Skip to content

[main][bugfix] Unify MoE routing init with standard torch_npu operator #2401

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

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
+13 −39
Open

[main][bugfix] Unify MoE routing init with standard torch_npu operator #2401

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
52 changes: 13 additions & 39 deletions vllm_ascend/quantization/w8a8_dynamic.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -365,29 +365,6 @@ def fused_experts_with_mc2(
return hidden_states, shared_output


def init_routing_quant(hidden_states, top_k, topk_ids, global_num_experts):
num_tokens, _ = hidden_states.shape
row_idx_len = num_tokens * top_k
row_idx = (torch.arange(0,
row_idx_len,
dtype=torch.int32,
device=hidden_states.device).view(
top_k, -1).permute(1, 0).contiguous())
hidden_states, expanded_row_idx, expanded_expert_idx = torch_npu.npu_moe_init_routing(
hidden_states,
row_idx=row_idx,
expert_idx=topk_ids,
active_num=num_tokens)

expanded_row_idx = (expanded_row_idx.view(top_k, -1).permute(
1, 0).contiguous().view(-1))
global_expert_tokens = torch.bincount(expanded_expert_idx,
minlength=global_num_experts)
global_expert_tokens = global_expert_tokens.to(torch.int32)
quantized_tokens, token_scales = torch_npu.npu_dynamic_quant(hidden_states)
return quantized_tokens, expanded_row_idx, global_expert_tokens, token_scales


# currently expert parallelism implemented with all2all
# is under-optimized.
def fused_experts_with_all2all(
Expand Down Expand Up @@ -417,22 +394,19 @@ def fused_experts_with_all2all(

if expert_map is not None:
global_num_experts = len(expert_map) + global_redundant_expert_num
if hasattr(torch_npu, "npu_moe_init_routing_quant"):
quantized_tokens, expanded_row_idx, global_expert_tokens, _, token_scales = torch_npu.npu_moe_init_routing_quant(
hidden_states,
expert_idx=topk_ids.to(torch.int32),
active_num=0,
expert_capacity=0,
expert_num=global_num_experts,
drop_pad_mode=0,
expert_tokens_num_mode=2,
expert_tokens_before_capacity_flag=False,
quant_mode=1,
)
else:
quantized_tokens, expanded_row_idx, global_expert_tokens, token_scales = init_routing_quant(
hidden_states, top_k, topk_ids, global_num_experts)

active_num = top_k * num_tokens
active_expert_range = [0, global_num_experts]
quantized_tokens, expanded_row_idx, global_expert_tokens, token_scales = torch_npu.npu_moe_init_routing_v2(
hidden_states,
expert_idx=topk_ids.to(torch.int32),
active_num=active_num,
expert_capacity=0,
expert_num=global_num_experts,
drop_pad_mode=0,
expert_tokens_num_type=1,
expert_tokens_num_flag=True,
quant_mode=1,
active_expert_range=active_expert_range)
gather_sizes = global_expert_tokens.new_empty(
global_expert_tokens.shape[0])
dist.all_to_all_single(gather_sizes, global_expert_tokens)
Expand Down
Loading