update llama.cpp submodule to 6eeaeba1 (#6039)

llm/ext_server/server.cpp

@@ -2438,15 +2438,6 @@ static void server_params_parse(int argc, char **argv, server_params &sparams, g
             params.lora_adapter.emplace_back(lora_adapter, std::stof(argv[i]));
             params.use_mmap = false;
         }
-        else if (arg == "--lora-base")
-        {
-            if (++i >= argc)
-            {
-                invalid_param = true;
-                break;
-            }
-            params.lora_base = argv[i];
-        }
         else if (arg == "-v" || arg == "--verbose")
         {
             server_verbose = true;

llm/llama.cpp

@@ -1 +1 @@
-Subproject commit d94c6e0ccbd29ee1ba4f44e9caa8682ad94df9fa
+Subproject commit 6eeaeba126ff701f3e8f79f246805b7023709972

llm/patches/05-default-pretokenizer.diff

@@ -1,8 +1,8 @@
 diff --git a/src/llama.cpp b/src/llama.cpp
-index 8fe51971..7113ba64 100644
+index a207451f..2ddf431d 100644
 --- a/src/llama.cpp
 +++ b/src/llama.cpp
-@@ -5433,16 +5433,7 @@ static void llm_load_vocab(
+@@ -5347,16 +5347,7 @@ static void llm_load_vocab(
          if (vocab.type == LLAMA_VOCAB_TYPE_BPE) {
              vocab.tokenizer_add_space_prefix = false;
              vocab.tokenizer_clean_spaces = true;
@@ -20,9 +20,9 @@ index 8fe51971..7113ba64 100644
                  vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEFAULT;
              } else if (
                      tokenizer_pre == "llama3"   ||
-@@ -5526,7 +5517,8 @@ static void llm_load_vocab(
+@@ -5443,7 +5434,8 @@ static void llm_load_vocab(
-                 vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_SMOLLM;
+                 tokenizer_pre == "codeshell") {
-                 vocab.tokenizer_clean_spaces = false;
+                 vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_CODESHELL;
              } else {
 -                throw std::runtime_error(format("unknown pre-tokenizer type: '%s'", tokenizer_pre.c_str()));
 +                LLAMA_LOG_WARN("%s: missing or unrecognized pre-tokenizer type, using: 'default'\n", __func__);

llm/patches/09-lora.diff

@@ -2,7 +2,7 @@ diff --git a/common/common.cpp b/common/common.cpp
 index dbb724fb..c26fe6ee 100644
 --- a/common/common.cpp
 +++ b/common/common.cpp
-@@ -2087,14 +2087,29 @@ std::tuple<struct llama_model *, struct llama_context *> llama_init_from_gpt_par
+@@ -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]);
@@ -20,9 +20,7 @@ index dbb724fb..c26fe6ee 100644
 +            int err = llama_model_apply_lora_from_file(model,
 +                                                    lora_adapter.c_str(),
 +                                                    lora_scale,
-+                                                    ((i > 0) || params.lora_base.empty())
++                                                    nullptr,
-+                                                        ? NULL
-+                                                        : params.lora_base.c_str(),
 +                                                    params.n_threads);
 +            if (err != 0) {
 +                fprintf(stderr, "%s: error: failed to apply lora adapter\n", __func__);

llm/patches/10-llama3-rope.diff

@@ -1,70 +0,0 @@
-From 2f872f294fb6f5c6e8f983b68c40ea656053dd92 Mon Sep 17 00:00:00 2001
-From: Michael Yang <mxyng@pm.me>
-Date: Tue, 23 Jul 2024 14:33:29 -0700
-Subject: [PATCH] llama 3.1 rope scaling
-
----
- src/llama.cpp | 14 ++++++++++++--
- 1 file changed, 12 insertions(+), 2 deletions(-)
-
-diff --git a/src/llama.cpp b/src/llama.cpp
-index 8fe51971..a9969df8 100644
---- a/src/llama.cpp
-+++ b/src/llama.cpp
-@@ -2472,6 +2472,7 @@ struct llama_layer {
-     // long rope factors
-     struct ggml_tensor * rope_long  = nullptr;
-     struct ggml_tensor * rope_short = nullptr;
-+    struct ggml_tensor * rope_freqs = nullptr;
- 
-     // bitnet scale
-     struct ggml_tensor * wq_scale;
-@@ -6143,6 +6144,8 @@ static bool llm_load_tensors(
- 
-                         layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd});
- 
-+                        layer.rope_freqs  = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ROPE_FREQS,  "weight"), { n_embd/n_head/2 }, llama_model_loader::TENSOR_NOT_REQUIRED | (i != 0 ? llama_model_loader::TENSOR_DUPLICATED : 0));
-+
-                         if (n_expert == 0) {
-                             layer.ffn_gate = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff});
-                             layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd});
-@@ -8620,6 +8623,10 @@ struct llm_build_context {
-         // choose long/short freq factors based on the context size
-         const auto n_ctx_pre_seq = cparams.n_ctx / cparams.n_seq_max;
- 
-+        if (model.layers[il].rope_freqs != nullptr) {
-+            return model.layers[il].rope_freqs;
-+        }
-+
-         if (n_ctx_pre_seq > hparams.n_ctx_orig_yarn) {
-             return model.layers[il].rope_long;
-         }
-@@ -8814,6 +8821,9 @@ struct llm_build_context {
- 
-             // self-attention
-             {
-+                // rope freq factors for llama3; may return nullptr for llama2 and other models
-+                struct ggml_tensor * rope_factors = build_rope_factors(il);
-+
-                 // compute Q and K and RoPE them
-                 struct ggml_tensor * Qcur = llm_build_lora_mm(lctx, ctx0, model.layers[il].wq, cur);
-                 cb(Qcur, "Qcur", il);
-@@ -8837,14 +8847,14 @@ struct llm_build_context {
-                 }
- 
-                 Qcur = ggml_rope_ext(
--                    ctx0, ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens), inp_pos, nullptr,
-+                    ctx0, ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens), inp_pos, rope_factors,
-                     n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
-                     ext_factor, attn_factor, beta_fast, beta_slow
-                 );
-                 cb(Qcur, "Qcur", il);
- 
-                 Kcur = ggml_rope_ext(
--                    ctx0, ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens), inp_pos, nullptr,
-+                    ctx0, ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens), inp_pos, rope_factors,
-                     n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
-                     ext_factor, attn_factor, beta_fast, beta_slow
-                 );
--- 
-2.45.2