support bf16 and quantized type (llama/20803)

diff --git a/src/ggml-sycl/ggml-sycl.cpp b/src/ggml-sycl/ggml-sycl.cpp
index 2ec1421841beec4911f596909c5137f861cfef02..456b1699fa3adaf557a94057beb849da1fcacd81 100644 (file)
--- a/src/ggml-sycl/ggml-sycl.cpp
+++ b/src/ggml-sycl/ggml-sycl.cpp
@@ -4667,22 +4667,8 @@ static bool ggml_backend_sycl_device_supports_op(ggml_backend_dev_t dev, const g
                 if (a->ne[3] != b->ne[3]) {
                     return false;
                 }
-                ggml_type a_type = a->type;
-                if (a_type == GGML_TYPE_IQ4_NL  || a_type == GGML_TYPE_IQ4_XS ||
-                    a_type == GGML_TYPE_IQ3_XXS || a_type == GGML_TYPE_IQ3_S  ||
-                    a_type == GGML_TYPE_IQ2_XXS || a_type == GGML_TYPE_IQ2_XS || a_type == GGML_TYPE_IQ2_S ||
-                    a_type == GGML_TYPE_IQ1_S || a_type == GGML_TYPE_IQ1_M
-                    ) {
-                    if (b->ne[1] == 1 && ggml_nrows(b) > 1) {
-                        return false;
-                    }
-                }
+
                 ggml_type src0_type = op->src[0]->type;
-                if (src0_type == GGML_TYPE_BF16 ) {
-                    // TODO: support GGML_TYPE_BF16
-                    // FIXME: keep a list of supported types to avoid breaking the backend when a new type is added
-                    return false;
-                }
 
                 // TODO: The configuration below needs more work to be supported with oneDNN
                 if (ggml_is_permuted(a) && !ggml_is_contiguous(a) &&