0b0c47fddd
Sub 655 root causes (confirmed from log analysis): 1. protocol.py: max_completion_tokens never folded into max_tokens → 162/881 replay requests rejected 400 (extra_forbidden) 2. max_num_seqs=1 → t2_n_2 test fails (needs n=2) 3. max_model_len=131072 → OOM crash at 62% replay, opencompass all 0 Fixes: - protocol.py: model_validator fold_max_completion_tokens - yaml: max_num_seqs=2, max_model_len=80000, PYTORCH_CUDA_ALLOC_CONF - topk_softmax stays =0 (corex CUB BlockReduce incompatible on BI-V100) xllm_latest import to ex_engine/: - npu_torch layers: GDN(1164L), Qwen3.5 GDN, attention, fused_moe - cuda/moe kernels: topk_softmax_kernels.cuh, moe_combine, moe_compute_index - npu kernels: causal_conv1d, recurrent_gated_delta_rule - model headers: qwen3_5.h, qwen3_next.h
71 lines
2.2 KiB
C++
71 lines
2.2 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);
|
|
|
|
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 sliding_window_;
|
|
};
|
|
TORCH_MODULE(Attention);
|
|
|
|
} // namespace layer
|
|
} // namespace xllm
|