Function ggml_compute_forward_flash_ff

ggml.c:12718–12740

}

static void ggml_compute_forward_flash_ff(
        const struct ggml_compute_params * params,
        const struct ggml_tensor * a,
        const struct ggml_tensor * b0,
        const struct ggml_tensor * b1,
        const struct ggml_tensor * c0,
        const struct ggml_tensor * c1,
        struct ggml_tensor * dst) {
    switch (b0->type) {
        case GGML_TYPE_F16:
            {
                ggml_compute_forward_flash_ff_f16(params, a, b0, b1, c0, c1, dst);
            } break;
        case GGML_TYPE_F32:
            {
                GGML_ASSERT(false); // TODO
            } break;
        default:
            {
                GGML_ASSERT(false);
            } break;
    }
}

// ggml_compute_forward_flash_attn_back