happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Improve dequantized_buffer fill

This commit is contained in:
Ed Addario 2025-08-22 09:17:58 +01:00
parent 897decbe8a
commit f05c8483d8
No known key found for this signature in database
GPG Key ID: E7875815A3230993
1 changed files with 3 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
src/llama-quant.cpp
View File

@ -843,12 +843,9 @@ static std::unordered_map<std::string, ggml_type> target_bpw_type(
LLAMA_LOG_WARN("%s: unsupported quantization type %s\n", __func__, ggml_type_name(quant_type));
return 1e35;
}
size_t done = 0;
while (done < sample_element_count) {
const size_t chunk = std::min((size_t)n_per_row, sample_element_count - done);
traits->to_float(quantized_buffer.data() + done / n_per_row * row_size, dequantized_buffer.data() + done, (int)chunk);
done += chunk;
const size_t row_size = ggml_row_size(quant_type, n_per_row);
for (size_t r = 0; r < sample_row_count; ++r) {
traits->to_float(quantized_buffer.data() + r * row_size, dequantized_buffer.data() + r * n_per_row, (int)n_per_row);
}
}
}