vrr/prima.cpp
2
0
Fork 0
mirror of https://github.com/Lizonghang/prima.cpp.git synced 2025-09-08 07:39:04 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

fix cuda support

Browse source
This commit is contained in:
Lizonghang 2024-11-04 11:00:01 +04:00
parent 9d6a6845ac
commit 976a4c3534
1 changed files with 36 additions and 55 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

91
src/llama.cpp
View file

@ -17317,8 +17317,8 @@ static void llama_graph_compute(
} }
struct input_tensors { struct input_tensors {
ggml_tensor * sub_gf_out = nullptr; ggml_tensor * sub_gf_out;
ggml_tensor * inp_pos = nullptr; ggml_tensor * inp_pos;
}; };
struct sync_meta { struct sync_meta {
@ -17340,7 +17340,7 @@ static void llama_send_meta(zmq::socket_t & socket, struct sync_meta * meta) {
} }
} }
static void llama_send_tensors(zmq::socket_t & socket, struct input_tensors * tensors) { static void llama_send_tensors(zmq::socket_t & socket, struct llama_ubatch * ubatch, struct input_tensors * tensors) {
try { try {
std::vector<zmq::message_t> send_msgs; std::vector<zmq::message_t> send_msgs;
size_t buf_size = 0; size_t buf_size = 0;
@ -17348,13 +17348,13 @@ static void llama_send_tensors(zmq::socket_t & socket, struct input_tensors * te
send_msgs.emplace_back("sub_gf_out", strlen("sub_gf_out")); send_msgs.emplace_back("sub_gf_out", strlen("sub_gf_out"));
send_msgs.emplace_back(tensors->sub_gf_out->ne, sizeof(tensors->sub_gf_out->ne)); send_msgs.emplace_back(tensors->sub_gf_out->ne, sizeof(tensors->sub_gf_out->ne));
buf_size = tensors->sub_gf_out->ne[0] * tensors->sub_gf_out->ne[1] * sizeof(float); buf_size = tensors->sub_gf_out->ne[0] * tensors->sub_gf_out->ne[1] * sizeof(float);
send_msgs.emplace_back(tensors->sub_gf_out->data, buf_size); send_msgs.emplace_back(ubatch->backend_embd, buf_size);
if (tensors->inp_pos) { if (tensors->inp_pos) {
send_msgs.emplace_back("inp_pos", strlen("inp_pos")); send_msgs.emplace_back("inp_pos", strlen("inp_pos"));
send_msgs.emplace_back(tensors->inp_pos->ne, sizeof(tensors->inp_pos->ne[0])); send_msgs.emplace_back(tensors->inp_pos->ne, sizeof(tensors->inp_pos->ne[0]));
buf_size = tensors->inp_pos->ne[0] * sizeof(int32_t); buf_size = tensors->inp_pos->ne[0] * sizeof(int32_t);
send_msgs.emplace_back(tensors->inp_pos->data, buf_size); send_msgs.emplace_back(ubatch->pos, buf_size);
} }
zmq::send_multipart(socket, send_msgs); zmq::send_multipart(socket, send_msgs);
@ -17385,7 +17385,7 @@ static int llama_recv_meta(zmq::socket_t & socket, struct sync_meta * meta) {
return 0; return 0;
} }
static void llama_recv_tensors(zmq::socket_t & socket, input_tensors * tensors) { static void llama_recv_tensors(zmq::socket_t & socket, struct llama_ubatch * ubatch, struct llama_context * lctx, const bool is_out_embd=false) {
std::vector<zmq::message_t> recv_msgs; std::vector<zmq::message_t> recv_msgs;
if (!zmq::recv_multipart(socket, std::back_inserter(recv_msgs))) { if (!zmq::recv_multipart(socket, std::back_inserter(recv_msgs))) {
LLAMA_LOG_INFO("Failed to receive tensor data.\n"); LLAMA_LOG_INFO("Failed to receive tensor data.\n");
@ -17396,19 +17396,15 @@ static void llama_recv_tensors(zmq::socket_t & socket, input_tensors * tensors)
zmq::message_t &dims_msg = recv_msgs[i + 1]; zmq::message_t &dims_msg = recv_msgs[i + 1];
zmq::message_t &data_msg = recv_msgs[i + 2]; zmq::message_t &data_msg = recv_msgs[i + 2];
if (key == "sub_gf_out" && tensors->sub_gf_out) { if (key == "sub_gf_out") {
int64_t * dims = static_cast<int64_t*>(dims_msg.data()); int64_t * dims = static_cast<int64_t *>(dims_msg.data());
size_t buf_size = dims[0] * dims[1] * sizeof(float); size_t buf_size = dims[0] * dims[1] * sizeof(float);
GGML_ASSERT(dims[0] == tensors->sub_gf_out->ne[0]); float * batch_embd = is_out_embd ? ubatch->out_embd : ubatch->backend_embd;
GGML_ASSERT(dims[1] == tensors->sub_gf_out->ne[1]); std::memcpy(batch_embd, data_msg.data(), buf_size);
GGML_ASSERT(data_msg.size() == buf_size); } else if (key == "inp_pos") {
std::memcpy(tensors->sub_gf_out->data, data_msg.data(), buf_size); int64_t * dims = static_cast<int64_t *>(dims_msg.data());
} else if (key == "inp_pos" && tensors->inp_pos) {
int64_t * dims = static_cast<int64_t*>(dims_msg.data());
