root/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-27 20:04:35 +00:00
Code Issues Actions 3 Packages Projects Releases Wiki Activity

llama : support offloading result_norm + comments

Browse Source
This commit is contained in:
Georgi Gerganov 2023-10-29 07:36:07 +02:00
parent 51c4f9ee9f
commit 4e98897ede
No known key found for this signature in database
GPG Key ID: 449E073F9DC10735
1 changed files with 15 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

17
llama.cpp
View File

@ -5452,12 +5452,16 @@ static struct ggml_cgraph * llama_build_graph(
} while (0); } while (0);
// offload layers // offload layers
// TODO: this code will be obsoleted with backend v2
{ {
const int n_layer = model.hparams.n_layer; const int n_layer = model.hparams.n_layer;
const int n_gpu_layers = model.n_gpu_layers; const int n_gpu_layers = model.n_gpu_layers;
const int i_gpu_start = n_layer - n_gpu_layers; const int i_gpu_start = n_layer - n_gpu_layers;
// should we offload the final norm? yes if we are not computing embeddings
const bool off_res_norm = !lctx.embedding.empty();
// offload functions set the tensor output backend to GPU // offload functions set the tensor output backend to GPU
// tensors are GPU-accelerated if any input or the output has been offloaded // tensors are GPU-accelerated if any input or the output has been offloaded
offload_func_t offload_func_nr = ggml_offload_nop; // nr = non-repeating offload_func_t offload_func_nr = ggml_offload_nop; // nr = non-repeating
@ -5566,7 +5570,7 @@ static struct ggml_cgraph * llama_build_graph(
{ "out_norm_0", offload_func_nr }, { "out_norm_0", offload_func_nr },
{ "out_norm_0_w", offload_func_nr }, { "out_norm_0_w", offload_func_nr },
//{ "result_norm", offload_func_nr }, // TODO CPU + GPU mirrored backend { "result_norm", off_res_norm ? offload_func_nr : ggml_offload_nop },
//{ "result_output", offload_func }, //{ "result_output", offload_func },
}; };
@ -5584,7 +5588,8 @@ static struct ggml_cgraph * llama_build_graph(
const std::string name = cur->name; const std::string name = cur->name;
if (k_offload_func.find(name) == k_offload_func.end()) { const auto it = k_offload_func.find(name);
if (it == k_offload_func.end()) {
// if a tensor that is not view hasn't been offloaded, we warn the user // if a tensor that is not view hasn't been offloaded, we warn the user
if (worst_case && cur->view_src == nullptr) { if (worst_case && cur->view_src == nullptr) {
LLAMA_LOG_WARN("%s: node %4d %32s: not offloaded (ref: %s)\n", __func__, LLAMA_LOG_WARN("%s: node %4d %32s: not offloaded (ref: %s)\n", __func__,
@ -5595,7 +5600,7 @@ static struct ggml_cgraph * llama_build_graph(
} }
// count the number of layers and respect the provided n_gpu_layers // count the number of layers and respect the provided n_gpu_layers
offload_func_t f = k_offload_func.at(name); offload_func_t f = it->second;
if (f == offload_func) { if (f == offload_func) {
if (ofn[name]++ < i_gpu_start) { if (ofn[name]++ < i_gpu_start) {
f = ggml_offload_nop; f = ggml_offload_nop;
@ -5753,11 +5758,13 @@ static int llama_decode_internal(
} }
// If all tensors can be run on the GPU then using more than 1 thread is detrimental. // If all tensors can be run on the GPU then using more than 1 thread is detrimental.
const bool full_offload_supported = model.arch == LLM_ARCH_LLAMA || const bool full_offload_supported =
model.arch == LLM_ARCH_LLAMA ||
model.arch == LLM_ARCH_BAICHUAN || model.arch == LLM_ARCH_BAICHUAN ||
model.arch == LLM_ARCH_FALCON || model.arch == LLM_ARCH_FALCON ||
model.arch == LLM_ARCH_REFACT || model.arch == LLM_ARCH_REFACT ||
model.arch == LLM_ARCH_MPT; model.arch == LLM_ARCH_MPT;
const bool fully_offloaded = model.n_gpu_layers >= (int) hparams.n_layer + 3; const bool fully_offloaded = model.n_gpu_layers >= (int) hparams.n_layer + 3;
if (ggml_cpu_has_cublas() && full_offload_supported && fully_offloaded) { if (ggml_cpu_has_cublas() && full_offload_supported && fully_offloaded) {
n_threads = 1; n_threads = 1;
@ -5803,6 +5810,8 @@ static int llama_decode_internal(
//} //}
// extract logits // extract logits
// TODO: do not compute and extract logits if only embeddings are needed
// need to update the graphs to skip "result_output"
{ {
auto & logits_out = lctx.logits; auto & logits_out = lctx.logits;