vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

imatrix: WIP

Browse source
This commit is contained in:
Iwan Kawrakow 2024-01-10 18:25:26 +02:00
parent 675819df2b
commit 055a0c2e12
3 changed files with 110 additions and 38 deletions
Download patch file Download diff file Expand all files Collapse all files

144
examples/imatrix/imatrix.cpp
View file

@ -33,31 +33,25 @@ struct StatParams {
bool collect_output_weight = false; bool collect_output_weight = false;
}; };
static void ik_save_statistics(const char * fname, const std::unordered_map<std::string, Stats>& stats, int ncall) { class IMatrixCollector {
std::ofstream out(fname, std::ios::binary); public:
int n_entries = stats.size(); IMatrixCollector() = default;
out.write((const char*)&n_entries, sizeof(n_entries)); void set_parameters(StatParams&& params) { m_params = std::move(params); }
for (auto& p : stats) { void collect_imatrix(const struct ggml_tensor * src0, const struct ggml_tensor * src1);
int len = p.first.size(); void save_imatrix() const;
out.write((const char*)&len, sizeof(len)); private:
out.write(p.first.c_str(), len); std::unordered_map<std::string, Stats> m_stats;
out.write((const char*)&p.second.ncall, sizeof(p.second.ncall)); StatParams m_params;
int nval = p.second.values.size(); std::mutex m_mutex;
out.write((const char*)&nval, sizeof(nval)); int m_last_call = 0;
if (nval > 0) out.write((const char*)p.second.values.data(), nval*sizeof(float)); };
}
fprintf(stderr, "%s: stored collected data after %d calls in %s\n",__func__,ncall,fname);
}
static void ik_collect_imatrix(const struct ggml_tensor * src0, const struct ggml_tensor * src1) { void IMatrixCollector::collect_imatrix(const struct ggml_tensor * src0, const struct ggml_tensor * src1) {
static int last_call = 0;
static std::mutex mutex;
if (src1->ne[1] < 16 || src1->type != GGML_TYPE_F32) return; if (src1->ne[1] < 16 || src1->type != GGML_TYPE_F32) return;
//if (strncmp(src0->name, "blk.", 4) != 0 && strcmp(src0->name, "output.weight") != 0) return; if (!(strncmp(src0->name, "blk.", 4) == 0 || (m_params.collect_output_weight && strcmp(src0->name, "output.weight") == 0))) return;
if (strncmp(src0->name, "blk.", 4) != 0) return; //if (strncmp(src0->name, "blk.", 4) != 0) return;
std::lock_guard<std::mutex> lock(mutex); std::lock_guard<std::mutex> lock(m_mutex);
auto& g_stats = ik_get_stats(); auto& e = m_stats[src0->name];
auto& e = g_stats[src0->name];
if (e.values.empty()) { if (e.values.empty()) {
e.values.resize(src1->ne[0], 0); e.values.resize(src1->ne[0], 0);
} }
@ -66,21 +60,89 @@ static void ik_collect_imatrix(const struct ggml_tensor * src0, const struct ggm
exit(1); //GGML_ASSERT(false); exit(1); //GGML_ASSERT(false);
} }
++e.ncall; ++e.ncall;
printf("%s[%d]: %s, %d x %d, %d\n",__func__,last_call,src0->name,(int)src1->ne[0],(int)src1->ne[1],(int)src1->type); printf("%s[%d]: %s, %d x %d, %d\n",__func__,m_last_call,src0->name,(int)src1->ne[0],(int)src1->ne[1],(int)src1->type);
for (int row = 0; row < (int)src1->ne[1]; ++row) { for (int row = 0; row < (int)src1->ne[1]; ++row) {
const float * x = (const float *)src1->data + row * src1->ne[0]; const float * x = (const float *)src1->data + row * src1->ne[0];
for (int j = 0; j < (int)src1->ne[0]; ++j) { for (int j = 0; j < (int)src1->ne[0]; ++j) {
e.values[j] += x[j]*x[j]; e.values[j] += x[j]*x[j];
} }
} }
if (e.ncall > last_call) { if (e.ncall > m_last_call) {
last_call = e.ncall; m_last_call = e.ncall;
if (last_call % 10 == 0) { if (m_last_call % m_params.n_output_frequency == 0) {
ik_save_statistics("stats.dat", g_stats, last_call); save_imatrix();
} }
} }
} }
void IMatrixCollector::save_imatrix() const {
const char * fname = m_params.ofile.empty() ? "imatrix.dat" : m_params.ofile.c_str();
std::ofstream out(fname, std::ios::binary);
int n_entries = m_stats.size();
out.write((const char*)&n_entries, sizeof(n_entries));
for (auto& p : m_stats) {
int len = p.first.size();
out.write((const char*)&len, sizeof(len));
out.write(p.first.c_str(), len);
out.write((const char*)&p.second.ncall, sizeof(p.second.ncall));
int nval = p.second.values.size();
out.write((const char*)&nval, sizeof(nval));
if (nval > 0) out.write((const char*)p.second.values.data(), nval*sizeof(float));
}
fprintf(stderr, "%s: stored collected data after %d calls in %s\n",__func__,m_last_call,fname);
}
static IMatrixCollector g_collector;
//static void ik_save_statistics(const char * fname, const std::unordered_map<std::string, Stats>& stats, int ncall) {
// std::ofstream out(fname, std::ios::binary);
// int n_entries = stats.size();
