llava : support MiniCPM-V-2.6 (#8967)

* init

* rename

* add run android for termux in readme

* add android readme

* add instructions in readme

* change name in readme

* Update README.md

* fixed line

* add result in readme

* random pos_embed

* add positions index

* change for ollama

* change for ollama

* better pos_embed in clip

* support ollama

* updata cmakelist

* updata cmakelist

* rename wrapper

* clear code

* replace and organize code

* add link

* sync master

* fix warnings

* fix warnings

* fix bug in bicubic resize when need resize iamge smaller

* receive review comments and modify

* receive review comments and modify

* put all code into llava dir

* fix quality problem in pr code

* change n_layer

* add space in "-1"

* imitate reshape bug of python code

* fix bug in clip

* fix issues for merging

* fix llama-minicpmv-cli in cmake file

* change pr readme

* fix code review

* remove in line 33 directory in the /cmakelists.txt (not in example, in the main dir

* fix cmakefile

* add warn

* fix KEY_HAS_MINICPMV_PROJ

* remove load_image_size into clip_ctx

* remove the extern "C", MINICPMV_API

* fix uhd code for review comment

* delete minicpmv-wrapper in pr

* remove uhd_image_embed

* Modify 2 notes

* support minicpmv2.6

* modify convert script of minicpmv

* modify convert

* modify convert

* add readme

* add resampler of v2.6

* modify clip

* modify readme

* fix type-check

* fix type-check

* fix type-check

* fix type-check

* modify convert script and readme

* fix convert script and readme

* fix convert

* fix num in convert

* fix type-check

---------

Co-authored-by: Hongji Zhu <fireyoucan@gmail.com>
Co-authored-by: harvestingmoon <leewenyeong@gmail.com>

.github/workflows/bench.yml.disabled

@@ -1,3 +1,6 @@
+# TODO: there have been some issues with the workflow, so disabling for now
+#       https://github.com/ggerganov/llama.cpp/issues/7893
+#
 # Benchmark
 name: Benchmark
 

Makefile

@@ -763,6 +763,10 @@ ifdef GGML_VULKAN_MEMORY_DEBUG
 	MK_CPPFLAGS  += -DGGML_VULKAN_MEMORY_DEBUG
 endif
 
+ifdef GGML_VULKAN_PERF
+	MK_CPPFLAGS  += -DGGML_VULKAN_PERF
+endif
+
 ifdef GGML_VULKAN_VALIDATE
 	MK_CPPFLAGS  += -DGGML_VULKAN_VALIDATE
 endif

README.md

@@ -105,6 +105,7 @@ Typically finetunes of the base models below are supported as well.
 - [x] [Open Elm models](https://huggingface.co/collections/apple/openelm-instruct-models-6619ad295d7ae9f868b759ca)
 - [x] [ChatGLM3-6b](https://huggingface.co/THUDM/chatglm3-6b) + [ChatGLM4-9b](https://huggingface.co/THUDM/glm-4-9b)
 - [x] [SmolLM](https://huggingface.co/collections/HuggingFaceTB/smollm-6695016cad7167254ce15966)
+- [x] [EXAONE-3.0-7.8B-Instruct](https://huggingface.co/LGAI-EXAONE/EXAONE-3.0-7.8B-Instruct)
 
 (instructions for supporting more models: [HOWTO-add-model.md](./docs/development/HOWTO-add-model.md))
 

common/common.cpp

@@ -110,8 +110,34 @@ int32_t cpu_get_num_physical_cores() {
     if (result == 0) {
         return num_physical_cores;
     }
-#elif defined(_WIN32)
-    //TODO: Implement
+#elif defined(_WIN32) && (_WIN32_WINNT >= 0x0601) && !defined(__MINGW64__) // windows 7 and later
+    // TODO: windows + arm64 + mingw64
+    unsigned int n_threads_win = std::thread::hardware_concurrency();
+    unsigned int default_threads = n_threads_win > 0 ? (n_threads_win <= 4 ? n_threads_win : n_threads_win / 2) : 4;
+
+    DWORD buffer_size = 0;
+    if (!GetLogicalProcessorInformationEx(RelationProcessorCore, nullptr, &buffer_size)) {
+        if (GetLastError() != ERROR_INSUFFICIENT_BUFFER) {
+            return default_threads;
+        }
+    }
+
+    std::vector<char> buffer(buffer_size);
+    if (!GetLogicalProcessorInformationEx(RelationProcessorCore, reinterpret_cast<PSYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX>(buffer.data()), &buffer_size)) {
+        return default_threads;
+    }
+
+    int32_t num_physical_cores = 0;
+    PSYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX info = reinterpret_cast<PSYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX>(buffer.data());
+    while (buffer_size > 0) {
+        if (info->Relationship == RelationProcessorCore) {
+            num_physical_cores += info->Processor.GroupCount;
+        }
+        buffer_size -= info->Size;
+        info = reinterpret_cast<PSYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX>(reinterpret_cast<char*>(info) + info->Size);
+    }
+
+    return num_physical_cores > 0 ? num_physical_cores : default_threads;
 #endif
     unsigned int n_threads = std::thread::hardware_concurrency();
     return n_threads > 0 ? (n_threads <= 4 ? n_threads : n_threads / 2) : 4;
@@ -1727,7 +1753,13 @@ std::string gpt_params_get_system_info(const gpt_params & params) {
     if (params.n_threads_batch != -1) {
         os << " (n_threads_batch = " << params.n_threads_batch << ")";
     }
+#if defined(_WIN32) && (_WIN32_WINNT >= 0x0601) && !defined(__MINGW64__) // windows 7 and later
+    // TODO: windows + arm64 + mingw64
+    DWORD logicalProcessorCount = GetActiveProcessorCount(ALL_PROCESSOR_GROUPS);
+    os << " / " << logicalProcessorCount << " | " << llama_print_system_info();
+#else
     os << " / " << std::thread::hardware_concurrency() << " | " << llama_print_system_info();
+#endif
 
     return os.str();
 }
@@ -2702,12 +2734,6 @@ std::string llama_detokenize(llama_context * ctx, const std::vector<llama_token>
     return text;
 }
 
-bool llama_should_add_bos_token(const llama_model * model) {
-    const int add_bos = llama_add_bos_token(model);
-
-    return add_bos != -1 ? bool(add_bos) : (llama_vocab_type(model) == LLAMA_VOCAB_TYPE_SPM);
-}
-
 //
 // Chat template utils
 //

common/common.h

@@ -380,10 +380,6 @@ std::string llama_detokenize(
         const std::vector<llama_token> & tokens,
                                   bool   special = true);
 
-// Uses the value from the model metadata if possible, otherwise
-// defaults to true when model type is SPM, otherwise false.
-bool llama_should_add_bos_token(const llama_model * model);
-
 //
 // Chat template utils
 //

convert_hf_to_gguf.py

@@ -590,6 +590,15 @@ class Model:
         if chkhsh == "855059429035d75a914d1eda9f10a876752e281a054a7a3d421ef0533e5b6249":
             # ref: https://huggingface.co/HuggingFaceTB/SmolLM-135M
             res = "smollm"
+        if chkhsh == "3c30d3ad1d6b64202cd222813e7736c2db6e1bd6d67197090fc1211fbc612ae7":
+            # ref: https://huggingface.co/bigscience/bloom
+            res = "bloom"
+        if chkhsh == "bc01ce58980e1db43859146dc51b1758b3b88729b217a74792e9f8d43e479d21":
+            # ref: https://huggingface.co/TurkuNLP/gpt3-finnish-small
+            res = "gpt3-finnish"
+        if chkhsh == "4e2b24cc4770243d65a2c9ec19770a72f08cffc161adbb73fcbb6b7dd45a0aae":
+            # ref: https://huggingface.co/LGAI-EXAONE/EXAONE-3.0-7.8B-Instruct
+            res = "exaone"
 
         if res is None:
             logger.warning("\n")
@@ -893,7 +902,7 @@ class GPTNeoXModel(Model):
         return tensors
 
 
-@Model.register("BloomForCausalLM")
+@Model.register("BloomForCausalLM", "BloomModel")
 class BloomModel(Model):
     model_arch = gguf.MODEL_ARCH.BLOOM
 
@@ -3734,6 +3743,118 @@ class ChatGLMModel(Model):
         name = name.removeprefix("transformer.")
         return [(self.map_tensor_name(name), data_torch)]
 
+
+@Model.register("NemotronForCausalLM")
+class NemotronModel(Model):
+    model_arch = gguf.MODEL_ARCH.NEMOTRON
+
+    def set_vocab(self):
+        self._set_vocab_sentencepiece()
+        self.gguf_writer.add_pad_token_id(0)
+        self.gguf_writer.add_unk_token_id(1)
+
+    def set_gguf_parameters(self):
+        super().set_gguf_parameters()
+        hparams = self.hparams
+        self.gguf_writer.add_vocab_size(hparams["vocab_size"])
+
+        f_norm_eps = self.find_hparam(["layer_norm_eps", "layer_norm_epsilon", "norm_epsilon", "norm_eps"])
+        self.gguf_writer.add_layer_norm_eps(f_norm_eps)
+
+        # * Partial RoPE
+        rot_pct = self.find_hparam(["partial_rotary_factor", "rope_pct", "rope_percent"])
+        n_embd = self.find_hparam(["hidden_size", "n_embd"])
+        n_head = self.find_hparam(["num_attention_heads", "n_head"])
+        self.gguf_writer.add_rope_dimension_count(int(rot_pct * n_embd) // n_head)
+
+        # * RopeScaling for Nemotron
+        if "rope_scaling" not in self.hparams or self.hparams["rope_scaling"] is None:
+            self.gguf_writer.add_rope_scaling_type(gguf.RopeScalingType.NONE)
+        else:
+            self.gguf_writer.add_rope_scaling_type(gguf.RopeScalingType.LINEAR)
+            self.gguf_writer.add_rope_scaling_factor(self.hparams["factor"])
+
+    def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iterable[tuple[str, Tensor]]:
+        # * Adding +1 to LayerNorm's weights here to implement layernorm1p w/o changing anything on the GGML engine side
+        #   model.layers.{l}.input_layernorm.weight
+        #   model.layers.{l}.post_attention_layernorm.weight
+        #   model.norm.weight
+        if name.endswith("norm.weight"):
+            data_torch = data_torch + 1
+
+        return [(self.map_tensor_name(name), data_torch)]
+
+
+@Model.register("ExaoneForCausalLM")
+class ExaoneModel(Model):
+    model_arch = gguf.MODEL_ARCH.EXAONE
+
+    def set_gguf_parameters(self):
+        hparams = self.hparams
+
+        assert(hparams["activation_function"] == "silu")
+
+        max_position_embeddings = hparams["max_position_embeddings"]
+        embed_dim = hparams["hidden_size"]
+        num_heads = hparams["num_attention_heads"]
+        num_kv_heads = hparams.get("num_key_value_heads", num_heads)
+        layer_norm_eps = hparams["layer_norm_epsilon"]
+        intermediate_size = hparams["intermediate_size"] if "intermediate_size" in hparams else 4 * embed_dim
+        num_layers = hparams["num_layers"]
+        # ignore for now as EXAONE-3.0-7.8B-Instruct attentino_dropout is 0.0
+        # attention_dropout_rate = hparams["attention_dropout"]
+        # ignore for now as EXAONE-3.0-7.8B-Instruct embed_dropout is 0.0
+        # embed_dropout_rate = hparams["embed_dropout"]
+        self.gguf_writer.add_embedding_length(embed_dim)
+        self.gguf_writer.add_head_count(num_heads)
+        self.gguf_writer.add_head_count_kv(num_kv_heads)
+        self.gguf_writer.add_context_length(max_position_embeddings)
+        self.gguf_writer.add_layer_norm_rms_eps(layer_norm_eps)
+        self.gguf_writer.add_feed_forward_length(intermediate_size)
+        self.gguf_writer.add_block_count(num_layers)
+        self.gguf_writer.add_file_type(self.ftype)
+
+        if (rope_theta := self.hparams.get("rope_theta")) is not None:
+            self.gguf_writer.add_rope_freq_base(rope_theta)
+        rotary_factor = self.find_hparam(["partial_rotary_factor", "rope_pct"], optional=True)
+        rotary_factor = rotary_factor if rotary_factor is not None else 1.0
+        self.gguf_writer.add_rope_dimension_count(int(rotary_factor * (hparams["hidden_size"] // hparams["num_attention_heads"])))
+        if hparams.get("rope_scaling") is not None and "factor" in hparams["rope_scaling"]:
+            if hparams["rope_scaling"].get("type") == "linear":
+                self.gguf_writer.add_rope_scaling_type(gguf.RopeScalingType.LINEAR)
+                self.gguf_writer.add_rope_scaling_factor(hparams["rope_scaling"]["factor"])
+
+    def prepare_tensors(self):
+        if rope_scaling := self.find_hparam(["rope_scaling"], optional=True):
+            if rope_scaling.get("rope_type", '').lower() == "llama3":
+                base = self.hparams.get("rope_theta", 10000.0)
+                dim = self.hparams["hidden_size"] // self.hparams["num_attention_heads"]
+                freqs = 1.0 / (base ** (torch.arange(0, dim, 2, dtype=torch.float32) / dim))
+
+                factor = rope_scaling.get("factor", 8.0)
+                low_freq_factor = rope_scaling.get("low_freq_factor", 1.0)
+                high_freq_factor = rope_scaling.get("high_freq_factor", 4.0)
+                old_context_len = self.hparams.get("original_max_position_embeddings", 8192)
+
+                low_freq_wavelen = old_context_len / low_freq_factor
+                high_freq_wavelen = old_context_len / high_freq_factor
+                assert low_freq_wavelen != high_freq_wavelen
+
+                rope_factors = []
+                for freq in freqs:
+                    wavelen = 2 * math.pi / freq
+                    if wavelen < high_freq_wavelen:
+                        rope_factors.append(1)
+                    elif wavelen > low_freq_wavelen:
+                        rope_factors.append(factor)
+                    else:
+                        smooth = (old_context_len / wavelen - low_freq_factor) / (high_freq_factor - low_freq_factor)
+                        rope_factors.append(1 / ((1 - smooth) / factor + smooth))
+
+                self.gguf_writer.add_tensor(self.format_tensor_name(gguf.MODEL_TENSOR.ROPE_FREQS), np.array(rope_factors, dtype=np.float32))
+
+        super().prepare_tensors()
+
 ###### CONVERSION LOGIC ######
 
