Method release_storage

src/framework/assets/tensor_source.cpp:319–321

        set_tensor_bytes(tensor, converted.data(), converted.size() * sizeof(ggml_fp16_t), name);
        return;
    }

    std::vector<ggml_bf16_t> converted(values.size());
    const int64_t count = static_cast<int64_t>(values.size());
    const int64_t chunks = (count + kF32ConvertChunkElements - 1) / kF32ConvertChunkElements;
    #pragma omp parallel for schedule(static)