size_t buf_size = dims[0] * sizeof(int32_t); size_t buf_size = dims[0] * sizeof(int32_t);
GGML_ASSERT(dims[0] == tensors->inp_pos->ne[0]); std::memcpy(ubatch->pos, data_msg.data(), buf_size);
GGML_ASSERT(data_msg.size() == buf_size);
std::memcpy(tensors->inp_pos->data, data_msg.data(), buf_size);
} }
} }
} }
@ -17725,29 +17721,14 @@ static int llama_decode_internal(
for (size_t i = 0; i < (size_t)gf.size(); ++i) { for (size_t i = 0; i < (size_t)gf.size(); ++i) {
sub_gf = gf[i]; sub_gf = gf[i];
// receive data from other nodes
if (n_world > 1 && !(my_rank == 0 && i == 0) && !(my_rank == 0 && is_last_l)) { if (n_world > 1 && !(my_rank == 0 && i == 0) && !(my_rank == 0 && is_last_l)) {
// receive data from previous nodes
input_tensors tensors;
const bool is_out_embd = my_rank == 0 && i == (size_t)gf.size() - 1; const bool is_out_embd = my_rank == 0 && i == (size_t)gf.size() - 1;
tensors.sub_gf_out = is_out_embd ? lctx.out_embd : lctx.backend_embd; llama_recv_tensors(*lctx.recv_socket, &ubatch, &lctx, is_out_embd);
tensors.inp_pos = lctx.inp_pos;
llama_recv_tensors(*lctx.recv_socket, &tensors);
is_last_l = my_rank == 0 && i == (size_t)gf.size() - 1;
size_t buf_size = tensors.sub_gf_out->ne[0] * tensors.sub_gf_out->ne[1] * ggml_element_size(tensors.sub_gf_out);
if (!is_last_l) {
memcpy(ubatch.backend_embd, tensors.sub_gf_out->data, buf_size);
} else {
memcpy(ubatch.out_embd, tensors.sub_gf_out->data, buf_size);
}
if (my_rank != 0 && i == 0) {
buf_size = tensors.inp_pos->ne[0] * ggml_element_size(tensors.inp_pos);
memcpy(ubatch.pos, tensors.inp_pos->data, buf_size);
}
} }
if (i > 0) { // ensure ggml_backend_tensor_get_async of the previous subgraph has finished
// ensure ggml_backend_tensor_get_async of the previous subgraph has finished if (i > 0 && (n_world == 1 || (my_rank == 0 && is_last_l))) {
ggml_backend_sched_synchronize(lctx.sched[i - 1]); ggml_backend_sched_synchronize(lctx.sched[i - 1]);
} }
@ -17772,28 +17753,28 @@ static int llama_decode_internal(
is_last_l = (cur_l == static_cast<int>(n_layer) - 1); is_last_l = (cur_l == static_cast<int>(n_layer) - 1);
} }
// send the result to the next node (or the master) float * embd_buf;
if (n_world == 1 || (my_rank == 0 && is_last_l)) { if (n_world == 1 || (my_rank == 0 && is_last_l)) {
size_t buf_size = sub_gf_out->ne[0] * sub_gf_out->ne[1] * sizeof(float); embd_buf = is_last_l ? ubatch.out_embd : ubatch.backend_embd;
float * embd_buf = is_last_l ? ubatch.out_embd : ubatch.backend_embd;
ggml_backend_t backend = ggml_backend_sched_get_tensor_backend(lctx.sched[i], sub_gf_out);
GGML_ASSERT(buf_size <= ggml_nbytes(sub_gf_out));
GGML_ASSERT(backend != nullptr);
GGML_ASSERT(embd_buf != nullptr);
ggml_backend_tensor_get_async(backend, sub_gf_out, embd_buf, 0, buf_size);
} else { } else {
input_tensors tensors; embd_buf = ubatch.backend_embd;
tensors.sub_gf_out = sub_gf_out; }
if (i == 0 && !is_last_l && my_rank != n_world - 1) { GGML_ASSERT(embd_buf != nullptr);
tensors.inp_pos = lctx.inp_pos;
const size_t buf_size = ubatch.n_tokens * ggml_element_size(tensors.inp_pos); // copy device data to cpu memory
memcpy(tensors.inp_pos->data, ubatch.pos, buf_size); size_t buf_size = sub_gf_out->ne[0] * sub_gf_out->ne[1] * sizeof(float);
} ggml_backend_t backend = ggml_backend_sched_get_tensor_backend(lctx.sched[i], sub_gf_out);
GGML_ASSERT(buf_size <= ggml_nbytes(sub_gf_out));
GGML_ASSERT(backend != nullptr);
ggml_backend_tensor_get_async(backend, sub_gf_out, embd_buf, 0, buf_size);
// send the result to the next node or the master
if (!(n_world == 1 || (my_rank == 0 && is_last_l))) {
struct input_tensors tensors = {sub_gf_out, lctx.inp_pos};
const bool is_to_master = my_rank != 0 && is_last_l; const bool is_to_master = my_rank != 0 && is_last_l;
zmq::socket_t * s = is_to_master ? lctx.master_socket : lctx.send_socket; zmq::socket_t * s = is_to_master ? lctx.master_socket : lctx.send_socket;
llama_send_tensors(*s, &tensors); ggml_backend_sched_synchronize(lctx.sched[i]);
llama_send_tensors(*s, &ubatch, &tensors);
} }
// overlap memory scheduling with other nodes' communication and computing // overlap memory scheduling with other nodes' communication and computing