e0e912f49b
* llama : add option to override tensor buffers * ggml : fix possible underflow in ggml_nbytes
* llama : add option to override tensor buffers * ggml : fix possible underflow in ggml_nbytes