bump scratch buffers to 256 MB

Author: NouamaneTazi

ggml.c

@@ -3983,7 +3983,8 @@ struct ggml_tensor * ggml_new_tensor_impl(
         };
     } else {
         if (ctx->scratch.offs + size_needed > ctx->scratch.size) {
-            GGML_PRINT("%s: not enough space in the scratch memory\n", __func__);
+            GGML_PRINT("%s: not enough space in the scratch memory pool (needed %zu, available %zu)\n",
+                    __func__, ctx->scratch.offs + size_needed, ctx->scratch.size);
             assert(false);
             return NULL;
         }

main.cpp

@@ -418,10 +418,10 @@ bool starcoder_eval(
 
     // use 2 scratch buffers
     // TODO: very hacky solution - reimplement in a more elegant way
-    static size_t scr0_size = 128u*1024*1024;
+    static size_t scr0_size = 256u*1024*1024;
     static void * scr0 = malloc(scr0_size);
 
-    static size_t scr1_size = 128u*1024*1024;
+    static size_t scr1_size = 256u*1024*1024;
     static void * scr1 = malloc(scr1_size);
 
     if (mem_per_token > 0 && mem_per_token*N > buf_size) {