Skip to content

[Refactor] Pre-transpose MoE weights for improved performance #2025

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
wants to merge 1 commit into from
Closed

[Refactor] Pre-transpose MoE weights for improved performance #2025

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
22 changes: 7 additions & 15 deletions vllm_ascend/ops/fused_moe.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -172,8 +172,6 @@ def fused_experts_with_mc2(
npu_wait_tensor(shared_gate_up, expand_x)
shared_act = shared_experts.act_fn(shared_gate_up)

w1 = w1.transpose(1, 2)

group_list = expert_token_nums.to(torch.int64)
gate_up_out_list = torch_npu.npu_grouped_matmul(
x=[expand_x],
Expand All @@ -189,7 +187,6 @@ def fused_experts_with_mc2(
gate_up_out = torch.cat(gate_up_out_list, dim=0)
gate_up_out = torch_npu.npu_swiglu(gate_up_out)

w2 = w2.transpose(1, 2)
down_out_list = torch_npu.npu_grouped_matmul(
x=[gate_up_out],
weight=[w2],
Expand Down Expand Up @@ -266,7 +263,6 @@ def apply_mlp(hidden_states_wrapper: List[torch.Tensor],
assert len(hidden_states_wrapper) == 1
hidden_states = hidden_states_wrapper.pop()

w1 = w1.transpose(1, 2)
hidden_states = torch_npu.npu_grouped_matmul(
x=[hidden_states],
weight=[w1],
Expand Down Expand Up @@ -369,7 +365,6 @@ def fused_experts_with_all2all(
expanded_expert_idx, num_experts)
expert_tokens = expert_tokens.to(torch.int64)

w1 = w1.transpose(1, 2)
gate_up_out_list = torch_npu.npu_grouped_matmul(
x=[hidden_states],
weight=[w1],
Expand Down Expand Up @@ -611,7 +606,6 @@ def fused_experts_moge(
0, sorted_topk_ids).unsqueeze(-1)
group_list = num_tokens_per_expert.cumsum(dim=0).to(torch.int64)

w1 = w1.transpose(1, 2)
gate_up_out = torch_npu.npu_grouped_matmul(
x=[sorted_hidden_states],
weight=[w1],
Expand All @@ -628,7 +622,6 @@ def fused_experts_moge(
gate_up_out = torch_npu.npu_swiglu(gate_up_out)
gate_up_out *= topk_scales

w2 = w2.transpose(1, 2)
down_out_list = torch_npu.npu_grouped_matmul(
x=[gate_up_out],
weight=[w2],
Expand Down Expand Up @@ -760,7 +753,6 @@ def fused_experts(
expanded_expert_idx, num_experts)
expert_tokens = expert_tokens.to(torch.int64)

w1 = w1.transpose(1, 2)
gate_up_out_list = torch_npu.npu_grouped_matmul(
x=[sorted_hidden_states],
weight=[w1],
Expand All @@ -774,7 +766,6 @@ def fused_experts(
gate_up_out = torch.cat(gate_up_out_list, dim=0)
gate_up_out = torch_npu.npu_swiglu(gate_up_out)

w2 = w2.transpose(1, 2)
down_out_list = torch_npu.npu_grouped_matmul(
x=[gate_up_out],
weight=[w2],
Expand Down Expand Up @@ -1003,12 +994,13 @@ def __init__(self, moe: FusedMoEConfig = None):
def process_weights_after_loading(self, layer):
super(UnquantizedFusedMoEMethod,
self).process_weights_after_loading(layer)
layer.w13_weight = torch.nn.Parameter(self._maybe_pad_weight(
layer.w13_weight.data),
requires_grad=False)
layer.w2_weight = torch.nn.Parameter(self._maybe_pad_weight(
layer.w2_weight.data),
requires_grad=False)
w13_data = self._maybe_pad_weight(layer.w13_weight.data).transpose(
1, 2).contiguous()
layer.w13_weight = torch.nn.Parameter(w13_data, requires_grad=False)

w2_data = self._maybe_pad_weight(layer.w2_weight.data).transpose(
1, 2).contiguous()
layer.w2_weight = torch.nn.Parameter(w2_data, requires_grad=False)

def apply(
self,
Expand Down
Loading