 

convert_hf_to_gguf_update.py

@@ -94,6 +94,9 @@ models = [
     {"name": "codeshell",      "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/WisdomShell/CodeShell-7B", },
     {"name": "tekken",         "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/mistralai/Mistral-Nemo-Base-2407", },
     {"name": "smollm",         "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/HuggingFaceTB/SmolLM-135M", },
+    {'name': "bloom",          "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/bigscience/bloom", },
+    {'name': "gpt3-finnish",   "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/TurkuNLP/gpt3-finnish-small", },
+    {"name": "exaone",         "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/LGAI-EXAONE/EXAONE-3.0-7.8B-Instruct", },
 ]
 
 

convert_llama_ggml_to_gguf.py

@@ -116,7 +116,7 @@ class Tensor:
         assert quant is not None, 'Unknown tensor type'
         (blksize, tysize) = quant
         offset += 12
-        self.dtype= dtype
+        self.dtype= gguf.GGMLQuantizationType(dtype)
         self.dims = struct.unpack(f'<{n_dims}I', data[offset:offset + (4 * n_dims)])
         offset += 4 * n_dims
         self.name = bytes(data[offset:offset + name_len])

examples/cvector-generator/cvector-generator.cpp

@@ -271,7 +271,7 @@ struct tokenized_prompt {
     size_t max_seq_len;
 
     tokenized_prompt(llama_context * ctx, std::string pos, std::string neg) {
-        const bool add_bos = llama_should_add_bos_token(llama_get_model(ctx));
+        const bool add_bos = llama_add_bos_token(llama_get_model(ctx));
         tokens_pos = ::llama_tokenize(ctx, pos, add_bos, true);
         tokens_neg = ::llama_tokenize(ctx, neg, add_bos, true);
         max_seq_len = std::max(tokens_pos.size(), tokens_neg.size());

examples/eval-callback/eval-callback.cpp

@@ -127,7 +127,7 @@ static bool ggml_debug(struct ggml_tensor * t, bool ask, void * user_data) {
 }
 
 static bool run(llama_context * ctx, const gpt_params & params) {
-    const bool add_bos = llama_should_add_bos_token(llama_get_model(ctx));
+    const bool add_bos = llama_add_bos_token(llama_get_model(ctx));
 
     std::vector<llama_token> tokens = ::llama_tokenize(ctx, params.prompt, add_bos);
 

examples/export-lora/README.md

@@ -17,9 +17,9 @@ For example:
 
 ```bash
 ./bin/llama-export-lora \
-    -m open-llama-3b-v2-q8_0.gguf \
-    -o open-llama-3b-v2-q8_0-english2tokipona-chat.gguf \
-    --lora lora-open-llama-3b-v2-q8_0-english2tokipona-chat-LATEST.gguf
+    -m open-llama-3b-v2.gguf \
+    -o open-llama-3b-v2-english2tokipona-chat.gguf \
+    --lora lora-open-llama-3b-v2-english2tokipona-chat-LATEST.gguf
 ```
 
 Multiple LORA adapters can be applied by passing multiple `--lora FNAME` or `--lora-scaled FNAME S` command line parameters:

examples/export-lora/export-lora.cpp

@@ -10,6 +10,12 @@
 
 static bool g_verbose = false;
 
+struct tensor_transformation {
+    struct ggml_tensor * in;
+    struct ggml_tensor * out;
+    bool is_copy;
+};
+
 static std::string get_kv_str(struct gguf_context * ctx_gguf, const std::string & key){
     int id = gguf_find_key(ctx_gguf, key.c_str());
     return id < 0 ? "" : std::string(gguf_get_val_str(ctx_gguf, id));
@@ -198,8 +204,7 @@ struct lora_merge_ctx {
         }
 
         // mapping base tensor to out tensor (same shape with base, but different type)
-        // if out_tensor == nullptr, we only copy it
-        std::vector<std::pair<struct ggml_tensor *, struct ggml_tensor *>> base_to_out_tensors;
+        std::vector<tensor_transformation> trans;
         for (auto & it : base_model.tensors) {
             bool t_a = true;
             bool t_b = true;
@@ -212,14 +217,22 @@ struct lora_merge_ctx {
                 // only copy
                 struct ggml_tensor * cpy_tensor = ggml_dup_tensor(ctx_out_ggml, base_tensor);
                 ggml_set_name(cpy_tensor, base_tensor->name);
-                base_to_out_tensors.push_back(std::make_pair(cpy_tensor, nullptr));
+                trans.push_back({
+                    cpy_tensor,
+                    cpy_tensor,
+                    true,
+                });
                 gguf_add_tensor(ctx_out, cpy_tensor);
             } else if (t_a && t_b) {
                 // need merging
                 struct ggml_tensor * out_tensor = ggml_new_tensor(
                     ctx_out_ggml, get_out_tensor_type(base_tensor), GGML_MAX_DIMS, base_tensor->ne);
                 ggml_set_name(out_tensor, base_tensor->name);
-                base_to_out_tensors.push_back(std::make_pair(base_tensor, out_tensor));
+                trans.push_back({
+                    base_tensor,
+                    out_tensor,
+                    false,
+                });
                 gguf_add_tensor(ctx_out, out_tensor);
             } else {
                 throw std::runtime_error("tensor " + it.first + " missing either lora_a or lora_b");
@@ -234,12 +247,12 @@ struct lora_merge_ctx {
 
         // process base model tensors
         size_t n_merged = 0;
-        for (auto & it : base_to_out_tensors) {
-            if (it.second != nullptr) {
-                merge_tensor(it.first, it.second);
+        for (auto & it : trans) {
+            if (!it.is_copy) {
+                merge_tensor(it.in, it.out);
                 n_merged++;
             } else {
-                copy_tensor(it.first);
+                copy_tensor(it.in);
             }
         }
 
@@ -252,7 +265,7 @@ struct lora_merge_ctx {
         }
 
         printf("%s : merged %ld tensors with lora adapters\n", __func__, n_merged);
-        printf("%s : wrote %ld tensors to output file\n", __func__, base_to_out_tensors.size());
+        printf("%s : wrote %ld tensors to output file\n", __func__, trans.size());
     }
 
     void copy_tensor(struct ggml_tensor * base) {
@@ -285,6 +298,10 @@ struct lora_merge_ctx {
         for (size_t i = 0; i < adapters.size(); ++i) {
             auto t_a = adapters[i]->get_tensor(name_lora_a);
             auto t_b = adapters[i]->get_tensor(name_lora_b);
+            // TODO: add support for quantized lora
+            if (ggml_is_quantized(t_a->type) || ggml_is_quantized(t_b->type)) {
+                throw std::runtime_error("quantized LoRA adapters is not supported, please retry with f16 or f32");
+            }
             inp_a[i] = ggml_dup_tensor(ctx, t_a);
             inp_b[i] = ggml_dup_tensor(ctx, t_b);
         }

examples/imatrix/imatrix.cpp

@@ -433,8 +433,8 @@ static void process_logits(
 }
 
 static bool compute_imatrix(llama_context * ctx, const gpt_params & params) {
-    const bool add_bos = llama_should_add_bos_token(llama_get_model(ctx));
-    GGML_ASSERT(llama_add_eos_token(llama_get_model(ctx)) != 1);
+    const bool add_bos = llama_add_bos_token(llama_get_model(ctx));
+    GGML_ASSERT(!llama_add_eos_token(llama_get_model(ctx)));
     const int n_ctx = llama_n_ctx(ctx);
 
     auto tim1 = std::chrono::high_resolution_clock::now();

examples/infill/infill.cpp

@@ -203,8 +203,8 @@ int main(int argc, char ** argv) {
         LOG_TEE("\n");
         LOG_TEE("%s\n", gpt_params_get_system_info(params).c_str());
     }
-    const bool add_bos = llama_should_add_bos_token(model);
-    GGML_ASSERT(llama_add_eos_token(model) != 1);
+    const bool add_bos = llama_add_bos_token(model);
+    GGML_ASSERT(!llama_add_eos_token(model));
     LOG("add_bos: %d\n", add_bos);
 
     std::vector<llama_token> embd_inp;

