llama : prevent usage of k-quants when tensor size is not a multiple of 256 (#1921)

* Fix examples/metal

* k-quants: prevent usage when tensor size is not divisible by 256

---------

Co-authored-by: Iwan Kawrakow <redacted>

diff --git a/llama.cpp b/llama.cpp
index c165d3239e63f09ebb8cea73eaa561412fdaa5b1..dfbb85a682baad745c620053eecc254f8c28590c 100644 (file)
--- a/llama.cpp
+++ b/llama.cpp
@@ -19,6 +19,11 @@
 #ifdef GGML_USE_METAL
 #include "ggml-metal.h"
 #endif
+#ifdef GGML_USE_K_QUANTS
+#ifndef QK_K
+#define QK_K 256
+#endif
+#endif
 
 #include <array>
 #include <ctime>
@@ -2491,6 +2496,17 @@ static void llama_model_quantize_internal(const std::string & fname_inp, const s
         } else {
             new_type = quantized_type;
 #ifdef GGML_USE_K_QUANTS
+            if (quantized_type == GGML_TYPE_Q2_K || quantized_type == GGML_TYPE_Q3_K || quantized_type == GGML_TYPE_Q4_K ||
+                quantized_type == GGML_TYPE_Q5_K || quantized_type == GGML_TYPE_Q6_K) {
+                int nx = tensor.ne.at(0);
+                int ny = tensor.ne.at(0);
+                if (nx % QK_K != 0 || ny % QK_K != 0) {
+                    fprintf(stderr, "\n\n========================= Tensor sizes %d x %d are not divisible by %d\n",nx,ny,QK_K);
+                    fprintf(stderr, "This is required to be able to use k-quants for now!\n");
+                    fprintf(stderr, "========================================================================================\n\n");
+                    throw std::runtime_error("Unsupported tensor size encountered\n");
+                }
+            }
             if (tensor.name == "output.weight") {
                new_type = GGML_TYPE_Q6_K;
             } else if (tensor.name.find("attention.wv.weight") != std::string::npos) {