From 7dd9b58e3304508db15696462774108e7380ec9b Mon Sep 17 00:00:00 2001 From: Daniel Han Date: Tue, 20 Oct 2026 19:00:00 -0700 Subject: [PATCH 1/2] pr-set: pin #303 (CUDA FA MMA in-kernel q4_0/q8_0 K/V dequant for 3-16 tokens) --- scripts/unsloth/feature-checks.json | 6 ++++++ scripts/unsloth/pr-set.json | 3 ++- 2 files changed, 8 insertions(+), 1 deletion(-) diff --git a/scripts/unsloth/feature-checks.json b/scripts/unsloth/feature-checks.json index 62ce39fe9417..5b9f90e3bd84 100644 --- a/scripts/unsloth/feature-checks.json +++ b/scripts/unsloth/feature-checks.json @@ -121,6 +121,12 @@ { "kind": "backend-op", "op": "GET_ROWS", "params": "type=mxfp8" }, { "kind": "mtmd", "projector": "deepseek41v" } ] + }, + "cuda-fa-quant-small-batch": { + "owner": "unslothai#303", + "checks": [ + { "kind": "backend-op", "op": "FLASH_ATTN_EXT", "params": "n_seq=4" } + ] } }, "unchecked": { diff --git a/scripts/unsloth/pr-set.json b/scripts/unsloth/pr-set.json index 8ecb7227a109..b7140a5299a4 100644 --- a/scripts/unsloth/pr-set.json +++ b/scripts/unsloth/pr-set.json @@ -37,6 +37,7 @@ "https://github.com/unslothai/llama.cpp/pull/266/commits/88035b0f74d0fd076df2012a957950fee0c4615f", "https://github.com/unslothai/llama.cpp/pull/283/commits/644ca7d817b8a591c61dee2e17778b43b1ecf3eb", "https://github.com/unslothai/llama.cpp/pull/287/commits/22fb037b1b84513e141d724bb9403ecb48925e32", - "https://github.com/unslothai/llama.cpp/pull/298/commits/9d907b7cfe0e41b03272e952a57b933c9fa08757" + "https://github.com/unslothai/llama.cpp/pull/298/commits/9d907b7cfe0e41b03272e952a57b933c9fa08757", + "https://github.com/unslothai/llama.cpp/pull/303/commits/e886a61c55320ce59e53b72489ad6e051e864ce8" ] } From 0f4a294c2b31f691932c9a630290fc287b5d062d Mon Sep 17 00:00:00 2001 From: Daniel Han Date: Wed, 21 Oct 2026 08:43:54 -0700 Subject: [PATCH 2/2] pr-set: pin #303 at 34a66f42b (composes after ggml-org#25731), repin ggml-org#25731 to e1b89cfa0 --- scripts/unsloth/pr-set.json | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/scripts/unsloth/pr-set.json b/scripts/unsloth/pr-set.json index b7140a5299a4..64bfc85c7500 100644 --- a/scripts/unsloth/pr-set.json +++ b/scripts/unsloth/pr-set.json @@ -19,7 +19,7 @@ ], "prs": [ "https://github.com/ggml-org/llama.cpp/pull/24423/commits/dd7f73b19deb9e79dd441904cf0158af459cd034", - "https://github.com/ggml-org/llama.cpp/pull/25731/commits/534c32182487ee3277908a91dd02f6d8e8ac7ed4", + "https://github.com/ggml-org/llama.cpp/pull/25731/commits/e1b89cfa035866efdc2857318f438609584395f5", "https://github.com/unslothai/llama.cpp/pull/70/commits/883f2c9ba78f3847148454adf025da29385fff3e", "https://github.com/unslothai/llama.cpp/pull/61/commits/05a66933fc06f6b2eda2288caf506e09f400ed2b", "https://github.com/unslothai/llama.cpp/pull/95/commits/496a8a646f19b244b5e2eaf4cf64f920ed31c1a3", @@ -38,6 +38,6 @@ "https://github.com/unslothai/llama.cpp/pull/283/commits/644ca7d817b8a591c61dee2e17778b43b1ecf3eb", "https://github.com/unslothai/llama.cpp/pull/287/commits/22fb037b1b84513e141d724bb9403ecb48925e32", "https://github.com/unslothai/llama.cpp/pull/298/commits/9d907b7cfe0e41b03272e952a57b933c9fa08757", - "https://github.com/unslothai/llama.cpp/pull/303/commits/e886a61c55320ce59e53b72489ad6e051e864ce8" + "https://github.com/unslothai/llama.cpp/pull/303/commits/34a66f42be7b0624b00e039725ea879fd772b14f" ] }