]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commitdiff
common : fix --n-cpu-moe, --cpu-moe for models with fused gate + up (#20416)
authorddh0 <redacted>
Wed, 11 Mar 2026 23:13:28 +0000 (18:13 -0500)
committerGitHub <redacted>
Wed, 11 Mar 2026 23:13:28 +0000 (00:13 +0100)
common/common.h

index ffaeefd7c9423617787b83f0fdfedad5e6f1fb46..c5645bba460e821f2753f362ff42aef919d632a4 100644 (file)
@@ -926,7 +926,7 @@ const char * const LLM_KV_SPLIT_TENSORS_COUNT = "split.tensors.count";
 // MoE utils
 //
 
-const char * const LLM_FFN_EXPS_REGEX = "\\.ffn_(up|down|gate)_(ch|)exps";
+const char * const LLM_FFN_EXPS_REGEX = "\\.ffn_(up|down|gate|gate_up)_(ch|)exps";
 
 inline std::string llm_ffn_exps_block_regex(int idx) {
     return string_format("blk\\.%d%s", idx, LLM_FFN_EXPS_REGEX);