examples/llava/README-minicpmv2.5.md

@@ -16,8 +16,8 @@ Convert PyTorch model to gguf files (You can also download the converted [gguf](
 
 ```bash
 python ./examples/minicpmv/minicpmv-surgery.py -m ../MiniCPM-Llama3-V-2_5
-python ./examples/minicpmv/minicpmv-convert-image-encoder-to-gguf.py -m ../MiniCPM-Llama3-V-2_5 --minicpmv-projector ../MiniCPM-Llama3-V-2_5/minicpmv.projector --output-dir ../MiniCPM-Llama3-V-2_5/ --image-mean 0.5 0.5 0.5 --image-std 0.5 0.5 0.5
-python ./convert-hf-to-gguf.py ../MiniCPM-Llama3-V-2_5/model
+python ./examples/minicpmv/minicpmv-convert-image-encoder-to-gguf.py -m ../MiniCPM-Llama3-V-2_5 --minicpmv-projector ../MiniCPM-Llama3-V-2_5/minicpmv.projector --output-dir ../MiniCPM-Llama3-V-2_5/ --image-mean 0.5 0.5 0.5 --image-std 0.5 0.5 0.5 --minicpmv_version 2
+python ./convert_hf_to_gguf.py ../MiniCPM-Llama3-V-2_5/model
 
 # quantize int4 version
 ./llama-quantize ../MiniCPM-Llama3-V-2_5/model/model-8B-F16.gguf ../MiniCPM-Llama3-V-2_5/model/ggml-model-Q4_K_M.gguf Q4_K_M

examples/llava/README-minicpmv2.6.md

@@ -0,0 +1,107 @@
+## MiniCPM-V 2.6
+
+### Prepare models and code
+
+Download [MiniCPM-V-2_6](https://huggingface.co/openbmb/MiniCPM-V-2_6) PyTorch model from huggingface to "MiniCPM-V-2_6" folder.
+
+Clone llama.cpp:
+```bash
+git clone git@github.com:OpenBMB/llama.cpp.git
+cd llama.cpp
+git checkout minicpmv-main
+```
+
+### Usage of MiniCPM-V 2.6
+
+Convert PyTorch model to gguf files (You can also download the converted [gguf](https://huggingface.co/openbmb/MiniCPM-V-2_6-gguf) by us)
+
+```bash
+python ./examples/llava/minicpmv-surgery.py -m ../MiniCPM-V-2_6
+python ./examples/llava/minicpmv-convert-image-encoder-to-gguf.py -m ../MiniCPM-V-2_6 --minicpmv-projector ../MiniCPM-V-2_6/minicpmv.projector --output-dir ../MiniCPM-V-2_6/ --image-mean 0.5 0.5 0.5 --image-std 0.5 0.5 0.5 --minicpmv_version 3
+python ./convert_hf_to_gguf.py ../MiniCPM-V-2_6/model
+
+# quantize int4 version
+./llama-quantize ../MiniCPM-V-2_6/model/ggml-model-f16.gguf ../MiniCPM-V-2_6/model/ggml-model-Q4_K_M.gguf Q4_K_M
+```
+
+Build for Linux or Mac
+
+```bash
+make
+make llama-minicpmv-cli
+```
+
+Inference on Linux or Mac
+```
+# run f16 version
+./llama-minicpmv-cli -m ../MiniCPM-V-2_6/model/ggml-model-f16.gguf --mmproj ../MiniCPM-V-2_6/mmproj-model-f16.gguf -c 4096 --temp 0.7 --top-p 0.8 --top-k 100 --repeat-penalty 1.05 --image xx.jpg -p "What is in the image?"
+
+# run quantized int4 version
+./llama-minicpmv-cli -m ../MiniCPM-V-2_6/model/ggml-model-Q4_K_M.gguf --mmproj ../MiniCPM-V-2_6/mmproj-model-f16.gguf -c 4096 --temp 0.7 --top-p 0.8 --top-k 100 --repeat-penalty 1.05 --image xx.jpg  -p "What is in the image?"
+
+# or run in interactive mode
+./llama-minicpmv-cli -m ../MiniCPM-V-2_6/model/ggml-model-Q4_K_M.gguf --mmproj ../MiniCPM-V-2_6/mmproj-model-f16.gguf -c 4096 --temp 0.7 --top-p 0.8 --top-k 100 --repeat-penalty 1.05 --image xx.jpg -i
+```
+
+### Video
+Install FFmpeg
+```
+brew install ffmpeg
+brew install pkg-config
+```
+
+### Android
+
+#### Build on Android device using Termux
+We found that build on Android device would bring better runtime performance, so we recommend to build on device.
+
+[Termux](https://github.com/termux/termux-app#installation) is a terminal app on Android device (no root required).
+
+Install tools in Termux:
+```
+apt update && apt upgrade -y
+apt install git make cmake
+```
+
+It's recommended to move your model inside the `~/` directory for best performance:
+```
+cd storage/downloads
+mv model.gguf ~/
+```
+
+#### Building the Project using Android NDK
+Obtain the [Android NDK](https://developer.android.com/ndk) and then build with CMake.
+
+Execute the following commands on your computer to avoid downloading the NDK to your mobile. Alternatively, you can also do this in Termux:
+
+```bash
+mkdir build-android
+cd build-android
+export NDK=/your_ndk_path
+cmake -DCMAKE_TOOLCHAIN_FILE=$NDK/build/cmake/android.toolchain.cmake -DANDROID_ABI=arm64-v8a -DANDROID_PLATFORM=android-23 -DCMAKE_C_FLAGS=-march=armv8.4a+dotprod ..
+make
+```
+
+Install [termux](https://github.com/termux/termux-app#installation) on your device and run `termux-setup-storage` to get access to your SD card (if Android 11+ then run the command twice).
+
+Finally, copy these built `llama` binaries and the model file to your device storage. Because the file permissions in the Android sdcard cannot be changed, you can copy the executable files to the `/data/data/com.termux/files/home/bin` path, and then execute the following commands in Termux to add executable permission:
+
+(Assumed that you have pushed the built executable files to the /sdcard/llama.cpp/bin path using `adb push`)
+```
+$cp -r /sdcard/llama.cpp/bin /data/data/com.termux/files/home/
+$cd /data/data/com.termux/files/home/bin
+$chmod +x ./*
+```
+
+Download models and push them to `/sdcard/llama.cpp/`, then move it to `/data/data/com.termux/files/home/model/`
+
+```
+$mv /sdcard/llama.cpp/ggml-model-Q4_K_M.gguf /data/data/com.termux/files/home/model/
+$mv /sdcard/llama.cpp/mmproj-model-f16.gguf /data/data/com.termux/files/home/model/
+```
+
+Now, you can start chatting:
+```
+$cd /data/data/com.termux/files/home/bin
+$./llama-minicpmv-cli -m ../model/ggml-model-Q4_K_M.gguf --mmproj ../model/mmproj-model-f16.gguf -c 4096 --temp 0.7 --top-p 0.8 --top-k 100 --repeat-penalty 1.05 --image xx.jpg  -p "What is in the image?"
+```

examples/llava/clip.cpp

@@ -81,6 +81,7 @@ static std::string format(const char * fmt, ...) {
 #define KEY_HAS_VIS_ENC         "clip.has_vision_encoder"
 #define KEY_HAS_LLAVA_PROJ      "clip.has_llava_projector"
 #define KEY_HAS_MINICPMV_PROJ   "clip.has_minicpmv_projector"
+#define KEY_MINICPMV_VERSION    "clip.minicpmv_version"
 #define KEY_USE_GELU            "clip.use_gelu"
 #define KEY_N_EMBD              "clip.%s.embedding_length"
 #define KEY_N_FF                "clip.%s.feed_forward_length"
@@ -526,6 +527,7 @@ struct clip_ctx {
     bool has_vision_encoder  = false;
     bool has_llava_projector = false;
     bool has_minicpmv_projector = false;
+    int minicpmv_version = 2;
 
     struct clip_vision_model vision_model;
     projector_type proj_type = PROJECTOR_TYPE_MLP;
@@ -641,7 +643,12 @@ static ggml_cgraph * clip_image_build_graph(clip_ctx * ctx, const clip_image_f32
     if (ctx->has_minicpmv_projector) {
         int pos_w = image_size_width/patch_size;
         int pos_h = image_size_height/patch_size;
-        pos_embed = ggml_new_tensor_3d(ctx0, GGML_TYPE_F32, 4096, pos_w * pos_h, 1);
+        if (ctx->minicpmv_version == 2) {
+            pos_embed = ggml_new_tensor_3d(ctx0, GGML_TYPE_F32, 4096, pos_w * pos_h, 1);
+        }
+        else if (ctx->minicpmv_version == 3) {
+            pos_embed = ggml_new_tensor_3d(ctx0, GGML_TYPE_F32, 3584, pos_w * pos_h, 1);
+        }
         ggml_set_name(pos_embed, "pos_embed");
         ggml_set_input(pos_embed);
     }
@@ -768,8 +775,8 @@ static ggml_cgraph * clip_image_build_graph(clip_ctx * ctx, const clip_image_f32
             embeddings = ggml_gelu(ctx0, embeddings);
             embeddings = ggml_mul_mat(ctx0, model.mm_2_w, embeddings);
             embeddings = ggml_add(ctx0, embeddings, model.mm_2_b);
-
-        } else if (ctx->proj_type == PROJECTOR_TYPE_MLP_NORM) {
+        }
+        else if (ctx->proj_type == PROJECTOR_TYPE_MLP_NORM) {
             embeddings = ggml_mul_mat(ctx0, model.mm_0_w, embeddings);
             embeddings = ggml_add(ctx0, embeddings, model.mm_0_b);
             // ggml_tensor_printf(embeddings, "mm_0_w",0,true,false);
@@ -949,10 +956,20 @@ static ggml_cgraph * clip_image_build_graph(clip_ctx * ctx, const clip_image_f32
             }
 
             { // attention
-                const int hidden_size = 4096;
+                int hidden_size = 4096;
                 const int d_head = 128;
-                const int n_head = hidden_size/d_head;
-                const int num_query = 96;
+                int n_head = hidden_size/d_head;
+                int num_query = 96;
+                if (ctx->minicpmv_version == 2) {
+                    hidden_size = 4096;
+                    n_head = hidden_size/d_head;
+                    num_query = 96;
+                }
+                else if (ctx->minicpmv_version == 3) {
+                    hidden_size = 3584;
+                    n_head = hidden_size/d_head;
+                    num_query = 64;
+                }
 
                 struct ggml_tensor * Q = ggml_add(ctx0, ggml_mul_mat(ctx0, model.mm_model_attn_q_w, q), model.mm_model_attn_q_b);
                 Q = ggml_scale_inplace(ctx0, Q, 1.0f / sqrt((float)d_head));
@@ -1149,6 +1166,11 @@ struct clip_ctx * clip_model_load(const char * fname, const int verbosity = 1) {
             new_clip->has_minicpmv_projector = gguf_get_val_bool(ctx, idx);
         }
 
+        idx = gguf_find_key(ctx, KEY_MINICPMV_VERSION);
+        if (idx != -1) {
+            new_clip->minicpmv_version = gguf_get_val_i32(ctx, idx);
+        }
+
         // GGML_ASSERT(new_clip->has_llava_projector); // see monatis/clip.cpp for image and/or text encoding for semantic search
 
         GGML_ASSERT(new_clip->has_vision_encoder);
@@ -1910,10 +1932,12 @@ int clip_uhd_num_image_embeds_col(struct clip_ctx * ctx_clip) {
 // returns the normalized float tensor for llava-1.5, for spatial_unpad with anyres processing for llava-1.6 it returns the normalized image patch tensors as a vector
 // res_imgs memory is being allocated here, previous allocations will be freed if found
 bool clip_image_preprocess(struct clip_ctx * ctx, const clip_image_u8 * img, clip_image_f32_batch * res_imgs) {
-    if (clip_is_minicpmv(ctx)) {
-        std::vector<std::vector<clip_image_u8 *>> imgs = uhd_slice_image(img);
+
+    if(clip_is_minicpmv(ctx)){
+        int max_slice_nums = 9;
+        std::vector<std::vector<clip_image_u8 *>> imgs = uhd_slice_image(img, max_slice_nums);
         res_imgs->size = 0;
-        for (size_t i = 0; i < imgs.size(); ++i) {
+        for (size_t i = 0; i < imgs.size(); ++i){
             res_imgs->size += imgs[i].size();
         }
         res_imgs->data = new clip_image_f32[res_imgs->size];
@@ -2146,7 +2170,12 @@ int clip_n_patches(const struct clip_ctx * ctx) {
     if (ctx->proj_type == PROJECTOR_TYPE_LDP || ctx->proj_type == PROJECTOR_TYPE_LDPV2) {
         n_patches /= 4;
     } else if (ctx->proj_type == PROJECTOR_TYPE_RESAMPLER) {
-        n_patches = 96;
+        if (ctx->minicpmv_version == 2) {
+            n_patches = 96;
+        }
+        else if (ctx->minicpmv_version == 3) {
+            n_patches = 64;
+        }
     }
 
     return n_patches;
@@ -2282,6 +2311,11 @@ bool clip_image_batch_encode(clip_ctx * ctx, const int n_threads, const clip_ima
     const int patch_size    = hparams.patch_size;
     const int num_patches   = ((image_size_width / patch_size) * (image_size_height / patch_size));
     const int num_positions = num_patches + (ctx->has_class_embedding ? 1 : 0);
+    if(ctx->load_image_size==nullptr){
+        ctx->load_image_size= clip_image_size_init();
+    }
+    const int pos_w = ctx->load_image_size->width/patch_size;
+    const int pos_h = ctx->load_image_size->height/patch_size;
 
     {
         struct ggml_tensor * inp_raw = ggml_graph_get_tensor(gf, "inp_raw");
@@ -2316,8 +2350,18 @@ bool clip_image_batch_encode(clip_ctx * ctx, const int n_threads, const clip_ima
             //    -> https://huggingface.co/HuggingFaceM4/siglip-so400m-14-980-flash-attn2-navit/blob/d66538faeba44480d0bfaa42145eef26f9423199/modeling_siglip.py#L316
             struct ggml_tensor * positions = ggml_graph_get_tensor(gf, "positions");
             int* positions_data = (int*)malloc(ggml_nbytes(positions));
-            for (int i = 0; i < num_positions; i++) {
-                positions_data[i] = std::floor(70.0*i/num_positions);
+            int bucket_coords_h[70];
+            int bucket_coords_w[70];
+            for (int i = 0; i < pos_h; i++){
+                bucket_coords_h[i] = std::floor(70.0*i/pos_h);
+            }
+            for (int i = 0; i < pos_w; i++){
+                bucket_coords_w[i] = std::floor(70.0*i/pos_w);
+            }
+            for (int i = 0, id = 0; i < pos_h; i++){
+                for (int j = 0; j < pos_w; j++){
+                    positions_data[id++] = bucket_coords_h[i]*70 + bucket_coords_w[j];
+                }
             }
             ggml_backend_tensor_set(positions, positions_data, 0, ggml_nbytes(positions));
             free(positions_data);
@@ -2328,12 +2372,13 @@ bool clip_image_batch_encode(clip_ctx * ctx, const int n_threads, const clip_ima
             //    -> https://huggingface.co/Qwen/Qwen-VL/tree/main
             //    -> https://huggingface.co/Qwen/Qwen-VL/blob/0547ed36a86561e2e42fecec8fd0c4f6953e33c4/visual.py#L23
             struct ggml_tensor * pos_embed = ggml_graph_get_tensor(gf, "pos_embed");
-            if(ctx->load_image_size==nullptr){
-                ctx->load_image_size= clip_image_size_init();
-            }
-            int pos_w = ctx->load_image_size->width/patch_size;
-            int pos_h = ctx->load_image_size->height/patch_size;
             int embed_dim = 4096;
+            if (ctx->minicpmv_version == 2) {
+                embed_dim = 4096;
+            }
+            else if (ctx->minicpmv_version == 3) {
+                embed_dim = 3584;
+            }
             auto pos_embed_t = get_2d_sincos_pos_embed(embed_dim, std::make_pair(pos_w, pos_h));
 
             float * pos_embed_data = (float *)malloc(ggml_nbytes(pos_embed));
@@ -2346,7 +2391,8 @@ bool clip_image_batch_encode(clip_ctx * ctx, const int n_threads, const clip_ima
             ggml_backend_tensor_set(pos_embed, pos_embed_data, 0, ggml_nbytes(pos_embed));
             free(pos_embed_data);
         }
-    } else {
+    }
+    else{
         {
             if (ctx->has_class_embedding) {
