6415249693
MoE call chain from ds_vllm (vllm-project/vllm latest): ex_engine/moe/ — 20 files, 8736 lines - modular_kernel.py (1630 lines) — base classes for modular MoE - experts/fused_batched_moe.py (972 lines) — NaiveBatchedExperts - prepare_finalize/batched.py (171 lines) — token grouping by expert - topk_weight_and_reduce.py (176 lines) — scatter-add finalize - fused_moe.py (1740 lines) — main fused_moe dispatch - config.py (1407 lines) — FusedMoEQuantConfig - activation.py, utils.py, layer.py, etc. xllm layer code (jd-opensource/xllm): ex_engine/xllm_layers/ — 39 files, 5859 lines - ilu/fused_moe.cpp (797 lines) — production ixformer 7-step MoE pipeline - ilu/attention.cpp (189 lines) — paged_attention + flash_attn bridge - npu_torch/qwen3_gated_delta_net_base.cpp (576 lines) — GDN reference - common/rms_norm.cpp, rotary_embedding.cpp, activation.cpp, dense_mlp.cpp xllm ILU kernels — synced 10 files to upstream (diffs from prior edits) These are reference implementations, NOT hand-written. Source repos: vllm-project/vllm, jd-opensource/xllm
83 lines
2.6 KiB
C++
83 lines
2.6 KiB
C++
/* Copyright 2025 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 <tuple>
|
|
|
|
#include "framework/kv_cache/kv_cache.h"
|
|
#include "framework/model/model_input_params.h"
|
|
#include "layers/common/attention_metadata.h"
|
|
|
|
namespace xllm {
|
|
namespace layer {
|
|
class AttentionImpl : public torch::nn::Module {
|
|
public:
|
|
AttentionImpl() = default;
|
|
|
|
AttentionImpl(int64_t num_heads,
|
|
int64_t head_size,
|
|
float scale,
|
|
int64_t num_kv_heads,
|
|
int64_t sliding_window);
|
|
AttentionImpl(int64_t num_heads,
|
|
int64_t head_size,
|
|
int64_t num_kv_heads,
|
|
int64_t v_head_dim,
|
|
int64_t sliding_window,
|
|
float scale,
|
|
bool use_fused_mla_qkv,
|
|
bool enable_lighting_indexer,
|
|
bool enable_mla);
|
|
|
|
std::tuple<torch::Tensor, std::optional<torch::Tensor>> forward(
|
|
const AttentionMetadata& attn_metadata,
|
|
torch::Tensor& query,
|
|
torch::Tensor& key,
|
|
torch::Tensor& value,
|
|
KVCache& kv_cache);
|
|
|
|
void prefill_forward(torch::Tensor& query,
|
|
torch::Tensor& key,
|
|
torch::Tensor& value,
|
|
torch::Tensor& output,
|
|
const torch::Tensor& k_cache,
|
|
const std::optional<torch::Tensor>& v_cache,
|
|
const AttentionMetadata& attn_metadata);
|
|
|
|
void decoder_forward(torch::Tensor& query,
|
|
torch::Tensor& output,
|
|
const torch::Tensor& k_cache,
|
|
const std::optional<torch::Tensor>& v_cache,
|
|
const AttentionMetadata& attn_metadata);
|
|
|
|
private:
|
|
int64_t num_heads_;
|
|
int64_t head_size_;
|
|
float scale_;
|
|
int64_t num_kv_heads_;
|
|
int64_t v_head_dim_;
|
|
bool use_fused_mla_qkv_;
|
|
bool enable_lighting_indexer_;
|
|
bool enable_mla_;
|
|
int64_t sliding_window_;
|
|
};
|
|
TORCH_MODULE(Attention);
|
|
|
|
} // namespace layer
|
|
} // namespace xllm
|