llama : add support for qwen3 reranker (#15824)

2025-09-25 03:53:09 -05:00
parent dfcd53f7ec
commit b5bd037832
9 changed files with 166 additions and 78 deletions
--- a/src/llama-model.cpp
+++ b/src/llama-model.cpp
@@ -3167,6 +3167,9 @@ bool llama_model::load_tensors(llama_model_loader & ml) {
                        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
                    }

+                    // output rerank head
+                    cls_out = create_tensor(tn(LLM_TENSOR_CLS_OUT, "weight"), {n_embd, hparams.n_cls_out}, TENSOR_NOT_REQUIRED);
+
                    for (int i = 0; i < n_layer; ++i) {
                        auto & layer = layers[i];