                 struct ggml_tensor * embeddings = ggml_graph_get_tensor(gf, "embeddings");
@@ -2548,13 +2594,21 @@ int clip_n_mmproj_embd(const struct clip_ctx * ctx) {
         return ctx->vision_model.mm_3_b->ne[0];
     }
     if (ctx->proj_type == PROJECTOR_TYPE_RESAMPLER) {
-        return 4096;
+        if (ctx->minicpmv_version == 2) {
+            return 4096;
+        }
+        else if (ctx->minicpmv_version == 3) {
+            return 3584;
+        }
     }
 
     std::string proj_type = PROJECTOR_TYPE_NAMES[ctx->proj_type];
     throw std::runtime_error(format("%s: don't support projector with: %s currently\n", __func__, proj_type.c_str()));
 }
 
-bool clip_is_minicpmv(const struct clip_ctx * ctx) {
-    return ctx->has_minicpmv_projector;
+int clip_is_minicpmv(const struct clip_ctx * ctx) {
+    if (ctx->has_minicpmv_projector) {
+        return ctx->minicpmv_version;
+    }
+    return 0;
 }

examples/llava/clip.h

@@ -85,7 +85,7 @@ CLIP_API bool clip_image_batch_encode(struct clip_ctx * ctx, int n_threads, cons
 
 CLIP_API bool clip_model_quantize(const char * fname_inp, const char * fname_out, int itype);
 
-CLIP_API bool clip_is_minicpmv(const struct clip_ctx * ctx);
+CLIP_API int clip_is_minicpmv(const struct clip_ctx * ctx);
 
 #ifdef __cplusplus
 }

examples/llava/llava.cpp

@@ -256,7 +256,14 @@ static bool encode_image_with_clip(clip_ctx * ctx_clip, int n_threads, const cli
             load_image_size->width = img_res_v.data[i].nx;
             load_image_size->height = img_res_v.data[i].ny;
             clip_add_load_image_size(ctx_clip, load_image_size);
-            const bool encoded = clip_image_encode(ctx_clip, n_threads, only_v2_5_reshape_by_patch(&img_res_v.data[i], patch_size), image_embd_v[i]);
+            bool encoded = false;
+            int has_minicpmv_projector = clip_is_minicpmv(ctx_clip);
+            if (has_minicpmv_projector == 2) {
+                encoded = clip_image_encode(ctx_clip, n_threads, only_v2_5_reshape_by_patch(&img_res_v.data[i], patch_size), image_embd_v[i]);
+            }
+            else if (has_minicpmv_projector == 3) {
+                encoded = clip_image_encode(ctx_clip, n_threads, &img_res_v.data[i], image_embd_v[i]);
+            }
             if (!encoded) {
                 LOG_TEE("Unable to encode image - spatial_unpad - subimage %d of %d\n", (int) i+1, (int) img_res_v.size);
                 return false;

examples/llava/minicpmv-cli.cpp

@@ -134,7 +134,13 @@ static void process_image(struct llava_context * ctx_llava, struct llava_image_e
     std::string system_prompt;
     int idx = 0;
     int num_image_embeds = embeds->n_image_pos / clip_n_patches(ctx_llava->ctx_clip);
-    system_prompt = "<|begin_of_text|><|start_header_id|>user<|end_header_id|>\n\n";
+    int has_minicpmv_projector = clip_is_minicpmv(ctx_llava->ctx_clip);
+    if (has_minicpmv_projector == 2) {
+        system_prompt = "<|begin_of_text|><|start_header_id|>user<|end_header_id|>\n\n";
+    }
+    else if (has_minicpmv_projector == 3) {
+        system_prompt = "<|im_start|>user\n";
+    }
     LOG_TEE("%s: image token past: %d\n", __func__, n_past);
     eval_string(ctx_llava->ctx_llama, (system_prompt+"<image>").c_str(), params->n_batch, &n_past, false);
     process_eval_image_embed(ctx_llava, embeds, params->n_batch, &n_past, idx++);
@@ -210,10 +216,24 @@ static struct llava_context * minicpmv_init(gpt_params * params, const std::stri
 
 static struct llama_sampling_context * llama_init(struct llava_context * ctx_llava, gpt_params * params, std::string prompt, int &n_past, bool is_first = false){
     std::string user_prompt = prompt;
-    if (!is_first) user_prompt = "<|begin_of_text|><|start_header_id|>user<|end_header_id|>\n\n" + prompt;
+    int has_minicpmv_projector = clip_is_minicpmv(ctx_llava->ctx_clip);
+    if (!is_first) {
+        if (has_minicpmv_projector == 2) {
+            user_prompt = "<|begin_of_text|><|start_header_id|>user<|end_header_id|>\n\n" + prompt;
+        }
+        else if (has_minicpmv_projector == 3) {
+            user_prompt = "<|im_start|>user\n" + prompt;
+        }
+    }
 
     eval_string(ctx_llava->ctx_llama, user_prompt.c_str(), params->n_batch, &n_past, false);
-    eval_string(ctx_llava->ctx_llama, "<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n", params->n_batch, &n_past, false);
+    if (has_minicpmv_projector == 2) {
+        eval_string(ctx_llava->ctx_llama, "<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n", params->n_batch, &n_past, false);
+    }
+    else if (has_minicpmv_projector == 3) {
+        eval_string(ctx_llava->ctx_llama, "<|im_end|><|im_start|>assistant\n", params->n_batch, &n_past, false);
+    }
+
     // generate the response
 
     LOG_TEE("\n");

examples/llava/minicpmv-convert-image-encoder-to-gguf.py

@@ -1,9 +1,416 @@
-import argparse
+# coding=utf-8
+# Copyright 2024 Google AI and The HuggingFace Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+""" PyTorch Siglip model. """
+# Copied from  HuggingFaceM4/siglip-so400m-14-980-flash-attn2-navit and add tgt_sizes
+
+
 import os
+import math
+import warnings
+
+import numpy as np
+import torch
+import torch.nn.functional as F
+import torch.utils.checkpoint
+from torch import nn
+from torch.nn.init import _calculate_fan_in_and_fan_out
+
+from transformers.activations import ACT2FN
+from transformers.modeling_utils import PreTrainedModel
+from transformers.configuration_utils import PretrainedConfig
+from transformers.utils import (
+    logging,
+)
+from transformers.utils import logging
+
+logger = logging.get_logger(__name__)
+
+class SiglipVisionConfig(PretrainedConfig):
+    r"""
+    This is the configuration class to store the configuration of a [`SiglipVisionModel`]. It is used to instantiate a
+    Siglip vision encoder according to the specified arguments, defining the model architecture. Instantiating a
+    configuration with the defaults will yield a similar configuration to that of the vision encoder of the Siglip
+    [google/siglip-base-patch16-224](https://huggingface.co/google/siglip-base-patch16-224) architecture.
+    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
+    documentation from [`PretrainedConfig`] for more information.
+    Args:
+        hidden_size (`int`, *optional*, defaults to 768):
+            Dimensionality of the encoder layers and the pooler layer.
+        intermediate_size (`int`, *optional*, defaults to 3072):
+            Dimensionality of the "intermediate" (i.e., feed-forward) layer in the Transformer encoder.
+        num_hidden_layers (`int`, *optional*, defaults to 12):
+            Number of hidden layers in the Transformer encoder.
+        num_attention_heads (`int`, *optional*, defaults to 12):
+            Number of attention heads for each attention layer in the Transformer encoder.
+        num_channels (`int`, *optional*, defaults to 3):
+            Number of channels in the input images.
+        image_size (`int`, *optional*, defaults to 224):
+            The size (resolution) of each image.
+        patch_size (`int`, *optional*, defaults to 16):
+            The size (resolution) of each patch.
+        hidden_act (`str` or `function`, *optional*, defaults to `"gelu_pytorch_tanh"`):
+            The non-linear activation function (function or string) in the encoder and pooler. If string, `"gelu"`,
+            `"relu"`, `"selu"` and `"gelu_new"` ``"quick_gelu"` are supported.
+        layer_norm_eps (`float`, *optional*, defaults to 1e-06):
+            The epsilon used by the layer normalization layers.
+        attention_dropout (`float`, *optional*, defaults to 0.0):
+            The dropout ratio for the attention probabilities.
+    Example:
+    ```python
+    >>> from transformers import SiglipVisionConfig, SiglipVisionModel
+    >>> # Initializing a SiglipVisionConfig with google/siglip-base-patch16-224 style configuration
+    >>> configuration = SiglipVisionConfig()
+    >>> # Initializing a SiglipVisionModel (with random weights) from the google/siglip-base-patch16-224 style configuration
+    >>> model = SiglipVisionModel(configuration)
+    >>> # Accessing the model configuration
+    >>> configuration = model.config
+    ```"""
+
+    model_type = "siglip_vision_model"
+
+    def __init__(
+        self,
+        hidden_size=768,
+        intermediate_size=3072,
+        num_hidden_layers=12,
+        num_attention_heads=12,
+        num_channels=3,
+        image_size=224,
+        patch_size=16,
+        hidden_act="gelu_pytorch_tanh",
+        layer_norm_eps=1e-6,
+        attention_dropout=0.0,
+        **kwargs,
+    ):
+        super().__init__(**kwargs)
+
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.num_channels = num_channels
+        self.patch_size = patch_size
+        self.image_size = image_size
+        self.attention_dropout = attention_dropout
+        self.layer_norm_eps = layer_norm_eps
+        self.hidden_act = hidden_act
+
+_CHECKPOINT_FOR_DOC = "google/siglip-base-patch16-224"
+
+SIGLIP_PRETRAINED_MODEL_ARCHIVE_LIST = [
+    "google/siglip-base-patch16-224",
+    # See all SigLIP models at https://huggingface.co/models?filter=siglip
+]
+
+# Copied from transformers.models.llama.modeling_llama._get_unpad_data
+def _get_unpad_data(attention_mask):
+    seqlens_in_batch = attention_mask.sum(dim=-1, dtype=torch.int32)
+    indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten()
+    max_seqlen_in_batch = seqlens_in_batch.max().item()
+    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.int32), (1, 0))
+    return (
+        indices,
+        cu_seqlens,
+        max_seqlen_in_batch,
+    )
+
+
+def _trunc_normal_(tensor, mean, std, a, b):
+    # Cut & paste from PyTorch official master until it's in a few official releases - RW
+    # Method based on https://people.sc.fsu.edu/~jburkardt/presentations/truncated_normal.pdf
+    def norm_cdf(x):
+        # Computes standard normal cumulative distribution function
+        return (1.0 + math.erf(x / math.sqrt(2.0))) / 2.0
+
+    if (mean < a - 2 * std) or (mean > b + 2 * std):
+        warnings.warn(
+            "mean is more than 2 std from [a, b] in nn.init.trunc_normal_. "
+            "The distribution of values may be incorrect.",
+            stacklevel=2,
+        )
+
+    # Values are generated by using a truncated uniform distribution and
+    # then using the inverse CDF for the normal distribution.
+    # Get upper and lower cdf values
+    l = norm_cdf((a - mean) / std)
+    u = norm_cdf((b - mean) / std)
+
+    # Uniformly fill tensor with values from [l, u], then translate to
+    # [2l-1, 2u-1].
+    tensor.uniform_(2 * l - 1, 2 * u - 1)
+
+    # Use inverse cdf transform for normal distribution to get truncated
+    # standard normal
+    if tensor.dtype in [torch.float16, torch.bfloat16]:
+        # The `erfinv_` op is not (yet?) defined in float16+cpu, bfloat16+gpu
+        og_dtype = tensor.dtype
+        tensor = tensor.to(torch.float32)
+        tensor.erfinv_()
+        tensor = tensor.to(og_dtype)
+    else:
+        tensor.erfinv_()
+
+    # Transform to proper mean, std
+    tensor.mul_(std * math.sqrt(2.0))
+    tensor.add_(mean)
+
+    # Clamp to ensure it's in the proper range
+    if tensor.dtype == torch.float16:
+        # The `clamp_` op is not (yet?) defined in float16+cpu
+        tensor = tensor.to(torch.float32)
+        tensor.clamp_(min=a, max=b)
+        tensor = tensor.to(torch.float16)
+    else:
+        tensor.clamp_(min=a, max=b)
+
+
+def trunc_normal_tf_(
+    tensor: torch.Tensor, mean: float = 0.0, std: float = 1.0, a: float = -2.0, b: float = 2.0
+):
+    """Fills the input Tensor with values drawn from a truncated
+    normal distribution. The values are effectively drawn from the
+    normal distribution :math:`\\mathcal{N}(\text{mean}, \text{std}^2)`
+    with values outside :math:`[a, b]` redrawn until they are within
+    the bounds. The method used for generating the random values works
+    best when :math:`a \\leq \text{mean} \\leq b`.
+    NOTE: this 'tf' variant behaves closer to Tensorflow / JAX impl where the
+    bounds [a, b] are applied when sampling the normal distribution with mean=0, std=1.0
+    and the result is subsquently scaled and shifted by the mean and std args.
+    Args:
+        tensor: an n-dimensional `torch.Tensor`
+        mean: the mean of the normal distribution
+        std: the standard deviation of the normal distribution
+        a: the minimum cutoff value
+        b: the maximum cutoff value
+    """
+    with torch.no_grad():
+        _trunc_normal_(tensor, 0, 1.0, a, b)
+        tensor.mul_(std).add_(mean)
+
+
+def variance_scaling_(tensor, scale=1.0, mode="fan_in", distribution="normal"):
+    fan_in, fan_out = _calculate_fan_in_and_fan_out(tensor)
+    denom = fan_in
+    if mode == "fan_in":
+        denom = fan_in
+    elif mode == "fan_out":
+        denom = fan_out
+    elif mode == "fan_avg":
+        denom = (fan_in + fan_out) / 2
+
+    variance = scale / denom
+
+    if distribution == "truncated_normal":
+        # constant is stddev of standard normal truncated to (-2, 2)
+        trunc_normal_tf_(tensor, std=math.sqrt(variance) / 0.87962566103423978)
+    elif distribution == "normal":
+        with torch.no_grad():
+            tensor.normal_(std=math.sqrt(variance))
+    elif distribution == "uniform":
+        bound = math.sqrt(3 * variance)
+        with torch.no_grad():
+            tensor.uniform_(-bound, bound)
+    else:
+        raise ValueError(f"invalid distribution {distribution}")
+
+
+def lecun_normal_(tensor):
+    variance_scaling_(tensor, mode="fan_in", distribution="truncated_normal")
+
+
+def default_flax_embed_init(tensor):
+    variance_scaling_(tensor, mode="fan_in", distribution="normal")
+
+class SiglipVisionEmbeddings(nn.Module):
+    def __init__(self, config: SiglipVisionConfig):
+        super().__init__()
