diff --git a/ggml/src/ggml-cpu/ggml-cpu.c b/ggml/src/ggml-cpu/ggml-cpu.c index 07df840334..1171431b8c 100644 --- a/ggml/src/ggml-cpu/ggml-cpu.c +++ b/ggml/src/ggml-cpu/ggml-cpu.c @@ -2858,7 +2858,7 @@ struct ggml_cplan ggml_graph_plan( case GGML_OP_CPY: case GGML_OP_DUP: { - if (ggml_is_quantized(node->type) || node->type == GGML_TYPE_F8_E4M3 || + if (ggml_is_quantized(node->type) || // F16 -> BF16 and BF16 -> F16 copies go through intermediate F32 (node->src[0]->type == GGML_TYPE_F16 && node->src[1] && node->src[1]->type == GGML_TYPE_BF16) || (node->src[0]->type == GGML_TYPE_BF16 && node->src[1] && node->src[1]->type == GGML_TYPE_F16) ||