Number of tokens mismatch"
Expand Down Expand Up @@ -142,10 +145,16 @@ def fused_marlin_moe(
intermediate_cache3 = intermediate_cache13[: M * topk_ids.shape[1] * K]
intermediate_cache3 = intermediate_cache3.view(-1, K)

use_atomic_add = (
hidden_states.dtype == torch.half
or torch.cuda.get_device_capability(hidden_states.device)[0] >= 9
)
def _should_use_atomic_add(m: int, n: int, k: int) -> bool:
return should_use_atomic_add_reduce(
m=m,
n=n,
k=k,
device=hidden_states.device,
dtype=hidden_states.dtype,
)

use_atomic_add_first = _should_use_atomic_add(M, 2 * N, K)

intermediate_cache1 = torch.ops.sgl_kernel.moe_wna16_marlin_gemm.default(
hidden_states,
Expand All @@ -169,7 +178,7 @@ def fused_marlin_moe(
size_n=2 * N,
size_k=K,
is_k_full=is_k_full,
use_atomic_add=use_atomic_add,
use_atomic_add=use_atomic_add_first,
use_fp32_reduce=True,
is_zp_float=False,
)
Expand All @@ -179,6 +188,8 @@ def fused_marlin_moe(
if expert_map is not None:
intermediate_cache3.zero_()

use_atomic_add_second = _should_use_atomic_add(M * topk, K, N)

intermediate_cache3 = torch.ops.sgl_kernel.moe_wna16_marlin_gemm.default(
intermediate_cache2,
intermediate_cache3,
Expand All @@ -201,7 +212,7 @@ def fused_marlin_moe(
size_n=K,
size_k=N,
is_k_full=is_k_full,
use_atomic_add=use_atomic_add,
use_atomic_add=use_atomic_add_second,
use_fp32_reduce=True,
is_zp_float=False,
).view(-1, topk, K)
Expand Down
Loading