+        self.config = config
+        self.embed_dim = config.hidden_size
+        self.image_size = config.image_size
+        self.patch_size = config.patch_size
+
+        self.patch_embedding = nn.Conv2d(
+            in_channels=config.num_channels,
+            out_channels=self.embed_dim,
+            kernel_size=self.patch_size,
+            stride=self.patch_size,
+            padding="valid",
+        )
+
+        self.num_patches_per_side = self.image_size // self.patch_size
+        self.num_patches = self.num_patches_per_side**2
+        self.num_positions = self.num_patches
+        self.position_embedding = nn.Embedding(self.num_positions, self.embed_dim)
+
+class SiglipAttention(nn.Module):
+    """Multi-headed attention from 'Attention Is All You Need' paper"""
+
+    # Copied from transformers.models.clip.modeling_clip.CLIPAttention.__init__
+    def __init__(self, config):
+        super().__init__()
+        self.config = config
+        self.embed_dim = config.hidden_size
+        self.num_heads = config.num_attention_heads
+        self.head_dim = self.embed_dim // self.num_heads
+        if self.head_dim * self.num_heads != self.embed_dim:
+            raise ValueError(
+                f"embed_dim must be divisible by num_heads (got `embed_dim`: {self.embed_dim} and `num_heads`:"
+                f" {self.num_heads})."
+            )
+        self.scale = self.head_dim**-0.5
+        self.dropout = config.attention_dropout
+
+        self.k_proj = nn.Linear(self.embed_dim, self.embed_dim)
+        self.v_proj = nn.Linear(self.embed_dim, self.embed_dim)
+        self.q_proj = nn.Linear(self.embed_dim, self.embed_dim)
+        self.out_proj = nn.Linear(self.embed_dim, self.embed_dim)
+
+# Copied from transformers.models.clip.modeling_clip.CLIPMLP with CLIP->Siglip
+class SiglipMLP(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.config = config
+        self.activation_fn = ACT2FN[config.hidden_act]
+        self.fc1 = nn.Linear(config.hidden_size, config.intermediate_size)
+        self.fc2 = nn.Linear(config.intermediate_size, config.hidden_size)
+
+
+# Copied from transformers.models.clip.modeling_clip.CLIPEncoderLayer with CLIP->Siglip
+class SiglipEncoderLayer(nn.Module):
+    def __init__(self, config: SiglipVisionConfig):
+        super().__init__()
+        self.embed_dim = config.hidden_size
+        self._use_flash_attention_2 = config._attn_implementation == "flash_attention_2"
+        self.self_attn = (
+            SiglipAttention(config)
+        )
+        self.layer_norm1 = nn.LayerNorm(self.embed_dim, eps=config.layer_norm_eps)
+        self.mlp = SiglipMLP(config)
+        self.layer_norm2 = nn.LayerNorm(self.embed_dim, eps=config.layer_norm_eps)
+
+class SiglipPreTrainedModel(PreTrainedModel):
+    """
+    An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained
+    models.
+    """
+
+    config_class = SiglipVisionConfig
+    base_model_prefix = "siglip"
+    supports_gradient_checkpointing = True
+
+    def _init_weights(self, module):
+        """Initialize the weights"""
+
+        if isinstance(module, SiglipVisionEmbeddings):
+            width = self.config.hidden_size
+            nn.init.normal_(module.position_embedding.weight, std=1 / np.sqrt(width))
+        elif isinstance(module, nn.Embedding):
+            default_flax_embed_init(module.weight)
+        elif isinstance(module, SiglipAttention):
+            nn.init.normal_(module.q_proj.weight)
+            nn.init.normal_(module.k_proj.weight)
+            nn.init.normal_(module.v_proj.weight)
+            nn.init.normal_(module.out_proj.weight)
+            nn.init.zeros_(module.q_proj.bias)
+            nn.init.zeros_(module.k_proj.bias)
+            nn.init.zeros_(module.v_proj.bias)
+            nn.init.zeros_(module.out_proj.bias)
+        elif isinstance(module, SiglipMLP):
+            nn.init.normal_(module.fc1.weight)
+            nn.init.normal_(module.fc2.weight)
+            nn.init.normal_(module.fc1.bias, std=1e-6)
+            nn.init.normal_(module.fc2.bias, std=1e-6)
+        elif isinstance(module, (nn.Linear, nn.Conv2d)):
+            lecun_normal_(module.weight)
+            if module.bias is not None:
+                nn.init.zeros_(module.bias)
+        elif isinstance(module, nn.LayerNorm):
+            module.bias.data.zero_()
+            module.weight.data.fill_(1.0)
+
+
+SIGLIP_START_DOCSTRING = r"""
+    This model inherits from [`PreTrainedModel`]. Check the superclass documentation for the generic methods the
+    library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads
+    etc.)
+    This model is also a PyTorch [torch.nn.Module](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) subclass.
+    Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage
+    and behavior.
+    Parameters:
+        config ([`SiglipVisionConfig`]): Model configuration class with all the parameters of the model.
+            Initializing with a config file does not load the weights associated with the model, only the
+            configuration. Check out the [`~PreTrainedModel.from_pretrained`] method to load the model weights.
+"""
+
+
+SIGLIP_VISION_INPUTS_DOCSTRING = r"""
+    Args:
+        pixel_values (`torch.FloatTensor` of shape `(batch_size, num_channels, height, width)`):
+            Pixel values. Padding will be ignored by default should you provide it. Pixel values can be obtained using
+            [`AutoImageProcessor`]. See [`CLIPImageProcessor.__call__`] for details.
+        output_attentions (`bool`, *optional*):
+            Whether or not to return the attentions tensors of all attention layers. See `attentions` under returned
+            tensors for more detail.
+        output_hidden_states (`bool`, *optional*):
+            Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for
+            more detail.
+        return_dict (`bool`, *optional*):
+            Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple.
+"""
+
+
+# Copied from transformers.models.clip.modeling_clip.CLIPEncoder with CLIP->Siglip
+class SiglipEncoder(nn.Module):
+    """
+    Transformer encoder consisting of `config.num_hidden_layers` self attention layers. Each layer is a
+    [`SiglipEncoderLayer`].
+    Args:
+        config: SiglipConfig
+    """
+
+    def __init__(self, config: SiglipVisionConfig):
+        super().__init__()
+        self.config = config
+        self.layers = nn.ModuleList([SiglipEncoderLayer(config) for _ in range(config.num_hidden_layers)])
+        self.gradient_checkpointing = False
+
+class SiglipVisionTransformer(SiglipPreTrainedModel):
+    config_class = SiglipVisionConfig
+    main_input_name = "pixel_values"
+    _supports_flash_attn_2 = True
+
+    def __init__(self, config: SiglipVisionConfig):
+        super().__init__(config)
+        self.config = config
+        embed_dim = config.hidden_size
+
+        self.embeddings = SiglipVisionEmbeddings(config)
+        self.encoder = SiglipEncoder(config)
+        self.post_layernorm = nn.LayerNorm(embed_dim, eps=config.layer_norm_eps)
+        self._use_flash_attention_2 = config._attn_implementation == "flash_attention_2"
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_input_embeddings(self) -> nn.Module:
+        return self.embeddings.patch_embedding
+
+import argparse
 import json
 import re
 
-import torch
 import numpy as np
 from gguf import *
 from transformers.models.idefics2.modeling_idefics2 import Idefics2VisionTransformer, Idefics2VisionConfig
@@ -94,6 +501,7 @@ default_image_mean = [0.48145466, 0.4578275, 0.40821073]
 default_image_std = [0.26862954, 0.26130258, 0.27577711]
 ap.add_argument('--image-mean', type=float, nargs='+', help='Mean of the images for normalization (overrides processor) ', default=None)
 ap.add_argument('--image-std', type=float, nargs='+', help='Standard deviation of the images for normalization (overrides processor)', default=None)
+ap.add_argument('--minicpmv_version', type=int, help='minicpmv_version: MiniCPM-V-2 use 1; MiniCPM-V-2.5 use 2; MiniCPM-V-2.6 use 3', default=2)
 
 # with proper
 args = ap.parse_args()
@@ -135,6 +543,15 @@ if args.use_f32:
 #     model = CLIPModel.from_pretrained(dir_model)
 #     processor = CLIPProcessor.from_pretrained(dir_model)
 
+minicpmv_version = args.minicpmv_version
+emb_dim = 4096
+if minicpmv_version == 1:
+    emb_dim = 2304
+elif minicpmv_version == 2:
+    emb_dim = 4096
+elif minicpmv_version == 3:
+    emb_dim = 3584
+
 default_vision_config = {
         "hidden_size": 1152,
         "image_size": 980,
@@ -144,8 +561,12 @@ default_vision_config = {
         "num_hidden_layers": 27,
         "patch_size": 14,
     }
+
 vision_config = Idefics2VisionConfig(**default_vision_config)
 model = Idefics2VisionTransformer(vision_config)
+if minicpmv_version == 3:
+    vision_config = SiglipVisionConfig(**default_vision_config)
+    model = SiglipVisionTransformer(vision_config)
 
 processor = None
 # if model.attn_pool is not None:
@@ -158,6 +579,7 @@ fname_middle = None
 has_text_encoder = True
 has_vision_encoder = True
 has_minicpmv_projector = False
+
 if args.text_only:
     fname_middle = "text-"
     has_vision_encoder = False
@@ -165,6 +587,7 @@ elif args.minicpmv_projector is not None:
     fname_middle = "mmproj-"
     has_text_encoder = False
     has_minicpmv_projector = True
+    minicpmv_version = 3
 elif args.vision_only:
     fname_middle = "vision-"
     has_text_encoder = False
@@ -189,6 +612,7 @@ elif has_minicpmv_projector:
     fout.add_description("image encoder for MiniCPM-V")
     # add projector type
     fout.add_string("clip.projector_type", "resampler")
+    fout.add_int32("clip.minicpmv_version", minicpmv_version)
 else:
     fout.add_description("two-tower CLIP model")
 
@@ -274,11 +698,11 @@ def _replace_name_resampler(s, v):
     if re.match("resampler.pos_embed", s):
         return {
             s: v,
-            re.sub("pos_embed", "pos_embed_k", s): torch.from_numpy(get_2d_sincos_pos_embed(4096, (70, 70))),
+            re.sub("pos_embed", "pos_embed_k", s): torch.from_numpy(get_2d_sincos_pos_embed(emb_dim, (70, 70))),
         }
     if re.match("resampler.proj", s):
         return {
-            re.sub("proj", "pos_embed_k", s): torch.from_numpy(get_2d_sincos_pos_embed(4096, (70, 70))),
+            re.sub("proj", "pos_embed_k", s): torch.from_numpy(get_2d_sincos_pos_embed(emb_dim, (70, 70))),
             re.sub("proj", "proj.weight", s): v.transpose(-1, -2).contiguous(),
         }
     if re.match("resampler.attn.in_proj_.*", s):

examples/llava/minicpmv-surgery.py

@@ -4,7 +4,7 @@ import torch
 from transformers import AutoModel, AutoTokenizer
 
 ap = argparse.ArgumentParser()
-ap.add_argument("-m", "--model", help="Path to MiniCPM-V-2.5 model")
+ap.add_argument("-m", "--model", help="Path to MiniCPM-V model")
 args = ap.parse_args()
 
 # find the model part that includes the the multimodal projector weights
@@ -29,7 +29,6 @@ if len(clip_tensors) > 0:
             f.write("{}\n")
 
 config = model.llm.config
-config._name_or_path = "openbmb/MiniCPM-Llama3-V-2.5"
 config.auto_map = {
     "AutoConfig": "configuration_minicpm.MiniCPMConfig",
     "AutoModel": "modeling_minicpm.MiniCPMModel",
@@ -40,7 +39,6 @@ config.auto_map = {
 model.llm.save_pretrained(f"{args.model}/model")
 tok = AutoTokenizer.from_pretrained(args.model, trust_remote_code=True)
 tok.save_pretrained(f"{args.model}/model")
-# os.system(f"cp {args.model}/modeling_minicpm.py {args.model}/MiniCPM_l3/modeling_minicpm.py")
 
 print("Done!")
 print(f"Now you can convert {args.model} to a regular LLaMA GGUF file.")

examples/main/main.cpp

@@ -267,9 +267,9 @@ int main(int argc, char ** argv) {
         }
     }
 
-    const bool add_bos = llama_should_add_bos_token(model);
+    const bool add_bos = llama_add_bos_token(model);
     if (!llama_model_has_encoder(model)) {
-        GGML_ASSERT(llama_add_eos_token(model) != 1);
+        GGML_ASSERT(!llama_add_eos_token(model));
     }
     LOG("add_bos: %d\n", add_bos);
 

examples/perplexity/perplexity.cpp

@@ -340,8 +340,8 @@ static results_perplexity perplexity_v2(llama_context * ctx, const gpt_params &
     // Output: `perplexity: 13.5106 [114/114]`
     // BOS tokens will be added for each chunk before eval
 
-    const bool add_bos = llama_should_add_bos_token(llama_get_model(ctx));
-    GGML_ASSERT(llama_add_eos_token(llama_get_model(ctx)) != 1);
+    const bool add_bos = llama_add_bos_token(llama_get_model(ctx));
+    GGML_ASSERT(!llama_add_eos_token(llama_get_model(ctx)));
 
     fprintf(stderr, "%s: tokenizing the input ..\n", __func__);
 
@@ -480,8 +480,8 @@ static results_perplexity perplexity(llama_context * ctx, const gpt_params & par
     // Output: `perplexity: 13.5106 [114/114]`
     // BOS tokens will be added for each chunk before eval
 
-    const bool add_bos = llama_should_add_bos_token(llama_get_model(ctx));
-    GGML_ASSERT(llama_add_eos_token(llama_get_model(ctx)) != 1);
+    const bool add_bos = llama_add_bos_token(llama_get_model(ctx));
+    GGML_ASSERT(!llama_add_eos_token(llama_get_model(ctx)));
 
     std::ofstream logits_stream;
     if (!params.logits_file.empty()) {
@@ -1733,8 +1733,8 @@ static void kl_divergence(llama_context * ctx, const gpt_params & params) {
     const int n_batch = params.n_batch;
     const int num_batches = (n_ctx + n_batch - 1)/n_batch;
     const int nv = 2*((n_vocab + 1)/2) + 4;
-    const bool add_bos = llama_should_add_bos_token(llama_get_model(ctx));
-    GGML_ASSERT(llama_add_eos_token(llama_get_model(ctx)) != 1);
+    const bool add_bos = llama_add_bos_token(llama_get_model(ctx));
+    GGML_ASSERT(!llama_add_eos_token(llama_get_model(ctx)));
 
     std::vector<uint16_t> log_probs_uint16(size_t(n_ctx - 1 - n_ctx/2) * nv);
     std::vector<float>    kld_values(size_t(n_ctx - 1 - n_ctx/2)*n_chunk);

examples/quantize/README.md

@@ -34,7 +34,7 @@ Run the quantized model:
 
