Merge branch 'fix' into speculative

This commit is contained in:
Li, Zonghang 2025-06-14 18:57:17 +04:00
commit 45de284f3d

View file

@ -1421,8 +1421,9 @@ static bool assign_layers_to_device(
if (n_m < static_cast<uint32_t>(std::floor(W * vec_z_gpu[m]))) { if (n_m < static_cast<uint32_t>(std::floor(W * vec_z_gpu[m]))) {
// if there is still free GPU memory // if there is still free GPU memory
has_free_gpu_memory = true; has_free_gpu_memory = true;
} else if (w_m > n_m) { }
// if the GPU is overloaded if (w_m > n_m) {
// if layers are offloaded to CPU
has_gpu_overload = true; has_gpu_overload = true;
} }
} else if (!in_set(m, M4)) { } else if (!in_set(m, M4)) {