// out.write((const char*)&n_entries, sizeof(n_entries));
// for (auto& p : stats) {
// int len = p.first.size();
// out.write((const char*)&len, sizeof(len));
// out.write(p.first.c_str(), len);
// out.write((const char*)&p.second.ncall, sizeof(p.second.ncall));
// int nval = p.second.values.size();
// out.write((const char*)&nval, sizeof(nval));
// if (nval > 0) out.write((const char*)p.second.values.data(), nval*sizeof(float));
// }
// fprintf(stderr, "%s: stored collected data after %d calls in %s\n",__func__,ncall,fname);
//}
static void ik_collect_imatrix(const struct ggml_tensor * src0, const struct ggml_tensor * src1) {
g_collector.collect_imatrix(src0, src1);
//static int last_call = 0;
//static std::mutex mutex;
//if (src1->ne[1] < 16 || src1->type != GGML_TYPE_F32) return;
////if (strncmp(src0->name, "blk.", 4) != 0 && strcmp(src0->name, "output.weight") != 0) return;
//if (strncmp(src0->name, "blk.", 4) != 0) return;
//std::lock_guard<std::mutex> lock(mutex);
//auto& g_stats = ik_get_stats();
//auto& e = g_stats[src0->name];
//if (e.values.empty()) {
// e.values.resize(src1->ne[0], 0);
//}
//else if (e.values.size() != (size_t)src1->ne[0]) {
// fprintf(stderr, "Oops: inconsistent size for %s (%d vs %d)\n", src0->name, (int)e.values.size(), (int)src1->ne[0]);
// exit(1); //GGML_ASSERT(false);
//}
//++e.ncall;
//printf("%s[%d]: %s, %d x %d, %d\n",__func__,last_call,src0->name,(int)src1->ne[0],(int)src1->ne[1],(int)src1->type);
//for (int row = 0; row < (int)src1->ne[1]; ++row) {
// const float * x = (const float *)src1->data + row * src1->ne[0];
// for (int j = 0; j < (int)src1->ne[0]; ++j) {
// e.values[j] += x[j]*x[j];
// }
//}
//if (e.ncall > last_call) {
// last_call = e.ncall;
// if (last_call % 10 == 0) {
// ik_save_statistics("stats.dat", g_stats, last_call);
// }
//}
}
struct results_log_softmax { struct results_log_softmax {
double log_softmax; double log_softmax;
@ -273,7 +335,9 @@ int main(int argc, char ** argv) {
StatParams sparams; StatParams sparams;
std::vector<char*> args; std::vector<char*> args;
for (int iarg = 1; iarg < argc-1; ++iarg) { args.push_back(argv[0]);
int iarg = 1;
for (; iarg < argc-1; ++iarg) {
std::string arg{argv[iarg]}; std::string arg{argv[iarg]};
if (arg == "-o" || arg == "--output-file") { if (arg == "-o" || arg == "--output-file") {
sparams.ofile = argv[++iarg]; sparams.ofile = argv[++iarg];
@ -287,6 +351,9 @@ int main(int argc, char ** argv) {
args.push_back(argv[iarg]); args.push_back(argv[iarg]);
} }
} }
if (iarg < argc) {
args.push_back(argv[iarg]);
}
gpt_params params; gpt_params params;
params.n_batch = 512; params.n_batch = 512;
@ -294,7 +361,10 @@ int main(int argc, char ** argv) {
return 1; return 1;
} }
ggml_set_stat_collection(ik_collect_imatrix); g_collector.set_parameters(std::move(sparams));
ggml_set_imatrix_collection(ik_collect_imatrix);
ggml_set_imatrix_collection(ik_collect_imatrix);
params.logits_all = true; params.logits_all = true;
params.n_batch = std::min(params.n_batch, params.n_ctx); params.n_batch = std::min(params.n_batch, params.n_ctx);
@ -340,12 +410,14 @@ int main(int argc, char ** argv) {
if (!OK) { if (!OK) {
return 1; return 1;
} }
auto& stats = ik_get_stats();
int ncall = 0; g_collector.save_imatrix();
for (auto& s : stats) { //auto& stats = ik_get_stats();
ncall = std::max(ncall, s.second.ncall); //int ncall = 0;
} //for (auto& s : stats) {
ik_save_statistics(sparams.ofile.c_str(), stats, ncall); // ncall = std::max(ncall, s.second.ncall);
//}
//ik_save_statistics(sparams.ofile.c_str(), stats, ncall);
llama_print_timings(ctx); llama_print_timings(ctx);

2
ggml.c
View file

@ -396,7 +396,7 @@ static void ggml_vec_dot_f16(const int n, float * restrict s, ggml_fp16_t * rest
ggml_collect_imatrix_t g_imatrix_collect = NULL; ggml_collect_imatrix_t g_imatrix_collect = NULL;
void ggml_set_stat_collection(ggml_collect_imatrix_t imatrix_collect) { void ggml_set_imatrix_collection(ggml_collect_imatrix_t imatrix_collect) {
g_imatrix_collect = imatrix_collect; g_imatrix_collect = imatrix_collect;
} }

2
ggml.h
View file

@ -2077,7 +2077,7 @@ extern "C" {
// Importance matrix // Importance matrix
// //
typedef void(*ggml_collect_imatrix_t)(const struct ggml_tensor * src0, const struct ggml_tensor * src1); typedef void(*ggml_collect_imatrix_t)(const struct ggml_tensor * src0, const struct ggml_tensor * src1);
GGML_API void ggml_set_stat_collection(ggml_collect_imatrix_t imatrix_collect); GGML_API void ggml_set_imatrix_collection(ggml_collect_imatrix_t imatrix_collect);
// //
// gguf // gguf