Skip to content

LLM inference in C/C++

C++ ggml

ggml-webgpu: fix flash_attn supports_op check for overlapping KV (#28205)

Masashi Yoshimura36a73916ee0c
diff --git a/ggml/src/ggml-webgpu/ggml-webgpu.cpp b/ggml/src/ggml-webgpu/ggml-webgpu.cpp
index ebc8a5e782..11a4fc46fe 100644
--- a/ggml/src/ggml-webgpu/ggml-webgpu.cpp
+++ b/ggml/src/ggml-webgpu/ggml-webgpu.cpp
@@ -4544,7 +4544,7 @@ static bool ggml_backend_webgpu_device_supports_op(ggml_backend_dev_t dev, const
break;
}
if (ggml_webgpu_tensor_binding_overlap(ctx->webgpu_global_ctx, src1, src2) &&
- src1->type != src2->type && !ggml_is_quantized(src1->type) && !ggml_is_quantized(src2->type)) {
+ src1->type != src2->type && !(ggml_is_quantized(src1->type) && ggml_is_quantized(src2->type))) {
supports_op = false;
break;
}