From: CausalLM <redacted>
Date: Tue, 27 Aug 2024 06:58:22 +0000 (+0800)
Subject: llama : fix ChatGLM4 wrong shape (#9194)
X-Git-Tag: upstream/0.0.4488~853
X-Git-Url: https://git.djapps.eu/?a=commitdiff_plain;h=2e59d61c1b321431c597c1f12249273427d5640d;p=pkg%2Fggml%2Fsources%2Fllama.cpp

llama : fix ChatGLM4 wrong shape (#9194)

This should fix THUDM/glm-4-9b-chat-1m and CausalLM/miniG
---

diff --git a/src/llama.cpp b/src/llama.cpp
index 7c148b83..f5097224 100644
--- a/src/llama.cpp
+++ b/src/llama.cpp
@@ -8116,8 +8116,8 @@ static bool llm_load_tensors(
 
                         layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd});
 
-                        layer.wqkv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + (hparams.n_embd_head_k << 2)});
-                        layer.bqkv = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_QKV, "bias", i),   {n_embd + (hparams.n_embd_head_k << 2)});
+                        layer.wqkv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa});
+                        layer.bqkv = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_QKV, "bias", i),   {n_embd + 2*n_embd_gqa});
 
                         layer.wo   = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd});