found a buffer overflow

[iacore]

Found this when reading the code

in llama.cpp when n_dims > 2

            int32_t nelements = 1;
            int32_t ne[2] = { 1, 1 };
            for (int i = 0; i < n_dims; ++i) {
                fin.read(reinterpret_cast<char *>(&ne[i]), sizeof(ne[i]));
                nelements *= ne[i];
            }

[sw]

#801 completely changed the loading code, read_tensor_metadata now checks that 1 <= n_dims <= 2, but only after calling file.read_raw(shard.ne.data(), sizeof(shard.ne[0]) * n_dims);, so that's still not perfect.

[iacore]

thanks!