apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-12-24 13:28:13 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[PD Disaggregation] support DP via v1 router and decouple DP and EP (#5197)

Browse Source 
* [fix] support DP via v1 router and decouple DP and EP

* [fix] fix scripts

* [fix] reset model path

* [fix] dp use get_output_ep, fix router port type, update scripts

* [merge] merge with latest code

* [chore] remove some debug log

* [fix] fix code style check

* [fix] fix test_multi_api_server for log_dir name

* [chore] reduce logs

* Apply suggestions from code review

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

---------

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
This commit is contained in:
Yonghua Li
2025-12-04 15:38:43 +08:00
committed by GitHub
parent 5cd17fd662
commit f4119d51b4
15 changed files with 394 additions and 146 deletions
Download Patch File Download Diff File Expand all files Collapse all files

97
custom_ops/gpu_ops/get_output.cc
View File

@@ -26,71 +26,70 @@
#define MAX_BSZ 512
// #define GET_OUTPUT_DEBUG
struct msgdata {
long mtype;
int mtext[MAX_BSZ + 2]; // stop_flag, bsz, tokens
long mtype;
int mtext[MAX_BSZ + 2]; // stop_flag, bsz, tokens
};
void GetOutput(const paddle::Tensor& x,
int64_t rank_id,
bool wait_flag,
int msg_queue_id) {
if (rank_id > 0) {
return;
}
static struct msgdata msg_rcv;
if (const char* inference_msg_queue_id_env_p =
std::getenv("INFERENCE_MSG_QUEUE_ID")) {
std::string inference_msg_queue_id_env_str(
inference_msg_queue_id_env_p);
int inference_msg_queue_id_from_env =
std::stoi(inference_msg_queue_id_env_str);
#ifdef GET_OUTPUT_DEBUG
std::cout << "Your INFERENCE_MSG_QUEUE_ID is: "
<< inference_msg_queue_id_from_env << std::endl;
#endif
msg_queue_id = inference_msg_queue_id_from_env;
}
static key_t key = ftok("/dev/shm", msg_queue_id);
static int msgid = msgget(key, IPC_CREAT | 0666);
#ifdef GET_OUTPUT_DEBUG
std::cout << "get_output_key: " << key << std::endl;
std::cout << "get_output msgid: " << msgid << std::endl;
#endif
int64_t* out_data = const_cast<int64_t*>(x.data<int64_t>());
int ret = -1;
if (!wait_flag) {
ret = msgrcv(msgid, &msg_rcv, (MAX_BSZ + 2) * 4, 0, IPC_NOWAIT);
} else {
ret = msgrcv(msgid, &msg_rcv, (MAX_BSZ + 2) * 4, 0, 0);
}
if (ret == -1) {
out_data[0] = -2;
out_data[1] = 0;
return;
}
int bsz = msg_rcv.mtext[1];
for (int64_t i = 0; i < bsz + 2; i++) {
out_data[i] = (int64_t)msg_rcv.mtext[i];
}
#ifdef GET_OUTPUT_DEBUG
std::cout << "get_output finished: " << msgid << std::endl;
#endif
if (rank_id > 0) {
return;
}
static struct msgdata msg_rcv;
if (const char* inference_msg_queue_id_env_p =
std::getenv("INFERENCE_MSG_QUEUE_ID")) {
std::string inference_msg_queue_id_env_str(inference_msg_queue_id_env_p);
int inference_msg_queue_id_from_env =
std::stoi(inference_msg_queue_id_env_str);
#ifdef GET_OUTPUT_DEBUG
std::cout << "Your INFERENCE_MSG_QUEUE_ID is: "
<< inference_msg_queue_id_from_env << std::endl;
#endif
msg_queue_id = inference_msg_queue_id_from_env;
}
static key_t key = ftok("/dev/shm", msg_queue_id);
static int msgid = msgget(key, IPC_CREAT | 0666);
#ifdef GET_OUTPUT_DEBUG
std::cout << "get_output_key: " << key << std::endl;
std::cout << "get_output msgid: " << msgid << std::endl;
#endif
int64_t* out_data = const_cast<int64_t*>(x.data<int64_t>());
int ret = -1;
if (!wait_flag) {
ret = msgrcv(msgid, &msg_rcv, (MAX_BSZ + 2) * 4, 0, IPC_NOWAIT);
} else {
ret = msgrcv(msgid, &msg_rcv, (MAX_BSZ + 2) * 4, 0, 0);
}
if (ret == -1) {
out_data[0] = -2;
out_data[1] = 0;
return;
}
int bsz = msg_rcv.mtext[1];
for (int64_t i = 0; i < bsz + 2; i++) {
out_data[i] = (int64_t)msg_rcv.mtext[i];
}
#ifdef GET_OUTPUT_DEBUG
std::cout << "get_output finished: " << msgid << std::endl;
#endif
return;
}
void GetOutputStatic(const paddle::Tensor& x, int64_t rank_id, bool wait_flag) {
GetOutput(x, rank_id, wait_flag, 1);
GetOutput(x, rank_id, wait_flag, 1);
}
void GetOutputDynamic(const paddle::Tensor& x,
int64_t rank_id,
bool wait_flag,
int msg_queue_id) {
GetOutput(x, rank_id, wait_flag, msg_queue_id);
GetOutput(x, rank_id, wait_flag, msg_queue_id);
}
PD_BUILD_STATIC_OP(get_output)