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

Merge 6414642d1a into a77feb5d71

Browse source
This commit is contained in:
Sigbjørn Skjæret 2024-08-27 10:30:30 +01:00 • committed by GitHub
commit ed1455ca77
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
2 changed files with 18 additions and 11 deletions
Download patch file Download diff file Expand all files Collapse all files

11
examples/imatrix/imatrix.cpp
View file

@ -250,8 +250,7 @@ void IMatrixCollector::save_imatrix(int ncall) const {
}
if (n_zeros > 0) {
fprintf(stderr, "%s: entry '%40s' has partial data (%.2f%%) - skipping\n", __func__, kv.first.c_str(), 100.0f * (n_all - n_zeros) / n_all);
continue;
fprintf(stderr, "%s: entry '%40s' has partial data (%.2f%%)\n", __func__, kv.first.c_str(), 100.0f * (n_all - n_zeros) / n_all);
}
n_entries++;
@ -275,7 +274,7 @@ void IMatrixCollector::save_imatrix(int ncall) const {
if (nval > 0) {
std::vector<float> tmp(nval);
for (int i = 0; i < nval; i++) {
tmp[i] = (stat.values[i] / static_cast<float>(stat.counts[i])) * static_cast<float>(stat.ncall);
tmp[i] = stat.counts[i] ? (stat.values[i] / static_cast<float>(stat.counts[i])) * static_cast<float>(stat.ncall) : 0.0f;
}
out.write((const char*)tmp.data(), nval*sizeof(float));
}
@ -344,8 +343,10 @@ bool IMatrixCollector::load_imatrix(const char * fname) {
// Recreate the state as expected by save_imatrix(), and corerct for weighted sum.
for (int i = 0; i < nval; i++) {
e.values[i] += tmp[i];
e.counts[i] += ncall;
if (std::isnormal(tmp[i])) {
e.values[i] += tmp[i];
e.counts[i] += ncall;
}
}
e.ncall += ncall;

18
src/llama.cpp
View file

@ -16767,19 +16767,25 @@ static void llama_model_quantize_internal(const std::string & fname_inp, const s
if (params->only_copy) {
ftype = model.ftype;
}
const std::unordered_map<std::string, std::vector<float>> * imatrix_data = nullptr;
std::unordered_map<std::string, std::vector<float>> * imatrix_data = nullptr;
if (params->imatrix) {
imatrix_data = static_cast<const std::unordered_map<std::string, std::vector<float>>*>(params->imatrix);
imatrix_data = static_cast<std::unordered_map<std::string, std::vector<float>>*>(params->imatrix);
if (imatrix_data) {
LLAMA_LOG_INFO("================================ Have weights data with %d entries\n",int(imatrix_data->size()));
qs.has_imatrix = true;
// check imatrix for nans or infs
for (const auto & kv : *imatrix_data) {
for (float f : kv.second) {
if (!std::isfinite(f)) {
throw std::runtime_error(format("imatrix contains non-finite value %f\n", f));
for (auto it = imatrix_data->begin(); it != imatrix_data->end();) {
bool remove_entry = false;
for (float f : it->second) {
if (!std::isnormal(f)) {
LLAMA_LOG_WARN("imatrix entry \"%s\" contains non-normal value %f, skipping!\n", it->first.c_str(), f);
remove_entry = true;
break;
}
}
it = remove_entry ? imatrix_data->erase(it) : ++it;
}
}
}