From 61b37f5f448e6c5d4f707ddcded3c797e01159c6 Mon Sep 17 00:00:00 2001 From: Kawrakow Date: Fri, 7 Aug 2026 17:09:18 +0300 Subject: [PATCH] Reduce the indexer temporary buffer size (#2271) --- ggml/src/ggml-cuda/indexer_topk.cu | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/ggml/src/ggml-cuda/indexer_topk.cu b/ggml/src/ggml-cuda/indexer_topk.cu index 0bdd7ea8..8cdc353b 100644 --- a/ggml/src/ggml-cuda/indexer_topk.cu +++ b/ggml/src/ggml-cuda/indexer_topk.cu @@ -146,7 +146,7 @@ void ggml_cuda_op_indexer_topk(ggml_backend_cuda_context & ctx, ggml_tensor * ds } - constexpr int64_t k_max_work_buffer_elements = 1 << 28; + constexpr int64_t k_max_work_buffer_elements = 1 << 26; int max_rows = k_max_work_buffer_elements / n_kv / n_head; if (max_rows < 1) max_rows = 1;