 ```bash
 # start inference on a gguf model
-./llama-cli -m ./models/mymodel/ggml-model-Q4_K_M.gguf -n 128
+./llama-cli -m ./models/mymodel/ggml-model-Q4_K_M.gguf -cnv -p "You are a helpful assistant"
 ```
 
 When running the larger models, make sure you have enough disk space to store all the intermediate files.

examples/retrieval/retrieval.cpp

@@ -253,6 +253,8 @@ int main(int argc, char ** argv) {
         chunks[i].tokens.clear();
     }
 
+    struct llama_batch query_batch = llama_batch_init(n_batch, 0, 1);
+
     // start loop, receive query and return top k similar chunks based on cosine similarity
     std::string query;
     while (true) {
@@ -260,7 +262,6 @@ int main(int argc, char ** argv) {
         std::getline(std::cin, query);
         std::vector<int32_t> query_tokens = llama_tokenize(ctx, query, true);
 
-        struct llama_batch query_batch = llama_batch_init(n_batch, 0, 1);
         batch_add_seq(query_batch, query_tokens, 0);
 
         std::vector<float> query_emb(n_embd, 0);
@@ -293,6 +294,7 @@ int main(int argc, char ** argv) {
     }
 
     // clean up
+    llama_batch_free(query_batch);
     llama_print_timings(ctx);
     llama_free(ctx);
     llama_free_model(model);

examples/server/server.cpp

@@ -693,9 +693,8 @@ struct server_context {
 
         n_ctx = llama_n_ctx(ctx);
 
-        add_bos_token = llama_should_add_bos_token(model);
-        has_eos_token = llama_add_eos_token(model) != 1;
-
+        add_bos_token = llama_add_bos_token(model);
+        has_eos_token = !llama_add_eos_token(model);
         return true;
     }
 
@@ -754,13 +753,13 @@ struct server_context {
         default_generation_settings_for_props = get_formated_generation(slots.front());
         default_generation_settings_for_props["seed"] = -1;
 
-        // the update_slots() logic will always submit a maximum of n_batch tokens
+        // the update_slots() logic will always submit a maximum of n_batch or n_parallel tokens
         // note that n_batch can be > n_ctx (e.g. for non-causal attention models such as BERT where the KV cache is not used)
         {
             const int32_t n_batch = llama_n_batch(ctx);
 
             // only a single seq_id per token is needed
-            batch = llama_batch_init(n_batch, 0, 1);
+            batch = llama_batch_init(std::max(n_batch, params.n_parallel), 0, 1);
         }
 
         metrics.init();
@@ -1137,28 +1136,19 @@ struct server_context {
         if (!system_prompt.empty()) {
             system_tokens = ::llama_tokenize(ctx, system_prompt, true);
 
-            llama_batch_clear(batch);
-
-            for (int i = 0; i < (int)system_tokens.size(); ++i) {
-                llama_batch_add(batch, system_tokens[i], i, { 0 }, false);
-            }
-
             const int32_t n_batch = llama_n_batch(ctx);
+            const int32_t n_tokens_prompt = system_tokens.size();
 
-            for (int32_t i = 0; i < batch.n_tokens; i += n_batch) {
-                const int32_t n_tokens = std::min(params.n_batch, batch.n_tokens - i);
-                llama_batch batch_view = {
-                    n_tokens,
-                    batch.token    + i,
-                    nullptr,
-                    batch.pos      + i,
-                    batch.n_seq_id + i,
-                    batch.seq_id   + i,
-                    batch.logits   + i,
-                    0, 0, 0, // unused
-                };
+            for (int32_t i = 0; i < n_tokens_prompt; i += n_batch) {
+                const int32_t n_tokens = std::min(n_batch, n_tokens_prompt - i);
 
-                if (llama_decode(ctx, batch_view) != 0) {
+                llama_batch_clear(batch);
+
+                for (int32_t j = 0; j < n_tokens; ++j) {
+                    llama_batch_add(batch, system_tokens[i + j], i + j, { 0 }, false);
+                }
+
+                if (llama_decode(ctx, batch) != 0) {
                     LOG_ERROR("llama_decode() failed", {});
                     return;
                 }
@@ -1331,7 +1321,7 @@ struct server_context {
 
         return json {
             {"n_ctx",                     slot.n_ctx},
-            {"n_predict",                 slot.n_predict},
+            {"n_predict",                 slot.n_predict},     // Server configured n_predict
             {"model",                     params.model_alias},
             {"seed",                      slot.sparams.seed},
             {"temperature",               slot.sparams.temp},
@@ -1353,7 +1343,7 @@ struct server_context {
             {"mirostat_eta",              slot.sparams.mirostat_eta},
             {"penalize_nl",               slot.sparams.penalize_nl},
             {"stop",                      slot.params.antiprompt},
-            {"n_predict",                 slot.params.n_predict}, // TODO: fix duplicate key n_predict
+            {"max_tokens",                slot.params.n_predict}, // User configured n_predict
             {"n_keep",                    slot.params.n_keep},
             {"n_discard",                 slot.params.n_discard},
             {"ignore_eos",                ignore_eos},
@@ -1861,6 +1851,8 @@ struct server_context {
                     llama_lora_adapters_apply(ctx, lora_adapters);
                     server_task_result result;
                     result.id = task.id;
+                    result.stop = true;
+                    result.error = false;
                     result.data = json{{ "success", true }};
                     queue_results.send(result);
                 } break;
@@ -2045,7 +2037,7 @@ struct server_context {
                         slot.t_start_generation = 0;
 
                         if (slot.infill) {
-                            const bool add_bos = llama_should_add_bos_token(model);
+                            const bool add_bos = llama_add_bos_token(model);
                             bool suff_rm_leading_spc = true;
                             if (params.input_suffix.find_first_of(' ') == 0 && params.input_suffix.size() > 1) {
                                 params.input_suffix.erase(0, 1);

examples/tokenize/tokenize.cpp

@@ -362,7 +362,7 @@ int main(int raw_argc, char ** raw_argv) {
         prompt = stdin_buffer.str();
     }
 
-    const bool model_wants_add_bos = llama_should_add_bos_token(model);
+    const bool model_wants_add_bos = llama_add_bos_token(model);
     const bool add_bos = model_wants_add_bos && !no_bos;
     const bool parse_special = !no_parse_special;
 

ggml/CMakeLists.txt

@@ -129,13 +129,13 @@ option(GGML_CUDA_NO_VMM                     "ggml: do not try to use CUDA VMM"
 option(GGML_CUDA_FA_ALL_QUANTS              "ggml: compile all quants for FlashAttention"     OFF)
 option(GGML_CUDA_USE_GRAPHS                 "ggml: use CUDA graphs (llama.cpp only)"          OFF)
 
-option(GGML_CURL                            "ggml: use libcurl to download model from an URL" OFF)
 option(GGML_HIPBLAS                         "ggml: use hipBLAS"                               OFF)
 option(GGML_HIP_UMA                         "ggml: use HIP unified memory architecture"       OFF)
 option(GGML_VULKAN                          "ggml: use Vulkan"                                OFF)
 option(GGML_VULKAN_CHECK_RESULTS            "ggml: run Vulkan op checks"                      OFF)
 option(GGML_VULKAN_DEBUG                    "ggml: enable Vulkan debug output"                OFF)
 option(GGML_VULKAN_MEMORY_DEBUG             "ggml: enable Vulkan memory debug output"         OFF)
+option(GGML_VULKAN_PERF                     "ggml: enable Vulkan perf output"                 OFF)
 option(GGML_VULKAN_VALIDATE                 "ggml: enable Vulkan validation"                  OFF)
 option(GGML_VULKAN_RUN_TESTS                "ggml: run Vulkan tests"                          OFF)
 option(GGML_KOMPUTE                         "ggml: use Kompute"                               OFF)

ggml/include/ggml.h

@@ -244,6 +244,8 @@
 #define GGML_EXIT_SUCCESS 0
 #define GGML_EXIT_ABORTED 1
 
+#define GGML_ROPE_TYPE_NEOX 2
+
 #define GGUF_MAGIC "GGUF"
 
 #define GGUF_VERSION 3
@@ -1453,8 +1455,8 @@ extern "C" {
             struct ggml_tensor  * b);
 
     // rotary position embedding
-    // if mode & 1 == 1, skip n_past elements (NOT SUPPORTED)
-    // if mode & 2 == 1, GPT-NeoX style
+    // if (mode & 1) - skip n_past elements (NOT SUPPORTED)
+    // if (mode & GGML_ROPE_TYPE_NEOX) - GPT-NeoX style
     //
     // b is an int32 vector with size a->ne[2], it contains the positions
     GGML_API struct ggml_tensor * ggml_rope(

ggml/src/CMakeLists.txt

@@ -602,6 +602,10 @@ if (GGML_VULKAN)
             add_compile_definitions(GGML_VULKAN_MEMORY_DEBUG)
         endif()
 
+        if (GGML_VULKAN_PERF)
+            add_compile_definitions(GGML_VULKAN_PERF)
+        endif()
+
         if (GGML_VULKAN_VALIDATE)
             add_compile_definitions(GGML_VULKAN_VALIDATE)
         endif()

ggml/src/ggml-backend.c

@@ -1018,10 +1018,6 @@ static bool ggml_is_view_op(enum ggml_op op) {
 #define GGML_SCHED_MAX_BACKENDS 16
 #endif
 
-#ifndef GGML_SCHED_MAX_SPLITS
-#define GGML_SCHED_MAX_SPLITS 2048
-#endif
-
 #ifndef GGML_SCHED_MAX_SPLIT_INPUTS
 #define GGML_SCHED_MAX_SPLIT_INPUTS GGML_MAX_SRC
 #endif
@@ -1125,7 +1121,8 @@ static int ggml_backend_sched_backend_from_buffer(ggml_backend_sched_t sched, co
 }
 
 #if 0
-static char causes[GGML_DEFAULT_GRAPH_SIZE*16 + GGML_SCHED_MAX_SPLITS*GGML_SCHED_MAX_SPLIT_INPUTS][128]; // debug only
+#define GGML_SCHED_MAX_SPLITS_DEBUG 4096
+static char causes[GGML_DEFAULT_GRAPH_SIZE*16 + GGML_SCHED_MAX_SPLITS_DEBUG*GGML_SCHED_MAX_SPLIT_INPUTS][128]; // debug only
 #define SET_CAUSE(node, ...) sprintf(causes[hash_id(node)], __VA_ARGS__)
 #define GET_CAUSE(node) causes[hash_id(node)]
 #else
@@ -1549,7 +1546,6 @@ static void ggml_backend_sched_split_graph(ggml_backend_sched_t sched, struct gg
