llama-cli made a change

Now instead of --log-disable we need to redirect to stderr:

ggml-org/llama.cpp#9418

Signed-off-by: Eric Curtin <ecurtin@redhat.com>

Author: ericcurtin

ramalama/common.py

@@ -34,10 +34,16 @@ def available(cmd):
     return shutil.which(cmd) is not None
 
 
-def exec_cmd(args):
+def exec_cmd(args, stderr=True):
     if x:
         print(*args)
 
+    if not stderr:
+        # Replace the current process with a new one
+        # Redirecting stderr to /dev/null
+        with open(os.devnull, 'w') as devnull:
+            os.dup2(devnull.fileno(), sys.stderr.fileno())
+
     try:
         return os.execvp(args[0], args)
     except Exception:

ramalama/model.py

@@ -86,7 +86,6 @@ def run(self, args):
             "llama-cli",
             "-m",
             symlink_path,
-            "--log-disable",
             "--in-prefix",
             "",
             "--in-suffix",
@@ -98,7 +97,7 @@ def run(self, args):
         if not args.ARGS:
             exec_args.append("-cnv")
 
-        exec_cmd(exec_args)
+        exec_cmd(exec_args, False)
 
     def serve(self, args):
         symlink_path = self.pull(args)