f4e2264a83
Copied from upstream_ref (NOT rewritten — exact upstream code):
ixformer C++ API (the authoritative header):
include/ixformer.h — ixformer::infer namespace: topk_softmax,
moe_compute_token_index_api, moe_w16a16_group_gemm, moe_expand_input,
moe_output_reduce_sum, silu_and_mul, rms_norm, xllm_paged_attention, etc.
include/ilu_ops_api.h — xllm::kernel::ilu namespace: moe_active_topk,
moe_gen_idx, moe_expand_input, group_gemm, moe_combine_result,
batch_prefill, batch_decode, rms_norm, matmul, act_and_mul, etc.
ILU kernel wrappers (call ixformer::infer directly):
csrc/ilu_kernel_fused_moe.cpp — topk routing + gen_idx + expand + combine
csrc/ilu_kernel_group_gemm.cpp — batched expert GEMM
csrc/ilu_kernel_{activation,norm,rope,matmul,attention}.cpp
ILU layer implementations (full pipeline):
csrc/ilu_layer_fused_moe.{cpp,h} — 797 lines, the complete MoE pipeline
that competitor 168 ran as corex_moe.py
csrc/ilu_layer_attention.{cpp,h} — prefill/decode attention dispatch
CUDA MoE kernels (from xllm + ds_vllm):
csrc/moe/moe_topk_softmax_kernels.cuh — CUB BlockReduce + warp topk
csrc/moe/moe_topk_sigmoid_kernels.cuh — sigmoid scoring variant
csrc/moe/moe_topk.cuh + moe_fused_topk.cu — entry points
csrc/moe/moeTopKFuncs.cuh — TRT-LLM derived vllm-compatible topk
csrc/moe/moe_ops.h + moe_align_sum_kernels.cu — alignment kernels
Common layer headers:
csrc/common_fused_moe{,_base}.h + common_moe_fused_topk.{cpp,h}
54 lines
1.5 KiB
C++
54 lines
1.5 KiB
C++
/* Copyright 2026 The xLLM Authors. All Rights Reserved.
|
|
|
|
Licensed under the Apache License, Version 2.0 (the "License");
|
|
you may not use this file except in compliance with the License.
|
|
You may obtain a copy of the License at
|
|
|
|
https://github.com/jd-opensource/xllm/blob/main/LICENSE
|
|
|
|
Unless required by applicable law or agreed to in writing, software
|
|
distributed under the License is distributed on an "AS IS" BASIS,
|
|
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
See the License for the specific language governing permissions and
|
|
limitations under the License.
|
|
==============================================================================*/
|
|
|
|
#pragma once
|
|
|
|
#include <torch/torch.h>
|
|
|
|
#include "framework/model/model_args.h"
|
|
#include "framework/quant_args.h"
|
|
#include "framework/state_dict/state_dict.h"
|
|
#include "framework/state_dict/utils.h"
|
|
|
|
namespace xllm {
|
|
namespace layer {
|
|
|
|
class MoEFusedTopkImpl : public torch::nn::Module {
|
|
public:
|
|
MoEFusedTopkImpl(const ModelArgs& model_args,
|
|
const QuantArgs& quant_args,
|
|
const torch::TensorOptions& options);
|
|
|
|
std::tuple<torch::Tensor, torch::Tensor> forward(
|
|
torch::Tensor& router_logits);
|
|
|
|
void load_state_dict(const StateDict& state_dict);
|
|
|
|
private:
|
|
int64_t topk_;
|
|
int64_t num_expert_group_;
|
|
int64_t topk_group_;
|
|
double route_scale_;
|
|
int64_t hidden_size_;
|
|
bool renormalize_;
|
|
std::string scoring_func_;
|
|
|
|
DEFINE_WEIGHT(e_score_correction_bias);
|
|
};
|
|
|
|
TORCH_MODULE(MoEFusedTopk);
|
|
} // namespace layer
|
|
} // namespace xllm
|