whisper : add Q4_1 model sizes

1 year ago · 3f08ce7004
parent b7621b4fda
commit 3f08ce7004
1 changed files with 9 additions and 0 deletions
--- a/examples/whisper/whisper.cpp
+++ b/examples/whisper/whisper.cpp
@ -271,6 +271,15 @@ static const std::map<ggml_type, std::map<e_model, size_t>> MEM_REQ_MODEL = {
            { MODEL_LARGE,   940ull*MB },
        },
    },
+    { GGML_TYPE_Q4_1,
+        {
+            { MODEL_TINY,     31ull*MB },
+            { MODEL_BASE,     57ull*MB },
+            { MODEL_SMALL,   181ull*MB },
+            { MODEL_MEDIUM,  559ull*MB },
+            { MODEL_LARGE,  1122ull*MB },
+        },
+    },
 };

 static const std::map<e_model, size_t> MEM_REQ_KV_SELF = {