llama : enable LLAMA_SET_ROWS=1 by default (#14959)

ggml-ci

diff --git a/src/llama-context.cpp b/src/llama-context.cpp
index bd637f3dffe9c5436769422ef48f98490c7a47b3..958bcc0477f7b4d47dce72b559badae1c6b8e799 100644 (file)
--- a/src/llama-context.cpp
+++ b/src/llama-context.cpp
@@ -105,7 +105,7 @@ llama_context::llama_context(
 
     {
         const char * LLAMA_SET_ROWS = getenv("LLAMA_SET_ROWS");
-        supports_set_rows = LLAMA_SET_ROWS ? (atoi(LLAMA_SET_ROWS) != 0) : false;
+        supports_set_rows = LLAMA_SET_ROWS ? (atoi(LLAMA_SET_ROWS) != 0) : supports_set_rows;
 
         if (!supports_set_rows && !cparams.kv_unified) {
             LLAMA_LOG_WARN("%s: non-unified KV cache requires ggml_set_rows() - forcing unified KV cache\n", __func__);

diff --git a/src/llama-context.h b/src/llama-context.h
index 7cfdc6a51731a31fde9c2555b4bdc47434aed915..25c143d56dfb2cc567e0abd021fca2237e74828c 100644 (file)
--- a/src/llama-context.h
+++ b/src/llama-context.h
@@ -289,7 +289,7 @@ private:
 
     // env: LLAMA_SET_ROWS (temporary)
     // ref: https://github.com/ggml-org/llama.cpp/pull/14285
-    bool supports_set_rows = false;
+    bool supports_set_rows = true;
 
     // env: LLAMA_GRAPH_REUSE_DISABLE
     bool graph_reuse_disable = false;

diff --git a/src/llama-kv-cache-unified.cpp b/src/llama-kv-cache-unified.cpp
index 321dc79fc36ab708a4ac96076b3fabf200568a3e..c741014cf375776e5e4df8a38865945507099497 100644 (file)
--- a/src/llama-kv-cache-unified.cpp
+++ b/src/llama-kv-cache-unified.cpp
@@ -193,7 +193,7 @@ llama_kv_cache_unified::llama_kv_cache_unified(
     debug = LLAMA_KV_CACHE_DEBUG ? atoi(LLAMA_KV_CACHE_DEBUG) : 0;
 
     const char * LLAMA_SET_ROWS = getenv("LLAMA_SET_ROWS");
-    supports_set_rows = LLAMA_SET_ROWS ? atoi(LLAMA_SET_ROWS) != 0 : 0;
+    supports_set_rows = LLAMA_SET_ROWS ? atoi(LLAMA_SET_ROWS) != 0 : supports_set_rows;
 
     if (!supports_set_rows) {
         // ref: https://github.com/ggml-org/llama.cpp/pull/14363

diff --git a/src/llama-kv-cache-unified.h b/src/llama-kv-cache-unified.h
index 3e28e346c3fcf8d1ec09fa15de1b2bd6c4dcb3b4..342a675962e2a85668d882888d98d4e71cf115c4 100644 (file)
--- a/src/llama-kv-cache-unified.h
+++ b/src/llama-kv-cache-unified.h
@@ -230,7 +230,7 @@ private:
 
     // env: LLAMA_SET_ROWS (temporary)
     // ref: https://github.com/ggml-org/llama.cpp/pull/14285
-    bool supports_set_rows = false;
+    bool supports_set_rows = true;
 
     const llama_swa_type swa_type = LLAMA_SWA_TYPE_NONE;