dylanyunlong/project_6
1
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

perf: MoE decode — pre-transposed bmm replaces F.linear (6.9ms vs 8.0ms, 14%)

Browse Source 
Probe data (probe_moe_fused_breakdown.sh on BI-V100):
  F.linear loop 8 experts:     8.060 ms
  bmm pre-transposed full MoE: 6.918 ms  ← 14% faster
  transpose+contiguous runtime: 22.219 ms ← why CUTLASS was 27ms

Changes:
  - Lazy-cache w13_t (E,H,2I) and w2_t (E,I,H) on first decode call
  - FC1: torch.bmm(x_expand, w13_t_sel) replaces F.linear(x, w13_sel.reshape)
  - FC2: torch.bmm(act, w2_t_sel) replaces torch.bmm(w2_sel, act^T)
  - Zero runtime transpose cost after first call
This commit is contained in:
dylan
2026-08-15 12:52:55 +00:00
parent 9f265894cc
commit 6f1904aa8c
1 changed files with 20 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

27
qwen3_6_scripts/qwen3_5.py
View File

@@ -1634,6 +1634,10 @@ class Qwen3_5MoeSparseBlock(nn.Module):
quant_config=quant_config)
self.act_fn = SiluAndMul()
# Pre-transposed weight cache for bmm decode path
self._w13_t = None
self._w2_t = None
def _router_shared_gate_weight_loader(
self,
param: torch.Tensor,
@@ -1770,19 +1774,28 @@ class Qwen3_5MoeSparseBlock(nn.Module):
H = hidden_states.shape[-1]
gate_up = F.linear(
hidden_states,
w13_sel.reshape(-1, H), # (K*2*I, H) — contiguous after indexing
) # (1, K*2*I)
gate_up = gate_up.view(self.top_k, -1) # (K, 2*I)
# --- Pre-transpose weights for bmm (cached after first call) ---
if not hasattr(self, '_w13_t') or self._w13_t is None:
# (E, 2*I, H) → (E, H, 2*I) — one-time cost at first decode
self._w13_t = self.experts.w13_weight.transpose(1, 2).contiguous()
self._w2_t = self.experts.w2_weight.transpose(1, 2).contiguous()
# (E, H, I) → (E, I, H)
w13_t_sel = self._w13_t[eids] # (K, H, 2*I)
w2_t_sel = self._w2_t[eids] # (K, I, H)
# FC1: bmm (K,1,H) @ (K,H,2I) → (K,1,2I)
x_expand = hidden_states.unsqueeze(0).expand(self.top_k, -1, -1) # (K, 1, H)
gate_up = torch.bmm(x_expand, w13_t_sel).squeeze(1) # (K, 2*I)
if _USE_FUSED_MOE_ACTIVATION:
act = self.act_fn(gate_up) # (K, I)
else:
gate, up = gate_up.chunk(2, dim=-1)
act = F.silu(gate) * up
# bmm: (K,H,I) @ (K,I,1) → (K,H,1) → (K,H)
expert_out = torch.bmm(w2_sel, act.unsqueeze(-1)).squeeze(-1) # (K, H)
# FC2: bmm (K,1,I) @ (K,I,H) → (K,1,H)
expert_out = torch.bmm(act.unsqueeze(1), w2_t_sel).squeeze(1) # (K, H)
if (_USE_COREX_MOE_EXACT_REDUCE
and expert_out.dtype == torch.float16