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

Parameterise type

This commit is contained in:
Ed Addario 2025-08-22 09:05:55 +01:00
parent bb0d912c1f
commit 2f13fee795
No known key found for this signature in database
GPG Key ID: E7875815A3230993
1 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
src/llama-quant.cpp
View File

@ -760,8 +760,8 @@ static std::unordered_map<std::string, ggml_type> target_bpw_type(
if (quantized_buffer.size() < buffer_size) { quantized_buffer.resize(buffer_size); } if (quantized_buffer.size() < buffer_size) { quantized_buffer.resize(buffer_size); }
if (dequantized_buffer.size() < sample_element_count) { dequantized_buffer.resize(sample_element_count); } if (dequantized_buffer.size() < sample_element_count) { dequantized_buffer.resize(sample_element_count); }
std::vector row_sq_norm(sample_row_count, 0.0); std::vector<double> row_sq_norm(sample_row_count, 0.0);
std::vector bias_denominator_per_slice(ne2, 0.0); std::vector<double> bias_denominator_per_slice(ne2, 0.0);
// Precompute bias denominator per slice // Precompute bias denominator per slice
const bool has_values = (values_sample != nullptr); const bool has_values = (values_sample != nullptr);