Fix _kill_process AttributeError when _stats_logger is unset (#6417)

* Fix _kill_process AttributeError when _stats_logger is unset

LlamaCppBackend._kill_process references self._stats_logger in its finally
block, but __init__ only sets self._stats_logger = None partway through. If
__init__ raises before that line, or the backend is built via __new__ (as the
kill-path unit test does), teardown crashes with AttributeError instead of
cleaning up the process.

Guard the reference with getattr, matching the existing
hasattr(self, '_chat_template_file') guard in the same finally block. Fixes
test_kill_process_records_timestamp_on_actual_kill.

* Also guard _stdout_thread in _kill_process teardown

Review follow-up: the same finally block also reads self._stdout_thread, which
is unset on a partially-built / __new__ backend. Guard it with getattr like
_llama_log_fh below, and add a test that _kill_process tolerates a backend with
those optional attrs unset. Trim the _stats_logger comment.

---------

Co-authored-by: Michael Han <michaelhan2050@gmail.com>

studio/backend/core/inference/llama_cpp.py

@@ -6125,7 +6125,9 @@ class LlamaCppBackend:
         except Exception as e:
             logger.warning(f"Error killing llama-server process: {e}")
         finally:
-            if self._stats_logger is not None:
+            # getattr: teardown must tolerate a partially-built backend (failed
+            # __init__ or a __new__-built instance), as with _llama_log_fh below.
+            if getattr(self, "_stats_logger", None) is not None:
                 self._stats_logger.stop()
                 self._stats_logger = None
             self._process = None
@@ -6135,8 +6137,9 @@ class LlamaCppBackend:
             # Drives _wait_for_vram_settle in the next load_model; set in finally
             # so both in-process and frontend Apply paths record the kill.
             self._last_kill_monotonic = time.monotonic()
-            if self._stdout_thread is not None:
-                self._stdout_thread.join(timeout = 2)
+            stdout_thread = getattr(self, "_stdout_thread", None)
+            if stdout_thread is not None:
+                stdout_thread.join(timeout = 2)
                 self._stdout_thread = None
             fh = getattr(self, "_llama_log_fh", None)
             if fh is not None:

studio/backend/tests/test_llama_cpp_wait_for_vram_settle.py

@@ -309,6 +309,26 @@ def test_kill_process_records_timestamp_on_actual_kill():
     assert before <= backend._last_kill_monotonic <= after
 
 
+def test_kill_process_tolerates_partially_constructed_backend():
+    # Teardown must not AttributeError on a __new__-built backend that never ran
+    # __init__: _stats_logger / _stdout_thread / _llama_log_fh are left unset.
+    backend = LlamaCppBackend.__new__(LlamaCppBackend)
+
+    class _FakeProcess:
+        def terminate(self):
+            pass
+
+        def wait(self, timeout = None):
+            return 0
+
+        def kill(self):
+            pass
+
+    backend._process = _FakeProcess()
+    backend._kill_process()
+    assert backend._process is None
+
+
 def test_helper_is_static_method_callable_off_class():
     """Pin the @staticmethod binding so call sites can invoke off the class."""
     ctx, _state = _patch_probe([[]])