cuda : mark I16 and I32 ops as unsupported

ggml-ci
ggerganov · Jan 3, 2024 · d55356d · d55356d
1 parent 75e3fd8
commit d55356d
Showing 1 changed file with 10 additions and 3 deletions.
diff --git a/ggml-cuda.cu b/ggml-cuda.cu
@@ -10039,14 +10039,22 @@ static bool ggml_backend_cuda_supports_op(ggml_backend_t backend, const ggml_ten
                 }
                 return false;
             } break;
+        case GGML_OP_DUP:
+        case GGML_OP_REPEAT:
+        case GGML_OP_CONCAT:
+            {
+                ggml_type src0_type = op->src[0]->type;
+                if (src0_type != GGML_TYPE_I32 && src0_type != GGML_TYPE_I16) {
+                    return true;
+                }
+                return false;
+            } break;
         case GGML_OP_NONE:
         case GGML_OP_RESHAPE:
         case GGML_OP_VIEW:
         case GGML_OP_PERMUTE:
         case GGML_OP_TRANSPOSE:
         case GGML_OP_NORM:
-        case GGML_OP_REPEAT:
-        case GGML_OP_DUP:
         case GGML_OP_ADD:
         case GGML_OP_MUL:
         case GGML_OP_DIV:
@@ -10063,7 +10071,6 @@ static bool ggml_backend_cuda_supports_op(ggml_backend_t backend, const ggml_ten
         case GGML_OP_SUM_ROWS:
         case GGML_OP_ARGSORT:
         case GGML_OP_ACC:
-        case GGML_OP_CONCAT:
         case GGML_OP_GROUP_NORM:
         case GGML_OP_UPSCALE:
         case GGML_OP_PAD: