apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2026-04-24 01:29:57 +08:00
Code Issues Actions 11 Packages Projects Releases Wiki Activity

c++ code format (#4527)

Browse Source
This commit is contained in:
zhupengyang
2025-10-22 17:59:50 +08:00
committed by GitHub
parent d7bcedf421
commit 3a6883ac1a
97 changed files with 8760 additions and 7382 deletions
Download Patch File Download Diff File Expand all files Collapse all files
custom_ops/xpu_ops/src/ops/update_inputs_v1.cc
+47 -48
View File
@@ -18,55 +18,54 @@
#include "xpu/plugin.h"
void UpdateInputesV1(const paddle::Tensor &stop_flags,
const paddle::Tensor &not_need_stop, // only on cpu
const paddle::Tensor &seq_lens_this_time,
const paddle::Tensor &seq_lens_encoder,
const paddle::Tensor &seq_lens_decoder,
const paddle::Tensor &step_seq_lens_decoder,
const paddle::Tensor &prompt_lens,
const paddle::Tensor &topk_ids,
const paddle::Tensor &input_ids,
const paddle::Tensor &block_tables,
const paddle::Tensor &stop_nums,
const paddle::Tensor &next_tokens,
const paddle::Tensor &is_block_step,
const int block_size) {
phi::XPUPlace place(phi::backends::xpu::GetXPUCurrentDeviceId());
auto dev_ctx =
paddle::experimental::DeviceContextPool::Instance().Get(place);
auto xpu_ctx = static_cast<const phi::XPUContext *>(dev_ctx);
const paddle::Tensor &not_need_stop, // only on cpu
const paddle::Tensor &seq_lens_this_time,
const paddle::Tensor &seq_lens_encoder,
const paddle::Tensor &seq_lens_decoder,
const paddle::Tensor &step_seq_lens_decoder,
const paddle::Tensor &prompt_lens,
const paddle::Tensor &topk_ids,
const paddle::Tensor &input_ids,
const paddle::Tensor &block_tables,
const paddle::Tensor &stop_nums,
const paddle::Tensor &next_tokens,
const paddle::Tensor &is_block_step,
const int block_size) {
phi::XPUPlace place(phi::backends::xpu::GetXPUCurrentDeviceId());
auto dev_ctx = paddle::experimental::DeviceContextPool::Instance().Get(place);
auto xpu_ctx = static_cast<const phi::XPUContext *>(dev_ctx);
const int max_bsz = stop_flags.shape()[0];
const int now_bsz = seq_lens_this_time.shape()[0];
// std::cout << "now_bsz: " << now_bsz << std::endl;
const int input_ids_stride = input_ids.shape()[1];
const int block_num_per_seq = block_tables.shape()[1];
auto not_need_stop_gpu = not_need_stop.copy_to(stop_flags.place(), false);
int r = baidu::xpu::api::plugin::update_inputs_v1(
xpu_ctx->x_context(),
const_cast<bool *>(not_need_stop_gpu.data<bool>()),
const_cast<int *>(seq_lens_this_time.data<int>()),
const_cast<int *>(seq_lens_encoder.data<int>()),
const_cast<int *>(seq_lens_decoder.data<int>()),
const_cast<int *>(step_seq_lens_decoder.data<int>()),
const_cast<int64_t *>(prompt_lens.data<int64_t>()),
const_cast<int64_t *>(topk_ids.data<int64_t>()),
const_cast<int64_t *>(input_ids.data<int64_t>()),
const_cast<int *>(block_tables.data<int>()),
stop_nums.data<int64_t>(),
const_cast<bool *>(stop_flags.data<bool>()),
const_cast<bool *>(is_block_step.data<bool>()),
next_tokens.data<int64_t>(),
now_bsz,
max_bsz,
input_ids_stride,
block_num_per_seq,
block_size);
PD_CHECK(r == 0, "baidu::xpu::api::plugin::update_inputs_kernel_v1 failed.");
auto not_need_stop_cpu =
not_need_stop_gpu.copy_to(not_need_stop.place(), false);
bool *not_need_stop_data = const_cast<bool *>(not_need_stop.data<bool>());
not_need_stop_data[0] = not_need_stop_cpu.data<bool>()[0];
const int max_bsz = stop_flags.shape()[0];
const int now_bsz = seq_lens_this_time.shape()[0];
// std::cout << "now_bsz: " << now_bsz << std::endl;
const int input_ids_stride = input_ids.shape()[1];
const int block_num_per_seq = block_tables.shape()[1];
auto not_need_stop_gpu = not_need_stop.copy_to(stop_flags.place(), false);
int r = baidu::xpu::api::plugin::update_inputs_v1(
xpu_ctx->x_context(),
const_cast<bool *>(not_need_stop_gpu.data<bool>()),
const_cast<int *>(seq_lens_this_time.data<int>()),
const_cast<int *>(seq_lens_encoder.data<int>()),
const_cast<int *>(seq_lens_decoder.data<int>()),
const_cast<int *>(step_seq_lens_decoder.data<int>()),
const_cast<int64_t *>(prompt_lens.data<int64_t>()),
const_cast<int64_t *>(topk_ids.data<int64_t>()),
const_cast<int64_t *>(input_ids.data<int64_t>()),
const_cast<int *>(block_tables.data<int>()),
stop_nums.data<int64_t>(),
const_cast<bool *>(stop_flags.data<bool>()),
const_cast<bool *>(is_block_step.data<bool>()),
next_tokens.data<int64_t>(),
now_bsz,
max_bsz,
input_ids_stride,
block_num_per_seq,
block_size);
PD_CHECK(r == 0, "baidu::xpu::api::plugin::update_inputs_kernel_v1 failed.");
auto not_need_stop_cpu =
not_need_stop_gpu.copy_to(not_need_stop.place(), false);
bool *not_need_stop_data = const_cast<bool *>(not_need_stop.data<bool>());
not_need_stop_data[0] = not_need_stop_cpu.data<bool>()[0];
}
PD_BUILD_OP(update_inputs_v1)