Fix memory allocation issues and seg faults

llama.cpp

@@ -102,6 +102,9 @@ struct llama_context {
     // decode output (2-dimensional array: [n_tokens][n_vocab])
     std::vector<float> logits;
     bool logits_all = false;
+
+    // work buffer for transformer evaluation
+    std::vector<uint8_t> buf_eval;
 };
 
 struct llama_context_params llama_context_default_params() {
@@ -627,27 +630,19 @@ static bool llama_eval_internal(
     const int n_rot   = hparams.n_embd/hparams.n_head;
 
     auto & mem_per_token = lctx.mem_per_token;
+    auto & buf_eval = lctx.buf_eval;
 
-    // TODO: fix this hardcoded size
+    if (mem_per_token*(n_past + N + 16) > buf_eval.size()) {
-    static size_t buf_size = 512u*1024*1024;
+        const size_t buf_size_new = 1.618*buf_eval.size();
-    static void * buf = malloc(buf_size);
 
-    if (mem_per_token > 0 && mem_per_token*N > buf_size) {
+        //fprintf(stderr, "\n%s: reallocating buffer from %zu to %zu bytes\n", __func__, buf_eval.size(), buf_size_new);
-        const size_t buf_size_new = 1.3*(mem_per_token*N); // add 30% to account for ggml object overhead
-        //fprintf(stderr, "\n%s: reallocating buffer from %zu to %zu bytes\n", __func__, buf_size, buf_size_new);
 
-        // reallocate
+        buf_eval.resize(buf_size_new);
-        buf_size = buf_size_new;
-        buf = realloc(buf, buf_size);
-        if (buf == nullptr) {
-            fprintf(stderr, "%s: failed to allocate %zu bytes\n", __func__, buf_size);
-            return false;
-        }
     }
 
     struct ggml_init_params params = {
-        /*.mem_size   =*/ buf_size,
+        /*.mem_size   =*/ buf_eval.size(),
-        /*.mem_buffer =*/ buf,
+        /*.mem_buffer =*/ buf_eval.data(),
     };
 
     struct ggml_context * ctx0 = ggml_init(params);
@@ -832,10 +827,11 @@ static bool llama_eval_internal(
         memcpy(logits_out.data(), (float *) ggml_get_data(inpL) + (n_vocab*(N-1)), sizeof(float)*n_vocab);
     }
 
-    if (mem_per_token == 0) {
+    if (N == 1) {
-        mem_per_token = ggml_used_mem(ctx0)/N;
+        mem_per_token = ggml_used_mem(ctx0)/(n_past + N);
     }
-    //fprintf(stderr, "used_mem = %zu\n", ggml_used_mem(ctx0));
+
+    //fprintf(stderr, "\nused_mem = %zu, %zu MB\n", ggml_used_mem(ctx0), ggml_used_mem(ctx0)/1024/1024);
 
     ggml_free(ctx0);
 
@@ -1416,6 +1412,8 @@ struct llama_context * llama_init_from_file(
         return nullptr;
     }
 
+    ctx->buf_eval.resize(512u*1024u*1024u);
+
     return ctx;
 }