Skip to content

fix: restore FP4 deepgemm path for Blackwell broken by #23686 #23948

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

Merged
Fridge003 merged 2 commits into from
Apr 28, 2026
Merged

fix: restore FP4 deepgemm path for Blackwell broken by #23686 #23948

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
018b03b
fix: restore FP4 deepgemm path for Blackwell broken by #23686
Apr 28, 2026
d3e63aa
refactor: address review feedback - simplify marlin branch and add ba…
Apr 28, 2026
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
77 changes: 35 additions & 42 deletions python/sglang/srt/layers/quantization/fp8.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -932,52 +932,45 @@ def process_weights_after_loading_block_quant(self, layer: Module) -> None:
will_use_deepgemm = self.is_deepgemm_moe_runner_backend_enabled()

if self.is_fp4_expert:
if get_moe_runner_backend().is_marlin():
layer.w13_weight.data = layer.w13_weight.data.view(torch.int8)
layer.w2_weight.data = layer.w2_weight.data.view(torch.int8)
elif not get_moe_runner_backend().is_flashinfer_mxfp4():
raise NotImplementedError(
"DeepSeekV4 FP4 experts now require a native FP4 MoE backend. "
"Use `--moe-runner-backend marlin` on Hopper or "
"`--moe-runner-backend flashinfer_mxfp4` when available."
# FP4 experts support three MoE backends:
# - marlin (Hopper w4a16): only needs int8 view
# - flashinfer_mxfp4: only needs int8 view
# - deepgemm/auto (Blackwell): int8 view + mega_moe or scale conversion
layer.w13_weight.data = layer.w13_weight.data.view(torch.int8)
layer.w2_weight.data = layer.w2_weight.data.view(torch.int8)

if envs.SGLANG_OPT_USE_DEEPGEMM_MEGA_MOE.get():
from sglang.srt.models.deepseek_v4 import (
build_mega_moe_experts_weights,
)

else:
layer.w13_weight.data = layer.w13_weight.data.view(torch.int8)
layer.w2_weight.data = layer.w2_weight.data.view(torch.int8)
build_mega_moe_experts_weights(layer)
return

if envs.SGLANG_OPT_USE_DEEPGEMM_MEGA_MOE.get():
from sglang.srt.models.deepseek_v4 import (
build_mega_moe_experts_weights,
if (
envs.SGLANG_OPT_DEEPGEMM_SCALE_CONVERT_AT_INIT.get()
and envs.SGLANG_DSV4_MODE.get() == "2604"
and deep_gemm_wrapper.DEEPGEMM_SCALE_UE8M0
and will_use_deepgemm
):
from deep_gemm import transform_sf_into_required_layout

for scale_param, weight_param in [
(layer.w13_weight_scale_inv, layer.w13_weight),
(layer.w2_weight_scale_inv, layer.w2_weight),
]:
num_experts, n, _ = scale_param.data.shape
k = weight_param.shape[2] * 2
scale_param.data = transform_sf_into_required_layout(
scale_param.data,
mn=n,
k=k,
recipe=(1, 32),
num_groups=num_experts,
disable_ue8m0_cast=False,
)

build_mega_moe_experts_weights(layer)
return

if (
envs.SGLANG_OPT_DEEPGEMM_SCALE_CONVERT_AT_INIT.get()
and envs.SGLANG_DSV4_MODE.get() == "2604"
and deep_gemm_wrapper.DEEPGEMM_SCALE_UE8M0
and will_use_deepgemm
):
from deep_gemm import transform_sf_into_required_layout

for scale_param, weight_param in [
(layer.w13_weight_scale_inv, layer.w13_weight),
(layer.w2_weight_scale_inv, layer.w2_weight),
]:
num_experts, n, _ = scale_param.data.shape
k = weight_param.shape[2] * 2
scale_param.data = transform_sf_into_required_layout(
scale_param.data,
mn=n,
k=k,
recipe=(1, 32),
num_groups=num_experts,
disable_ue8m0_cast=False,
)
layer.w13_weight_scale_inv.format_ue8m0 = True
layer.w2_weight_scale_inv.format_ue8m0 = True
layer.w13_weight_scale_inv.format_ue8m0 = True
layer.w2_weight_scale_inv.format_ue8m0 = True

if (
not self.is_fp4_expert
Expand Down
Loading