revert supports_op change

JohannesGaessler · JohannesGaessler · commit 33c94caf13d5 · 2025-02-10T18:39:05.000+01:00
diff --git a/ggml/src/ggml-cuda/ggml-cuda.cu b/ggml/src/ggml-cuda/ggml-cuda.cu
@@ -3205,7 +3205,8 @@ static bool ggml_backend_cuda_device_supports_op(ggml_backend_dev_t dev, const g
             if (op->src[0]->ne[0] == 256 && op->src[1]->type == GGML_TYPE_F16 && op->src[2]->type == GGML_TYPE_F16) {
                 return true;
             }
-            return fp16_mma_available(ggml_cuda_info().devices[dev_ctx->device].cc);
+            return fp16_mma_available(ggml_cuda_info().devices[dev_ctx->device].cc) &&
+                op->src[1]->type == GGML_TYPE_F16 && op->src[2]->type == GGML_TYPE_F16;
         }
         case GGML_OP_CROSS_ENTROPY_LOSS:
         case GGML_OP_CROSS_ENTROPY_LOSS_BACK:

Original file line number	Diff line number	Diff line change
`@@ -3205,7 +3205,8 @@ static bool ggml_backend_cuda_device_supports_op(ggml_backend_dev_t dev, const g`
`3205`	`3205`	`if (op->src[0]->ne[0] == 256 && op->src[1]->type == GGML_TYPE_F16 && op->src[2]->type == GGML_TYPE_F16) {`
`3206`	`3206`	`return true;`
`3207`	`3207`	`}`
`3208`		`- return fp16_mma_available(ggml_cuda_info().devices[dev_ctx->device].cc);`
	`3208`	`+ return fp16_mma_available(ggml_cuda_info().devices[dev_ctx->device].cc) &&`
	`3209`	`+ op->src[1]->type == GGML_TYPE_F16 && op->src[2]->type == GGML_TYPE_F16;`
`3209`	`3210`	`}`
`3210`	`3211`	`case GGML_OP_CROSS_ENTROPY_LOSS:`
`3211`	`3212`	`case GGML_OP_CROSS_ENTROPY_LOSS_BACK:`