                     sched->splits = realloc(sched->splits, sched->splits_capacity * sizeof(struct ggml_backend_sched_split));
                     GGML_ASSERT(sched->splits != NULL);
                 }
-                GGML_ASSERT(i_split < GGML_SCHED_MAX_SPLITS);
                 split = &sched->splits[i_split];
                 split->backend_id = node_backend_id;
                 split->i_start = i;
@@ -1865,13 +1861,14 @@ ggml_backend_sched_t ggml_backend_sched_new(
     sched->hv_tensor_backend_ids = malloc(sched->hash_set.size * sizeof(sched->hv_tensor_backend_ids[0]));
     sched->hv_tensor_copies      = malloc(sched->hash_set.size * sched->n_backends * sched->n_copies * sizeof(struct ggml_tensor *));
 
-    const size_t nodes_size = graph_size + GGML_SCHED_MAX_SPLITS*GGML_SCHED_MAX_SPLIT_INPUTS*2;
+    const size_t ggml_sched_max_splits = graph_size; // at most there is one split for each node in the graph
+    const size_t nodes_size = graph_size + ggml_sched_max_splits*GGML_SCHED_MAX_SPLIT_INPUTS*2;
     sched->node_backend_ids = calloc(nodes_size, sizeof(sched->node_backend_ids[0]));
     sched->leaf_backend_ids = calloc(nodes_size, sizeof(sched->leaf_backend_ids[0]));
     sched->prev_node_backend_ids = calloc(nodes_size, sizeof(sched->prev_node_backend_ids[0]));
     sched->prev_leaf_backend_ids = calloc(nodes_size, sizeof(sched->prev_leaf_backend_ids[0]));
 
-    sched->context_buffer_size = GGML_SCHED_MAX_SPLITS*GGML_SCHED_MAX_SPLIT_INPUTS*2*sizeof(struct ggml_tensor) + ggml_graph_overhead_custom(graph_size, false);
+    sched->context_buffer_size = ggml_sched_max_splits*GGML_SCHED_MAX_SPLIT_INPUTS*2*sizeof(struct ggml_tensor) + ggml_graph_overhead_custom(graph_size, false);
     sched->context_buffer = malloc(sched->context_buffer_size);
 
     const int initial_splits_capacity = 16;

ggml/src/ggml-cann/aclnn_ops.cpp

@@ -2881,7 +2881,7 @@ void ggml_cann_rope(ggml_backend_cann_context& ctx, ggml_tensor* dst) {
     ggml_rope_yarn_corr_dims(n_dims, n_ctx_orig, freq_base, beta_fast,
                              beta_slow, corr_dims);
 
-    const bool is_neox = mode & 2;
+    const bool is_neox = mode & GGML_ROPE_TYPE_NEOX;
 
     // init cos/sin cache
     ggml_cann_pool_alloc sin_allocator(

ggml/src/ggml-cuda/rope.cu

@@ -226,7 +226,7 @@ void ggml_cuda_op_rope(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
     memcpy(&beta_fast,   (int32_t *) dst->op_params +  9, sizeof(float));
     memcpy(&beta_slow,   (int32_t *) dst->op_params + 10, sizeof(float));
 
-    const bool is_neox = mode & 2;
+    const bool is_neox = mode & GGML_ROPE_TYPE_NEOX;
 
     const int32_t * pos = (const int32_t *) src1_d;
 

ggml/src/ggml-metal.m

@@ -2313,7 +2313,7 @@ static enum ggml_status ggml_metal_graph_compute(
                         memcpy(&beta_fast,   (int32_t *) dst->op_params +  9, sizeof(float));
                         memcpy(&beta_slow,   (int32_t *) dst->op_params + 10, sizeof(float));
 
-                        const bool is_neox = mode & 2;
+                        const bool is_neox = mode & GGML_ROPE_TYPE_NEOX;
 
                         id<MTLComputePipelineState> pipeline = nil;
 

ggml/src/ggml-sycl/rope.cpp

@@ -226,7 +226,7 @@ void ggml_sycl_op_rope(
     memcpy(&beta_fast,   (int32_t *) dst->op_params +  9, sizeof(float));
     memcpy(&beta_slow,   (int32_t *) dst->op_params + 10, sizeof(float));
 
-    const bool is_neox = mode & 2;
+    const bool is_neox = mode & GGML_ROPE_TYPE_NEOX;
 
     const int32_t * pos = (const int32_t *) src1_dd;
 

ggml/src/ggml.c

@@ -14094,7 +14094,7 @@ static void ggml_compute_forward_rope_f32(
     float corr_dims[2];
     ggml_rope_yarn_corr_dims(n_dims, n_ctx_orig, freq_base, beta_fast, beta_slow, corr_dims);
 
-    const bool is_neox = mode & 2;
+    const bool is_neox = mode & GGML_ROPE_TYPE_NEOX;
 
     const float * freq_factors = NULL;
     if (src2 != NULL) {
@@ -14219,7 +14219,7 @@ static void ggml_compute_forward_rope_f16(
     float corr_dims[2];
     ggml_rope_yarn_corr_dims(n_dims, n_ctx_orig, freq_base, beta_fast, beta_slow, corr_dims);
 
-    const bool is_neox = mode & 2;
+    const bool is_neox = mode & GGML_ROPE_TYPE_NEOX;
 
     const float * freq_factors = NULL;
     if (src2 != NULL) {

ggml/src/kompute-shaders/op_rope_f16.comp

@@ -11,7 +11,7 @@ void main() {
     const uint i2 = gl_WorkGroupID.y;
     const uint i1 = gl_WorkGroupID.x;
 
-    const bool is_neox = (pcs.mode & 2) != 0;
+    const bool is_neox = (pcs.mode & GGML_ROPE_TYPE_NEOX) != 0;
 
     float corr_dims[2];
     rope_yarn_corr_dims(pcs.n_dims, pcs.n_ctx_orig, pcs.freq_base, pcs.beta_fast, pcs.beta_slow, corr_dims);

ggml/src/kompute-shaders/op_rope_f32.comp

@@ -11,7 +11,7 @@ void main() {
     const uint i2 = gl_WorkGroupID.y;
     const uint i1 = gl_WorkGroupID.x;
 
-    const bool is_neox = (pcs.mode & 2) != 0;
+    const bool is_neox = (pcs.mode & GGML_ROPE_TYPE_NEOX) != 0;
 
     float corr_dims[2];
     rope_yarn_corr_dims(pcs.n_dims, pcs.n_ctx_orig, pcs.freq_base, pcs.beta_fast, pcs.beta_slow, corr_dims);

ggml/src/kompute-shaders/rope_common.comp

@@ -1,5 +1,7 @@
 #include "common.comp"
 
+#define GGML_ROPE_TYPE_NEOX 2
+
 // TODO: use a local size of 32 or more (Metal uses 1024)
 layout(local_size_x = 1) in;
 

ggml/src/vulkan-shaders/concat.comp

@@ -30,6 +30,10 @@ void main() {
 #ifndef OPTIMIZATION_ERROR_WORKAROUND
     data_d[p.d_offset + dst_idx] = D_TYPE(is_src0 ? data_a[src0_idx] : data_b[src1_idx]);
 #else
-    data_d[p.d_offset + dst_idx] = is_src0 ? data_a[src0_idx] : data_b[src1_idx];
+    if (is_src0) {
+        data_d[p.d_offset + dst_idx] = data_a[src0_idx];
+    } else {
+        data_d[p.d_offset + dst_idx] = data_b[src1_idx];
+    }
 #endif
 }

ggml/src/vulkan-shaders/mul_mat_vec.comp

@@ -39,8 +39,7 @@ void main() {
         vec2 v = dequantize(ib, iqs, a_offset / QUANT_K);
 
         // matrix multiplication
-        tmp[tid] += FLOAT_TYPE(v.x) * FLOAT_TYPE(data_b[b_offset + iybs + iqs]) +
-                    FLOAT_TYPE(v.y) * FLOAT_TYPE(data_b[b_offset + iybs + iqs + y_offset]);
+        tmp[tid] = fma(FLOAT_TYPE(v.x), FLOAT_TYPE(data_b[b_offset + iybs + iqs]), fma(FLOAT_TYPE(v.y), FLOAT_TYPE(data_b[b_offset + iybs + iqs + y_offset]), tmp[tid]));
     }
 
     // sum up partial sums and write back result

ggml/src/vulkan-shaders/mul_mat_vec_nc.comp

@@ -53,7 +53,7 @@ void main() {
 
         const FLOAT_TYPE xi = FLOAT_TYPE(data_a[ix]);
 
-        tmp[tid] += xi * FLOAT_TYPE(data_b[iy]);
+        tmp[tid] = fma(xi, FLOAT_TYPE(data_b[iy]), tmp[tid]);
     }
 
     // sum up partial sums and write back result

ggml/src/vulkan-shaders/mul_mat_vec_p021.comp

@@ -52,7 +52,7 @@ void main() {
         // y is not transposed but permuted
         const uint iy = channel*nrows_y + row_y;
 
-        tmp[tid] += xi * FLOAT_TYPE(data_b[iy]);
+        tmp[tid] = fma(xi, FLOAT_TYPE(data_b[iy]), tmp[tid]);
     }
 
     // dst is not transposed and not permuted

ggml/src/vulkan-shaders/mul_mat_vec_q2_k.comp

@@ -39,24 +39,25 @@ void main() {
         FLOAT_TYPE sum1 = FLOAT_TYPE(0.0);
         FLOAT_TYPE sum2 = FLOAT_TYPE(0.0);
         for (int l = 0; l < K_QUANTS_PER_ITERATION; ++l) {
-            sum1 += FLOAT_TYPE(data_b[b_offset + y_idx + l +  0]) * FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 0] & 0xF) * FLOAT_TYPE((data_a[ib0 + i].qs[q_offset + l + 0] >> 0) & 3)
-                  + FLOAT_TYPE(data_b[b_offset + y_idx + l + 16]) * FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 1] & 0xF) * FLOAT_TYPE((data_a[ib0 + i].qs[q_offset + l +16] >> 0) & 3)
-                  + FLOAT_TYPE(data_b[b_offset + y_idx + l + 32]) * FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 2] & 0xF) * FLOAT_TYPE((data_a[ib0 + i].qs[q_offset + l + 0] >> 2) & 3)
-                  + FLOAT_TYPE(data_b[b_offset + y_idx + l + 48]) * FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 3] & 0xF) * FLOAT_TYPE((data_a[ib0 + i].qs[q_offset + l +16] >> 2) & 3)
-                  + FLOAT_TYPE(data_b[b_offset + y_idx + l + 64]) * FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 4] & 0xF) * FLOAT_TYPE((data_a[ib0 + i].qs[q_offset + l + 0] >> 4) & 3)
-                  + FLOAT_TYPE(data_b[b_offset + y_idx + l + 80]) * FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 5] & 0xF) * FLOAT_TYPE((data_a[ib0 + i].qs[q_offset + l +16] >> 4) & 3)
-                  + FLOAT_TYPE(data_b[b_offset + y_idx + l + 96]) * FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 6] & 0xF) * FLOAT_TYPE((data_a[ib0 + i].qs[q_offset + l + 0] >> 6) & 3)
-                  + FLOAT_TYPE(data_b[b_offset + y_idx + l +112]) * FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 7] & 0xF) * FLOAT_TYPE((data_a[ib0 + i].qs[q_offset + l +16] >> 6) & 3);
-            sum2 += FLOAT_TYPE(data_b[b_offset + y_idx + l +  0]) * FLOAT_TYPE((data_a[ib0 + i].scales[s_offset + 0] >> 4) & 0xF)
-                  + FLOAT_TYPE(data_b[b_offset + y_idx + l + 16]) * FLOAT_TYPE((data_a[ib0 + i].scales[s_offset + 1] >> 4) & 0xF)
-                  + FLOAT_TYPE(data_b[b_offset + y_idx + l + 32]) * FLOAT_TYPE((data_a[ib0 + i].scales[s_offset + 2] >> 4) & 0xF)
-                  + FLOAT_TYPE(data_b[b_offset + y_idx + l + 48]) * FLOAT_TYPE((data_a[ib0 + i].scales[s_offset + 3] >> 4) & 0xF)
-                  + FLOAT_TYPE(data_b[b_offset + y_idx + l + 64]) * FLOAT_TYPE((data_a[ib0 + i].scales[s_offset + 4] >> 4) & 0xF)
-                  + FLOAT_TYPE(data_b[b_offset + y_idx + l + 80]) * FLOAT_TYPE((data_a[ib0 + i].scales[s_offset + 5] >> 4) & 0xF)
-                  + FLOAT_TYPE(data_b[b_offset + y_idx + l + 96]) * FLOAT_TYPE((data_a[ib0 + i].scales[s_offset + 6] >> 4) & 0xF)
-                  + FLOAT_TYPE(data_b[b_offset + y_idx + l +112]) * FLOAT_TYPE((data_a[ib0 + i].scales[s_offset + 7] >> 4) & 0xF);
