ggml : use F16 instead of F32 in Q4_0, Q4_1, Q8_0 (#1508)

* ggml : use F16 instead of F32 in Q4_0, Q4_1 and Q8_0

* llama : bump LLAMA_FILE_VERSION to 3

* cuda : update Q4 and Q8 dequantize kernels

* ggml : fix AVX dot products

* readme : update performance table + hot topics

1 files changed, 1 insertions, 1 deletions

diff --git a/llama.h b/llama.h
index f955fa2..fd3f21e 100644
--- a/llama.h
+++ b/llama.h
@@ -19,7 +19,7 @@
 #    define LLAMA_API
 #endif
 
-#define LLAMA_FILE_VERSION           2
+#define LLAMA_FILE_VERSION           3
 #define LLAMA_FILE_MAGIC             'ggjt'
 #define LLAMA_FILE_MAGIC_UNVERSIONED 'ggml'
 #define LLAMA_SESSION_MAGIC          'ggsn'