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

Exit loop if no better solution found

This commit is contained in:
Ed Addario 2025-09-22 20:09:11 +01:00
parent b748a1efa7
commit c855094dff
No known key found for this signature in database
GPG Key ID: E7875815A3230993
1 changed files with 3 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
src/llama-quant.cpp
View File

@ -1347,9 +1347,12 @@ static std::unordered_map<std::string, ggml_type> target_bpw_type(
// increase mu until we get under budget or hit a safety cap // increase mu until we get under budget or hit a safety cap
{ {
int expand = 0; int expand = 0;
size_t prev_bytes_hi = std::numeric_limits<size_t>::max();
while (true) { while (true) {
lagrange_penalty(mu_hi, choice_hi, bytes_hi, err_hi); lagrange_penalty(mu_hi, choice_hi, bytes_hi, err_hi);
if (bytes_hi <= budget_bytes) { break; } if (bytes_hi <= budget_bytes) { break; }
if (bytes_hi >= prev_bytes_hi) { break; }
prev_bytes_hi = bytes_hi;
mu_hi *= 2.0; mu_hi *= 2.0;
if (++expand > 60) { break; } // safety cap if (++expand > 60) { break; } // safety cap