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

fix quantizing of merged experts

Browse source
This commit is contained in:
slaren 2024-04-02 17:07:14 +02:00
parent 68d21debe4
commit f27cbf3610
1 changed files with 13 additions and 4 deletions
Download patch file Download diff file Expand all files Collapse all files

17
llama.cpp
View file

@ -13534,7 +13534,7 @@ static void llama_model_quantize_internal(const std::string & fname_inp, const s
if (it == imatrix_data->end()) { if (it == imatrix_data->end()) {
LLAMA_LOG_INFO("\n====== %s: did not find weights for %s\n", __func__, tensor->name); LLAMA_LOG_INFO("\n====== %s: did not find weights for %s\n", __func__, tensor->name);
} else { } else {
if (it->second.size() == (size_t)tensor->ne[0]) { if (it->second.size() == (size_t)tensor->ne[0]*tensor->ne[2]) {
imatrix = it->second.data(); imatrix = it->second.data();
} else { } else {
LLAMA_LOG_INFO("\n====== %s: imatrix size %d is different from tensor size %d for %s\n", __func__, LLAMA_LOG_INFO("\n====== %s: imatrix size %d is different from tensor size %d for %s\n", __func__,
@ -13575,15 +13575,24 @@ static void llama_model_quantize_internal(const std::string & fname_inp, const s
new_data = work.data(); new_data = work.data();
const int n_per_row = tensor->ne[0]; const int n_per_row = tensor->ne[0];
const int nrows = nelements / n_per_row; const int nrows = tensor->ne[1];
static const int min_chunk_size = 32 * 512; static const int min_chunk_size = 32 * 512;
const int chunk_size = n_per_row >= min_chunk_size ? n_per_row : n_per_row * ((min_chunk_size + n_per_row - 1)/n_per_row); const int chunk_size = n_per_row >= min_chunk_size ? n_per_row : n_per_row * ((min_chunk_size + n_per_row - 1)/n_per_row);
const int nchunk = (nelements + chunk_size - 1)/chunk_size; const int nelements_matrix = tensor->ne[0] * tensor->ne[1];
const int nchunk = (nelements_matrix + chunk_size - 1)/chunk_size;
const int nthread_use = nthread > 1 ? std::max(1, std::min(nthread, nchunk)) : 1; const int nthread_use = nthread > 1 ? std::max(1, std::min(nthread, nchunk)) : 1;
new_size = llama_tensor_quantize_internal(new_type, f32_data, new_data, chunk_size, nrows, n_per_row, imatrix, workers, nthread_use);
// quantize each expert separately since they have different importance matrices
new_size = 0;
for (int64_t i03 = 0; i03 < tensor->ne[2]; ++i03) {
const float * f32_data_03 = f32_data + i03 * nelements_matrix;
void * new_data_03 = (char *)new_data + ggml_row_size(new_type, n_per_row) * i03 * nrows;
const float * imatrix_03 = imatrix ? imatrix + i03 * n_per_row : nullptr;
new_size += llama_tensor_quantize_internal(new_type, f32_data_03, new_data_03, chunk_size, nrows, n_per_row, imatrix_03, workers, nthread_use);
}
LLAMA_LOG_INFO("size = %8.2f MiB -> %8.2f MiB\n", ggml_nbytes(tensor)/1024.0/1024.0, new_size/1024.0/1024.0); LLAMA_LOG_INFO("size = %8.2f MiB -> %8.2f MiB\n", ggml_nbytes(tensor)/1024.0/1024.0, new_size/1024.0/1024.0);
} }
total_size_org += ggml_nbytes(tensor); total_size_org += ggml_nbytes(tensor);