Illegal instruction on Intel sandy bridge

[pikalover6]

Following the instructions on the readme, I get all the way to running the model. Then ./bin/gpt-2 -m models/gpt-2-117M/ggml-model.bin -p "This is an example" gives me this output:

main: seed = 1677652321
gpt2_model_load: loading model from 'models/gpt-2-117M/ggml-model.bin'
gpt2_model_load: n_vocab = 50257
gpt2_model_load: n_ctx = 1024
gpt2_model_load: n_embd = 768
gpt2_model_load: n_head = 12
gpt2_model_load: n_layer = 12
gpt2_model_load: f16 = 1
gpt2_model_load: ggml ctx size = 311.12 MB
Illegal instruction (core dumped)

According to everything I had read, my processor should support all of the required instruction sets. I have installed libmkl-avx21 and tried make clean/make all. Specs are i7 2600k and Ubuntu 22.04

[ggerganov]

Try to make this change to src/CMakeLists.txt and see if it fixes the issue:

diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
index da46be3..c40ca97 100644
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -30,7 +30,7 @@ if (${CMAKE_SYSTEM_PROCESSOR} MATCHES "arm" OR ${CMAKE_SYSTEM_PROCESSOR} MATCHES
     #set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mcpu=apple-m1")
 else()
     message(STATUS "x86 detected")
-    set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mavx -mavx2 -mfma -mf16c")
+    set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -march=native")
 endif()