From bdff7729b1716284ea3234a87d52930cf2fdce80 Mon Sep 17 00:00:00 2001 From: Georgi Gerganov Date: Mon, 8 Sep 2025 21:01:25 +0300 Subject: [PATCH] metal : fix batch size for MUL_MAT_ID --- ggml/src/ggml-metal/ggml-metal.m | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/ggml/src/ggml-metal/ggml-metal.m b/ggml/src/ggml-metal/ggml-metal.m index 958b14b180..e6e4557f2c 100644 --- a/ggml/src/ggml-metal/ggml-metal.m +++ b/ggml/src/ggml-metal/ggml-metal.m @@ -6743,8 +6743,9 @@ static bool ggml_backend_metal_device_supports_buft(ggml_backend_dev_t dev, ggml static int64_t get_op_batch_size(const struct ggml_tensor * op) { switch (op->op) { case GGML_OP_MUL_MAT: - case GGML_OP_MUL_MAT_ID: return op->ne[1]; + case GGML_OP_MUL_MAT_ID: + return op->ne[2]; default: return ggml_nrows(op); }