add patches

llama/patches/02-default-pretokenizer.diff

@@ -0,0 +1,32 @@
+diff --git a/llama.cpp b/llama.cpp
+index 40d2ec2c..74f3ee9c 100644
+--- a/llama.cpp
++++ b/llama.cpp
+@@ -4642,16 +4642,7 @@ static void llm_load_vocab(
+ 
+         // for now, only BPE models have pre-tokenizers
+         if (vocab.type == LLAMA_VOCAB_TYPE_BPE) {
+-            if (tokenizer_pre.empty()) {
+-                LLAMA_LOG_WARN("%s: missing pre-tokenizer type, using: 'default'\n", __func__);
+-                LLAMA_LOG_WARN("%s:                                             \n", __func__);
+-                LLAMA_LOG_WARN("%s: ************************************        \n", __func__);
+-                LLAMA_LOG_WARN("%s: GENERATION QUALITY WILL BE DEGRADED!        \n", __func__);
+-                LLAMA_LOG_WARN("%s: CONSIDER REGENERATING THE MODEL             \n", __func__);
+-                LLAMA_LOG_WARN("%s: ************************************        \n", __func__);
+-                LLAMA_LOG_WARN("%s:                                             \n", __func__);
+-                vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEFAULT;
+-            } else if (
++            if (
+                     tokenizer_pre == "default") {
+                 vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEFAULT;
+             } else if (
+@@ -4703,7 +4694,8 @@ static void llm_load_vocab(
+                 tokenizer_pre == "smaug-bpe") {
+                 vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_SMAUG;
+             } 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__);
++                vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEFAULT;
+             }
+         } else {
+             vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEFAULT;

llama/patches/03-metal.diff

@@ -0,0 +1,45 @@
+diff --git a/ggml-metal.m b/ggml-metal.m
+index 0207b787..b5e9884b 100644
+--- a/ggml-metal.m
++++ b/ggml-metal.m
+@@ -1396,27 +1396,23 @@ static enum ggml_status ggml_metal_graph_compute(
+                         // to the matrix-vector kernel
+                         int ne11_mm_min = 1;
+ 
+-#if 0
+                         // the numbers below are measured on M2 Ultra for 7B and 13B models
+                         // these numbers do not translate to other devices or model sizes
+                         // TODO: need to find a better approach
+-                        if ([ctx->device.name isEqualToString:@"Apple M2 Ultra"]) {
+-                            switch (src0t) {
+-                                case GGML_TYPE_F16:  ne11_mm_min = 2;  break;
+-                                case GGML_TYPE_Q8_0: ne11_mm_min = 7;  break;
+-                                case GGML_TYPE_Q2_K: ne11_mm_min = 15; break;
+-                                case GGML_TYPE_Q3_K: ne11_mm_min = 7;  break;
+-                                case GGML_TYPE_Q4_0:
+-                                case GGML_TYPE_Q4_1: ne11_mm_min = 15; break;
+-                                case GGML_TYPE_Q4_K: ne11_mm_min = 11; break;
+-                                case GGML_TYPE_Q5_0:                          // not tested yet
+-                                case GGML_TYPE_Q5_1: ne11_mm_min = 13; break; // not tested yet
+-                                case GGML_TYPE_Q5_K: ne11_mm_min = 7;  break;
+-                                case GGML_TYPE_Q6_K: ne11_mm_min = 7;  break;
+-                                default:             ne11_mm_min = 1;  break;
+-                            }
++                        switch (src0t) {
++                            case GGML_TYPE_F16:  ne11_mm_min = 2;  break;
++                            case GGML_TYPE_Q8_0: ne11_mm_min = 7;  break;
++                            case GGML_TYPE_Q2_K: ne11_mm_min = 15; break;
++                            case GGML_TYPE_Q3_K: ne11_mm_min = 7;  break;
++                            case GGML_TYPE_Q4_0:
++                            case GGML_TYPE_Q4_1: ne11_mm_min = 15; break;
++                            case GGML_TYPE_Q4_K: ne11_mm_min = 11; break;
++                            case GGML_TYPE_Q5_0:                          // not tested yet
++                            case GGML_TYPE_Q5_1: ne11_mm_min = 13; break; // not tested yet
++                            case GGML_TYPE_Q5_K: ne11_mm_min = 7;  break;
++                            case GGML_TYPE_Q6_K: ne11_mm_min = 7;  break;
++                            default:             ne11_mm_min = 1;  break;
+                         }
+-#endif
+ 
+                         // for now the matrix-matrix multiplication kernel only works on A14+/M1+ SoCs
+                         // AMD GPU and older A-chips will reuse matrix-vector multiplication kernel

llama/patches/04-qwen2.diff

@@ -0,0 +1,13 @@
+diff --git a/llama.cpp b/llama.cpp
+index 40d2ec2c..f34eb79a 100644
+--- a/llama.cpp
++++ b/llama.cpp
+@@ -6943,7 +6943,7 @@ static struct ggml_tensor * llm_build_kqv(
+         struct ggml_tensor * kq = ggml_mul_mat(ctx, k, q);
+         cb(kq, "kq", il);
+ 
+-        if (model.arch == LLM_ARCH_PHI2 || model.arch == LLM_ARCH_PHI3 || model.arch == LLM_ARCH_GPTNEOX) {
++        if (model.arch == LLM_ARCH_PHI2 || model.arch == LLM_ARCH_PHI3 || model.arch == LLM_ARCH_GPTNEOX || model.arch == LLM_ARCH_QWEN2) {
+             // for this arch, we need to perform the KQ multiplication with F32 precision, otherwise we get NaNs
+             // ref: https://github.com/ggerganov/llama.cpp/pull/4490#issuecomment-1859055847
+             ggml_mul_mat_set_prec(kq, GGML_PREC_F32);