Skip to content

[NPU] bug fix for mtp and w4a8 #14806

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
iforgetmyname merged 2 commits into from
Dec 10, 2025
Merged

[NPU] bug fix for mtp and w4a8 #14806

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
ddfda45
[NPU] bug fix
liupeng374 Dec 10, 2025
168f952
Merge branch 'main' into mtp_bugfix
iforgetmyname Dec 10, 2025
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
2 changes: 2 additions & 0 deletions python/sglang/srt/hardware_backend/npu/attention/ascend_backend.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -297,6 +297,8 @@ def init_forward_metadata(self, forward_batch: ForwardBatch):
if (
self.use_mla
and forward_batch.forward_mode.is_extend()
and not forward_batch.forward_mode.is_draft_extend(include_v2=True)
and not forward_batch.forward_mode.is_target_verify()
and sum(forward_batch.extend_prefix_lens_cpu) > 0
):
self.forward_metadata.prefix_lens = forward_batch.extend_prefix_lens.to(
Expand Down
20 changes: 11 additions & 9 deletions python/sglang/srt/hardware_backend/npu/quantization/fused_moe_method_npu.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -567,11 +567,13 @@ def apply_without_routing_weights(
group_list,
output_dtype,
):
from sgl_kernel_npu.activation.swiglu_quant import swiglu_quant
Copy link
Copy Markdown
Contributor

@gemini-code-assist gemini-code-assist bot Dec 10, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

For better code organization and to avoid potential performance overhead from repeated imports, it's recommended to move this import to the top of the file. This makes dependencies clearer and follows PEP 8 guidelines.

You can add the following at the top of the file:

from sglang.srt.utils import is_npu

if is_npu():
    from sgl_kernel_npu.activation.swiglu_quant import swiglu_quant


hidden_states = torch.ops.npu.npu_grouped_matmul(
x=[hidden_states],
weight=[self.w13_weight],
scale=[self.w13_weight_scale],
bias=[self.w13_scale_bias],
weight=[layer.w13_weight],
scale=[layer.w13_weight_scale],
bias=[layer.w13_scale_bias],
per_token_scale=[hidden_states_scale],
group_list=group_list,
split_item=2,
Expand All @@ -580,15 +582,15 @@ def apply_without_routing_weights(
output_dtype=output_dtype,
)[0]

# act_fn: swiglu
hidden_states = torch.ops.npu.npu_swiglu(hidden_states)
hidden_states, swiglu_out_scale = torch.ops.npu.npu_dynamic_quant(hidden_states)
hidden_states, swiglu_out_scale = swiglu_quant(
hidden_states, group_list, group_list_type
)

hidden_states = torch.ops.npu.npu_grouped_matmul(
x=[hidden_states],
weight=[self.w2_weight],
scale=[self.w2_weight_scale],
bias=[self.w2_scale_bias],
weight=[layer.w2_weight],
scale=[layer.w2_weight_scale],
bias=[layer.w2_scale_bias],
per_token_scale=[swiglu_out_scale],
group_list=group_list,
split_item=2,
Expand Down
7 changes: 6 additions & 1 deletion python/sglang/srt/layers/moe/ep_moe/layer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,6 +6,9 @@
import torch

from sglang.srt.environ import envs
from sglang.srt.hardware_backend.npu.quantization.fused_moe_method_npu import (
NPUW4A16Int4DynamicMoEMethod,
)
from sglang.srt.layers import deep_gemm_wrapper
from sglang.srt.layers.moe import (
get_deepep_mode,
Expand Down Expand Up @@ -347,7 +350,9 @@ def forward_npu(
)
else:
input_quant = get_bool_env_var("DEEP_NORMAL_MODE_USE_INT8_QUANT")
if not input_quant and self.w13_weight.dtype != torch.int32:
if not input_quant and not isinstance(
self.quant_method, NPUW4A16Int4DynamicMoEMethod
):
hidden_states, hidden_states_scale = torch_npu.npu_dynamic_quant(
hidden_states
)
Expand Down
Loading