From dccb0126378da098af757d858e99dca4b5cfd883 Mon Sep 17 00:00:00 2001
From: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
Date: Sun, 7 Apr 2024 05:09:17 +0200
Subject: [PATCH] llama: dbrx: quantize fix n_attention_wv tensor name

---
 llama.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/llama.cpp b/llama.cpp
index a76d5f7ec..5922db5a5 100644
--- a/llama.cpp
+++ b/llama.cpp
@@ -13717,7 +13717,7 @@ static void llama_model_quantize_internal(const std::string & fname_inp, const s
         const std::string name = ggml_get_name(meta);
 
         // TODO: avoid hardcoded tensor names - use the TN_* constants
-        if (name.find("attn_v.weight") != std::string::npos || name.find("attn_qkv.weight") != std::string::npos || name.find("attn.Wqkv.weight") != std::string::npos) {
+        if (name.find("attn_v.weight") != std::string::npos || name.find("attn_qkv.weight") != std::string::npos || name == LLM_TN(model.arch)(LLM_TENSOR_ATTN_QKV, "weight")) {
             ++qs.n_attention_wv;
         } else if (name == LLM_TN(model.arch)(LLM_TENSOR_OUTPUT, "weight")) {
             qs.has_output = true;