From: Georgi Gerganov <redacted>
Date: Tue, 6 Jun 2023 19:54:39 +0000 (+0300)
Subject: llama : fix vram_scratch var
X-Git-Tag: gguf-v0.4.0~675
X-Git-Url: https://git.djapps.eu/?a=commitdiff_plain;h=2d7bf110edd8c49209401a16132052cba706ffd0;p=pkg%2Fggml%2Fsources%2Fllama.cpp

llama : fix vram_scratch var
---

diff --git a/llama.cpp b/llama.cpp
index cf512ccd..16d6f6ef 100644
--- a/llama.cpp
+++ b/llama.cpp
@@ -1076,6 +1076,7 @@ static void llama_model_load_internal(
 
     // prepare memory for the weights
     size_t vram_weights = 0;
+    size_t vram_scratch = 0;
     {
         const uint32_t n_embd  = hparams.n_embd;
         const uint32_t n_layer = hparams.n_layer;
@@ -1152,8 +1153,9 @@ static void llama_model_load_internal(
         fprintf(stderr, "%s: mem required  = %7.2f MB (+ %7.2f MB per state)\n", __func__,
                 mem_required / 1024.0 / 1024.0, mem_required_state / 1024.0 / 1024.0);
 
+        (void) vram_scratch;
 #ifdef GGML_USE_CUBLAS
-        const size_t vram_scratch = n_batch * MB;
+        vram_scratch = n_batch * MB;
         ggml_cuda_set_scratch_size(vram_scratch);
         if (n_gpu_layers > 0) {
             fprintf(stderr, "%s: allocating batch_size x 1 MB = %ld MB VRAM for the scratch buffer\n",