update llama.cpp submodule to 1e6f6554 (#6208)

llm/ext_server/server.cpp

@@ -403,7 +403,9 @@ struct llama_server_context
             }
         }
 
-        std::tie(model, ctx) = llama_init_from_gpt_params(params);
+        auto init_result = llama_init_from_gpt_params(params);
+        model = init_result.model;
+        ctx = init_result.context;
         if (model == nullptr)
         {
             LOG_ERROR("unable to load model", {{"model", params.model}});
@@ -2422,7 +2424,10 @@ static void server_params_parse(int argc, char **argv, server_params &sparams, g
                 invalid_param = true;
                 break;
             }
-            params.lora_adapter.emplace_back(argv[i], 1.0f);
+            params.lora_adapters.push_back({
+                std::string(argv[i]),
+                1.0,
+            });
             params.use_mmap = false;
         }
         else if (arg == "--lora-scaled")
@@ -2438,7 +2443,10 @@ static void server_params_parse(int argc, char **argv, server_params &sparams, g
                 invalid_param = true;
                 break;
             }
-            params.lora_adapter.emplace_back(lora_adapter, std::stof(argv[i]));
+            params.lora_adapters.push_back({
+                lora_adapter,
+                std::stof(argv[i])
+            });
             params.use_mmap = false;
         }
         else if (arg == "-v" || arg == "--verbose")

llm/llama.cpp

@@ -1 +1 @@
-Subproject commit 6eeaeba126ff701f3e8f79f246805b7023709972
+Subproject commit 1e6f6554aa11fa10160a5fda689e736c3c34169f

llm/patches/09-lora.diff

@@ -1,40 +1,32 @@
 diff --git a/common/common.cpp b/common/common.cpp
-index dbb724fb..c26fe6ee 100644
+index 2e8374d5..70d0afde 100644
 --- a/common/common.cpp
 +++ b/common/common.cpp
-@@ -2087,14 +2087,27 @@ std::tuple<struct llama_model *, struct llama_context *> llama_init_from_gpt_par
-     for (unsigned int i = 0; i < params.lora_adapter.size(); ++i) {
-         const std::string & lora_adapter = std::get<0>(params.lora_adapter[i]);
-         float lora_scale = std::get<1>(params.lora_adapter[i]);
-+
-+        // try to load as gguf
-         auto adapter = llama_lora_adapter_init(model, lora_adapter.c_str());
-         if (adapter == nullptr) {
--            fprintf(stderr, "%s: error: failed to apply lora adapter\n", __func__);
+@@ -2110,9 +2110,21 @@ struct llama_init_result llama_init_from_gpt_params(gpt_params & params) {
+         loaded_la.adapter = llama_lora_adapter_init(model, la.path.c_str());
+         if (loaded_la.adapter == nullptr) {
+             fprintf(stderr, "%s: error: failed to apply lora adapter '%s'\n", __func__, la.path.c_str());
 -            llama_free(lctx);
 -            llama_free_model(model);
--            return std::make_tuple(nullptr, nullptr);
-+            fprintf(stderr, "%s: error: failed to apply lora adapter, trying ggla\n", __func__);
+-            return iparams;
 +
 +            // if that fails, try loading as ggla for compatibility
 +            int err = llama_model_apply_lora_from_file(model,
-+                                                    lora_adapter.c_str(),
-+                                                    lora_scale,
++                                                    la.path.c_str(),
++                                                    la.scale,
 +                                                    nullptr,
 +                                                    params.n_threads);
 +            if (err != 0) {
 +                fprintf(stderr, "%s: error: failed to apply lora adapter\n", __func__);
 +                llama_free(lctx);
 +                llama_free_model(model);
-+                return std::make_tuple(nullptr, nullptr);
++                return iparams;
++            } else {
++                break;
 +            }
-+        } else {
-+            llama_lora_adapter_set(lctx, adapter, lora_scale);
          }
--        llama_lora_adapter_set(lctx, adapter, lora_scale);
+         iparams.lora_adapters.push_back(loaded_la); // copy to list of loaded adapters
      }
- 
-     if (params.ignore_eos) {
 diff --git a/include/llama.h b/include/llama.h
 index 93fd77ca..b0fb37a6 100644
 --- a/include/llama.h
@@ -355,4 +347,4 @@ index 80a0dd0f..9d7b0e17 100644
 +        return 1;
 +    }
 +}
-\ No newline at end of file
+\ No newline at end of file

llm/patches/10-params.diff

@@ -1,20 +0,0 @@
-diff --git a/src/llama.cpp b/src/llama.cpp
-index a207451f..fba6b175 100644
---- a/src/llama.cpp
-+++ b/src/llama.cpp
-@@ -4969,6 +4969,7 @@ static void llm_load_hparams(
-                 hparams.attn_soft_cap = true;
- 
-                 switch (hparams.n_layer) {
-+                    case 26: model.type = e_model::MODEL_2B; break;
-                     case 42: model.type = e_model::MODEL_9B; break;
-                     case 46: model.type = e_model::MODEL_27B; break;
-                     default: model.type = e_model::MODEL_UNKNOWN;
-@@ -11736,6 +11737,7 @@ struct llm_build_context {
- 
-                 // ref: https://github.com/google/gemma_pytorch/commit/03e657582d17cb5a8617ebf333c1c16f3694670e
-                 switch (model.type) {
-+                    case e_model::MODEL_2B: Qcur = ggml_scale(ctx0, Qcur, 1.0f / sqrtf(float(n_embd_head_k))); break;
-                     case e_model::MODEL_9B:  Qcur = ggml_scale(ctx0, Qcur, 1.0f / sqrtf(float(n_embd_head_k)));   break;
-                     case e_model::MODEL_27B: Qcur = ggml_scale(ctx0, Qcur, 1.0f / sqrtf(float(n_embd / n_head))); break;
-                     default: GGML_ABORT("fatal error");