From: Sigbjørn Skjæret <redacted>
Date: Thu, 18 Sep 2025 11:28:22 +0000 (+0200)
Subject: cuda : add missing F32<->I32 entries in ggml_cuda_cpy_fn (llama/16060)
X-Git-Tag: v0.9.1~13
X-Git-Url: https://git.djapps.eu/?a=commitdiff_plain;h=9c9ca2ce4e42d1066d9aca3497c0020e3076445e;p=pkg%2Fggml%2Fsources%2Fggml

cuda : add missing F32<->I32 entries in ggml_cuda_cpy_fn (llama/16060)
---

diff --git a/src/ggml-cuda/cpy.cu b/src/ggml-cuda/cpy.cu
index 8567c3d5..1b763a62 100644
--- a/src/ggml-cuda/cpy.cu
+++ b/src/ggml-cuda/cpy.cu
@@ -441,6 +441,10 @@ void* ggml_cuda_cpy_fn(const ggml_tensor * src0, ggml_tensor * src1) {
         return (void*) cpy_flt<cpy_1_flt<nv_bfloat16, nv_bfloat16>>;
     } else if (src0->type == GGML_TYPE_BF16 && src1->type == GGML_TYPE_F32) {
         return (void*) cpy_flt<cpy_1_flt<nv_bfloat16, float>>;
+    } else if (src0->type == GGML_TYPE_F32 && src1->type == GGML_TYPE_I32) {
+        return (void*) cpy_flt<cpy_1_flt<float, int32_t>>;
+    } else if (src0->type == GGML_TYPE_I32 && src1->type == GGML_TYPE_F32) {
+        return (void*) cpy_flt<cpy_1_flt<int32_t, float>>;
     } else {
         GGML_ABORT("%s: unsupported type combination (%s to %s)\n", __func__,
                 ggml_type_name(src0->type), ggml_type_name(src1->type));