llama : fix ChatGLM4 wrong shape (#9194)

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

Author: RealJosephus

src/llama.cpp

@@ -9144,8 +9144,11 @@ static bool llm_load_tensors(
 
                         layer.attn_norm = create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd});
 
-                        layer.wqkv = create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + (hparams.n_embd_head_k << 2)});
-                        layer.bqkv = create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_QKV, "bias", i),   {n_embd + (hparams.n_embd_head_k << 2)});
+                        // layer.wqkv = create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + (hparams.n_embd_head_k << 2)});
+                        // layer.bqkv = create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_QKV, "bias", i),   {n_embd + (hparams.n_embd_head_k << 2)});
+
+                        layer.wqkv = create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa});
+                        layer.bqkv = create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_QKV, "bias", i),   {n_embd + 2*n_embd_gqa});
 
                         layer.wo   = create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd});