thek0tyara/llama-cpp-turboquant
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 4

model : add LFM2-ColBert-350M (#18607)

Browse source 
* model : add LFM2-ColBert-350M

* llama_model_n_embd_out() - returns `hparams.n_embd_out` if set and fallbacks to `hparams.n_embd`
This commit is contained in:
Tarek Dakhran 2026-01-05 19:52:56 +01:00 • committed by GitHub
parent df17a4c94f
commit 73d284a250
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
16 changed files with 118 additions and 60 deletions
Download patch file Download diff file Expand all files Collapse all files

14
examples/model-conversion/logits.cpp
View file

@ -161,9 +161,9 @@ int main(int argc, char ** argv) {
std::vector<float> embd_out;
if (embedding_mode) {
const int n_embd = llama_model_n_embd(model);
const int n_embd_out = llama_model_n_embd_out(model);
const int n_embd_count = pooling_enabled ? 1 : batch.n_tokens;
const int n_embeddings = n_embd * n_embd_count;
const int n_embeddings = n_embd_out * n_embd_count;
float * embeddings;
type = "-embeddings";
@ -177,7 +177,7 @@ int main(int argc, char ** argv) {
embeddings = llama_get_embeddings(ctx);
}
printf("Embedding dimension: %d\n", n_embd);
printf("Embedding dimension: %d\n", n_embd_out);
printf("\n");
// Print embeddings in the specified format
@ -185,16 +185,16 @@ int main(int argc, char ** argv) {
printf("embedding %d: ", j);
// Print first 3 values
for (int i = 0; i < 3 && i < n_embd; i++) {
printf("%9.6f ", embeddings[j * n_embd + i]);
for (int i = 0; i < 3 && i < n_embd_out; i++) {
printf("%9.6f ", embeddings[j * n_embd_out + i]);
}
printf(" ... ");
// Print last 3 values
for (int i = n_embd - 3; i < n_embd; i++) {
for (int i = n_embd_out - 3; i < n_embd_out; i++) {
if (i >= 0) {
printf("%9.6f ", embeddings[j * n_embd + i]);
printf("%9.6f ", embeddings[j * n_embd_out + i]);
}
}