mirror of
https://github.com/ollama/ollama.git
synced 2026-03-27 02:58:43 +07:00
87d21c7fc0
The CLI now links to the lazy-load MLX code, but that still happens in init functions. On internal MLX errors, the CLI exits before it has a chance to start. This change re-wires the MLX error handling so it doesn't exit by default. The MLX based runners currently expect exits on failure, so they re-initialize the default error handling. We can refine error handling for better go stack traces in the future.
23 lines
816 B
C
23 lines
816 B
C
// mlx_error_handler.h - Safe error handling for MLX initialization
|
|
// This replaces the default exit(-1) MLX error handler during init()
|
|
// so that GPU failures don't kill the process.
|
|
|
|
#ifndef MLX_ERROR_HANDLER_H
|
|
#define MLX_ERROR_HANDLER_H
|
|
|
|
// Enter safe mode before any MLX compute calls during init().
|
|
// Replaces the default exit(-1) handler with one that silently stores errors.
|
|
void mlx_set_safe_init_mode(void);
|
|
|
|
// Restore the default MLX error handler (exit on error).
|
|
// Call from runner entry points after confirming MLX is available.
|
|
void mlx_set_default_error_mode(void);
|
|
|
|
// Check whether an error occurred while in safe init mode.
|
|
int mlx_had_init_error(void);
|
|
|
|
// Get the error message from the last init error, or NULL if none.
|
|
const char* mlx_get_init_error(void);
|
|
|
|
#endif // MLX_ERROR_HANDLER_H
|