From: Aman Gupta <redacted>
Date: Tue, 24 Mar 2026 04:57:57 +0000 (+0800)
Subject: llama-fit: fix regex pattern for gate_up tensors (#20910)
X-Git-Tag: upstream/0.0.8611~114
X-Git-Url: https://git.djapps.eu/?a=commitdiff_plain;h=e852eb490136dfa335586b9dd6efb94b02c8fdd3;p=pkg%2Fggml%2Fsources%2Fllama.cpp

llama-fit: fix regex pattern for gate_up tensors (#20910)

* llama-fit: fix regex pattern for gate_up tensors

* Apply suggestions from code review

Co-authored-by: Johannes Gäßler <redacted>

---------

Co-authored-by: Johannes Gäßler <redacted>
---

diff --git a/src/llama.cpp b/src/llama.cpp
index 872e659ed..4a8a71b08 100644
--- a/src/llama.cpp
+++ b/src/llama.cpp
@@ -365,14 +365,14 @@ static void llama_params_fit_impl(
             case LAYER_FRACTION_ATTN: {
                 static std::array<std::string, n_strings> patterns;
                 if (patterns[il].empty()) {
-                    patterns[il] = "blk\\." + std::to_string(il) + "\\.ffn_(up|gate|down).*";
+                    patterns[il] = "blk\\." + std::to_string(il) + "\\.ffn_(gate|up|gate_up|down).*";
                 }
                 return patterns[il].c_str();
             }
             case LAYER_FRACTION_UP: {
                 static std::array<std::string, n_strings> patterns;
                 if (patterns[il].empty()) {
-                    patterns[il] = "blk\\." + std::to_string(il) + "\\.ffn_(gate|down).*";
+                    patterns[il] = "blk\\." + std::to_string(il) + "\\.ffn_(gate|gate_up|down).*";
                 }
                 return patterns[il].c_str();
             }
@@ -386,7 +386,7 @@ static void llama_params_fit_impl(
             case LAYER_FRACTION_MOE: {
                 static std::array<std::string, n_strings> patterns;
                 if (patterns[il].empty()) {
-                    patterns[il] = "blk\\." + std::to_string(il) + "\\.ffn_(up|down|gate)_(ch|)exps";
+                    patterns[il] = "blk\\." + std::to_string(il) + "\\.ffn_(up|down|gate_up|gate)_(ch|)exps";
                 }
                 return patterns[il].c_str();
             }
@@ -480,7 +480,7 @@ static void llama_params_fit_impl(
 
     int64_t global_surplus_cpu_moe = 0;
     if (hp_nex > 0) {
-        const static std::string pattern_moe_all = "blk\\.\\d+\\.ffn_(up|down|gate)_(ch|)exps"; // matches all MoE tensors
+        const static std::string pattern_moe_all = "blk\\.\\d+\\.ffn_(up|down|gate_up|gate)_(ch|)exps"; // matches all MoE tensors
         ggml_backend_buffer_type_t cpu_buft = ggml_backend_cpu_buffer_type();
         tensor_buft_overrides[0] = {pattern_moe_all.c_str(), cpu_buft};
         tensor_buft_overrides[1] = {nullptr, nullptr};