Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

ggml : fix handling of zero blocks in IQ quants #7955

Merged
merged 1 commit into from
Jun 16, 2024
Merged

ggml : fix handling of zero blocks in IQ quants #7955

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 3 additions & 3 deletions ggml-quants.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13139,7 +13139,7 @@ static int iq1_find_best_neighbour(const uint16_t * restrict neighbours, const u
const float * restrict xval, const float * restrict weight, float * scale, int8_t * restrict L, int ngrid) {
int num_neighbors = neighbours[0];
GGML_ASSERT(num_neighbors > 0);
float best_score = 0;
float best_score = -FLT_MAX;
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This function isn't used any more, but why negative FLT_MAX?

Copy link
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Any negative value works, but used -FLT_MAX for consistency with usage of FLT_MAX elsewhere in the source

int grid_index = -1;
for (int j = 1; j <= num_neighbors; ++j) {
const int8_t * pg = (const int8_t *)(grid + neighbours[j]);
Expand Down Expand Up @@ -13337,7 +13337,7 @@ static void quantize_row_iq1_s_impl(const float * restrict x, void * restrict vy
sumw[j+1] = sumw[j] + weight[i];
}
}
float best_score = 0, scale = max;
float best_score = -FLT_MIN, scale = max;
int besti1 = -1, besti2 = -1, best_shift = 0;
for (int i1 = 0; i1 <= block_size; ++i1) {
for (int i2 = i1; i2 <= block_size; ++i2) {
Expand Down Expand Up @@ -13513,7 +13513,7 @@ static void quantize_row_iq1_m_impl(const float * restrict x, void * restrict vy
idx[2*j] = j;
}
qsort(pairs, block_size, 2*sizeof(float), iq1_sort_helper);
float best_score = 0, scale = max;
float best_score = -FLT_MIN, scale = max;
int besti1 = -1, besti2 = -1, best_k = -1;
// 0: +, +
// 1: +, -
Expand Down
Loading