+            sum1 = fma(FLOAT_TYPE(data_b[b_offset + y_idx + l +  0]), FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 0] & 0xF) * FLOAT_TYPE((data_a[ib0 + i].qs[q_offset + l + 0] >> 0) & 3),
+                   fma(FLOAT_TYPE(data_b[b_offset + y_idx + l + 16]), FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 1] & 0xF) * FLOAT_TYPE((data_a[ib0 + i].qs[q_offset + l +16] >> 0) & 3),
+                   fma(FLOAT_TYPE(data_b[b_offset + y_idx + l + 32]), FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 2] & 0xF) * FLOAT_TYPE((data_a[ib0 + i].qs[q_offset + l + 0] >> 2) & 3),
+                   fma(FLOAT_TYPE(data_b[b_offset + y_idx + l + 48]), FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 3] & 0xF) * FLOAT_TYPE((data_a[ib0 + i].qs[q_offset + l +16] >> 2) & 3),
+                   fma(FLOAT_TYPE(data_b[b_offset + y_idx + l + 64]), FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 4] & 0xF) * FLOAT_TYPE((data_a[ib0 + i].qs[q_offset + l + 0] >> 4) & 3),
+                   fma(FLOAT_TYPE(data_b[b_offset + y_idx + l + 80]), FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 5] & 0xF) * FLOAT_TYPE((data_a[ib0 + i].qs[q_offset + l +16] >> 4) & 3),
+                   fma(FLOAT_TYPE(data_b[b_offset + y_idx + l + 96]), FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 6] & 0xF) * FLOAT_TYPE((data_a[ib0 + i].qs[q_offset + l + 0] >> 6) & 3),
+                   fma(FLOAT_TYPE(data_b[b_offset + y_idx + l +112]), FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 7] & 0xF) * FLOAT_TYPE((data_a[ib0 + i].qs[q_offset + l +16] >> 6) & 3), sum1))))))));
+            sum2 = fma(FLOAT_TYPE(data_b[b_offset + y_idx + l +  0]), FLOAT_TYPE((data_a[ib0 + i].scales[s_offset + 0] >> 4) & 0xF),
+                   fma(FLOAT_TYPE(data_b[b_offset + y_idx + l + 16]), FLOAT_TYPE((data_a[ib0 + i].scales[s_offset + 1] >> 4) & 0xF),
+                   fma(FLOAT_TYPE(data_b[b_offset + y_idx + l + 32]), FLOAT_TYPE((data_a[ib0 + i].scales[s_offset + 2] >> 4) & 0xF),
+                   fma(FLOAT_TYPE(data_b[b_offset + y_idx + l + 48]), FLOAT_TYPE((data_a[ib0 + i].scales[s_offset + 3] >> 4) & 0xF),
+                   fma(FLOAT_TYPE(data_b[b_offset + y_idx + l + 64]), FLOAT_TYPE((data_a[ib0 + i].scales[s_offset + 4] >> 4) & 0xF),
+                   fma(FLOAT_TYPE(data_b[b_offset + y_idx + l + 80]), FLOAT_TYPE((data_a[ib0 + i].scales[s_offset + 5] >> 4) & 0xF),
+                   fma(FLOAT_TYPE(data_b[b_offset + y_idx + l + 96]), FLOAT_TYPE((data_a[ib0 + i].scales[s_offset + 6] >> 4) & 0xF),
+                   fma(FLOAT_TYPE(data_b[b_offset + y_idx + l +112]), FLOAT_TYPE((data_a[ib0 + i].scales[s_offset + 7] >> 4) & 0xF), sum2))))))));
         }
-        tmp[16 * ix + tid] += dall * sum1 - dmin * sum2;
+        const uint tmp_idx = 16 * ix + tid;
+        tmp[tmp_idx] = fma(dall, sum1, fma(-dmin, sum2, tmp[tmp_idx]));
     }
 
     // sum up partial sums and write back result

ggml/src/vulkan-shaders/mul_mat_vec_q3_k.comp

@@ -40,16 +40,17 @@ void main() {
 
         FLOAT_TYPE sum = FLOAT_TYPE(0.0);
         for (int l = 0; l < K_QUANTS_PER_ITERATION; ++l) {
-            sum += FLOAT_TYPE(data_b[b_offset + y_idx + l +  0]) * FLOAT_TYPE(int8_t(((data_a[ib0 + i].scales[0] >> s_shift) & 0xF) | ((data_a[ib0 + i].scales[ 8] >> (s_shift + 0) & 0x3) << 4)) - 32) * FLOAT_TYPE(((data_a[ib0 + i].qs[q_offset + l   ]     ) & 3) - (((data_a[ib0 + i].hmask[l0 + l   ] & (m << 0)) != 0) ? 0 : 4))
-                 + FLOAT_TYPE(data_b[b_offset + y_idx + l + 32]) * FLOAT_TYPE(int8_t(((data_a[ib0 + i].scales[2] >> s_shift) & 0xF) | ((data_a[ib0 + i].scales[10] >> (s_shift + 0) & 0x3) << 4)) - 32) * FLOAT_TYPE(((data_a[ib0 + i].qs[q_offset + l   ] >> 2) & 3) - (((data_a[ib0 + i].hmask[l0 + l   ] & (m << 1)) != 0) ? 0 : 4))
-                 + FLOAT_TYPE(data_b[b_offset + y_idx + l + 64]) * FLOAT_TYPE(int8_t(((data_a[ib0 + i].scales[4] >> s_shift) & 0xF) | ((data_a[ib0 + i].scales[ 8] >> (s_shift + 2) & 0x3) << 4)) - 32) * FLOAT_TYPE(((data_a[ib0 + i].qs[q_offset + l   ] >> 4) & 3) - (((data_a[ib0 + i].hmask[l0 + l   ] & (m << 2)) != 0) ? 0 : 4))
-                 + FLOAT_TYPE(data_b[b_offset + y_idx + l + 96]) * FLOAT_TYPE(int8_t(((data_a[ib0 + i].scales[6] >> s_shift) & 0xF) | ((data_a[ib0 + i].scales[10] >> (s_shift + 2) & 0x3) << 4)) - 32) * FLOAT_TYPE(((data_a[ib0 + i].qs[q_offset + l   ] >> 6) & 3) - (((data_a[ib0 + i].hmask[l0 + l   ] & (m << 3)) != 0) ? 0 : 4))
-                 + FLOAT_TYPE(data_b[b_offset + y_idx + l + 16]) * FLOAT_TYPE(int8_t(((data_a[ib0 + i].scales[1] >> s_shift) & 0xF) | ((data_a[ib0 + i].scales[ 9] >> (s_shift + 0) & 0x3) << 4)) - 32) * FLOAT_TYPE(((data_a[ib0 + i].qs[q_offset + l+16]     ) & 3) - (((data_a[ib0 + i].hmask[l0 + l+16] & (m << 0)) != 0) ? 0 : 4))
-                 + FLOAT_TYPE(data_b[b_offset + y_idx + l + 48]) * FLOAT_TYPE(int8_t(((data_a[ib0 + i].scales[3] >> s_shift) & 0xF) | ((data_a[ib0 + i].scales[11] >> (s_shift + 0) & 0x3) << 4)) - 32) * FLOAT_TYPE(((data_a[ib0 + i].qs[q_offset + l+16] >> 2) & 3) - (((data_a[ib0 + i].hmask[l0 + l+16] & (m << 1)) != 0) ? 0 : 4))
-                 + FLOAT_TYPE(data_b[b_offset + y_idx + l + 80]) * FLOAT_TYPE(int8_t(((data_a[ib0 + i].scales[5] >> s_shift) & 0xF) | ((data_a[ib0 + i].scales[ 9] >> (s_shift + 2) & 0x3) << 4)) - 32) * FLOAT_TYPE(((data_a[ib0 + i].qs[q_offset + l+16] >> 4) & 3) - (((data_a[ib0 + i].hmask[l0 + l+16] & (m << 2)) != 0) ? 0 : 4))
-                 + FLOAT_TYPE(data_b[b_offset + y_idx + l +112]) * FLOAT_TYPE(int8_t(((data_a[ib0 + i].scales[7] >> s_shift) & 0xF) | ((data_a[ib0 + i].scales[11] >> (s_shift + 2) & 0x3) << 4)) - 32) * FLOAT_TYPE(((data_a[ib0 + i].qs[q_offset + l+16] >> 6) & 3) - (((data_a[ib0 + i].hmask[l0 + l+16] & (m << 3)) != 0) ? 0 : 4));
+            sum = fma(FLOAT_TYPE(data_b[b_offset + y_idx + l +  0]) * FLOAT_TYPE(int8_t(((data_a[ib0 + i].scales[0] >> s_shift) & 0xF) | ((data_a[ib0 + i].scales[ 8] >> (s_shift + 0) & 0x3) << 4)) - 32), FLOAT_TYPE(((data_a[ib0 + i].qs[q_offset + l   ]     ) & 3) - (((data_a[ib0 + i].hmask[l0 + l   ] & (m << 0)) != 0) ? 0 : 4)),
+                  fma(FLOAT_TYPE(data_b[b_offset + y_idx + l + 32]) * FLOAT_TYPE(int8_t(((data_a[ib0 + i].scales[2] >> s_shift) & 0xF) | ((data_a[ib0 + i].scales[10] >> (s_shift + 0) & 0x3) << 4)) - 32), FLOAT_TYPE(((data_a[ib0 + i].qs[q_offset + l   ] >> 2) & 3) - (((data_a[ib0 + i].hmask[l0 + l   ] & (m << 1)) != 0) ? 0 : 4)),
+                  fma(FLOAT_TYPE(data_b[b_offset + y_idx + l + 64]) * FLOAT_TYPE(int8_t(((data_a[ib0 + i].scales[4] >> s_shift) & 0xF) | ((data_a[ib0 + i].scales[ 8] >> (s_shift + 2) & 0x3) << 4)) - 32), FLOAT_TYPE(((data_a[ib0 + i].qs[q_offset + l   ] >> 4) & 3) - (((data_a[ib0 + i].hmask[l0 + l   ] & (m << 2)) != 0) ? 0 : 4)),
+                  fma(FLOAT_TYPE(data_b[b_offset + y_idx + l + 96]) * FLOAT_TYPE(int8_t(((data_a[ib0 + i].scales[6] >> s_shift) & 0xF) | ((data_a[ib0 + i].scales[10] >> (s_shift + 2) & 0x3) << 4)) - 32), FLOAT_TYPE(((data_a[ib0 + i].qs[q_offset + l   ] >> 6) & 3) - (((data_a[ib0 + i].hmask[l0 + l   ] & (m << 3)) != 0) ? 0 : 4)),
+                  fma(FLOAT_TYPE(data_b[b_offset + y_idx + l + 16]) * FLOAT_TYPE(int8_t(((data_a[ib0 + i].scales[1] >> s_shift) & 0xF) | ((data_a[ib0 + i].scales[ 9] >> (s_shift + 0) & 0x3) << 4)) - 32), FLOAT_TYPE(((data_a[ib0 + i].qs[q_offset + l+16]     ) & 3) - (((data_a[ib0 + i].hmask[l0 + l+16] & (m << 0)) != 0) ? 0 : 4)),
+                  fma(FLOAT_TYPE(data_b[b_offset + y_idx + l + 48]) * FLOAT_TYPE(int8_t(((data_a[ib0 + i].scales[3] >> s_shift) & 0xF) | ((data_a[ib0 + i].scales[11] >> (s_shift + 0) & 0x3) << 4)) - 32), FLOAT_TYPE(((data_a[ib0 + i].qs[q_offset + l+16] >> 2) & 3) - (((data_a[ib0 + i].hmask[l0 + l+16] & (m << 1)) != 0) ? 0 : 4)),
+                  fma(FLOAT_TYPE(data_b[b_offset + y_idx + l + 80]) * FLOAT_TYPE(int8_t(((data_a[ib0 + i].scales[5] >> s_shift) & 0xF) | ((data_a[ib0 + i].scales[ 9] >> (s_shift + 2) & 0x3) << 4)) - 32), FLOAT_TYPE(((data_a[ib0 + i].qs[q_offset + l+16] >> 4) & 3) - (((data_a[ib0 + i].hmask[l0 + l+16] & (m << 2)) != 0) ? 0 : 4)),
+                  fma(FLOAT_TYPE(data_b[b_offset + y_idx + l +112]) * FLOAT_TYPE(int8_t(((data_a[ib0 + i].scales[7] >> s_shift) & 0xF) | ((data_a[ib0 + i].scales[11] >> (s_shift + 2) & 0x3) << 4)) - 32), FLOAT_TYPE(((data_a[ib0 + i].qs[q_offset + l+16] >> 6) & 3) - (((data_a[ib0 + i].hmask[l0 + l+16] & (m << 3)) != 0) ? 0 : 4)), sum))))))));
         }
-        tmp[16 * ix + tid] += d * sum;
+        const uint tmp_idx = 16 * ix + tid;
+        tmp[tmp_idx] = fma(d, sum, tmp[tmp_idx]);
     }
 
     // sum up partial sums and write back result

ggml/src/vulkan-shaders/mul_mat_vec_q4_k.comp

@@ -67,17 +67,17 @@ void main() {
         const uint8_t q4_14 = uint8_t(data_a[ib0 + i].qs[q_offset + 66]  >> 4);
         const uint8_t q4_15 = uint8_t(data_a[ib0 + i].qs[q_offset + 67]  >> 4);
 
-        const FLOAT_TYPE sx = FLOAT_TYPE(FLOAT_TYPE(data_b[b_offset + y1_idx]) * q4_0 + FLOAT_TYPE(data_b[b_offset + y1_idx + 1]) * q4_1 + FLOAT_TYPE(data_b[b_offset + y1_idx + 2]) * q4_2 + FLOAT_TYPE(data_b[b_offset + y1_idx + 3]) * q4_3);
-        const FLOAT_TYPE sy = FLOAT_TYPE(FLOAT_TYPE(data_b[b_offset + y1_idx + 32]) * q4_4 + FLOAT_TYPE(data_b[b_offset + y1_idx + 33]) * q4_5 + FLOAT_TYPE(data_b[b_offset + y1_idx + 34]) * q4_6 + FLOAT_TYPE(data_b[b_offset + y1_idx + 35]) * q4_7);
-        const FLOAT_TYPE sz = FLOAT_TYPE(FLOAT_TYPE(data_b[b_offset + y2_idx]) * q4_8 + FLOAT_TYPE(data_b[b_offset + y2_idx + 1]) * q4_9 + FLOAT_TYPE(data_b[b_offset + y2_idx + 2]) * q4_10 + FLOAT_TYPE(data_b[b_offset + y2_idx + 3]) * q4_11);
-        const FLOAT_TYPE sw = FLOAT_TYPE(FLOAT_TYPE(data_b[b_offset + y2_idx + 32]) * q4_12 + FLOAT_TYPE(data_b[b_offset + y2_idx + 33]) * q4_13 + FLOAT_TYPE(data_b[b_offset + y2_idx + 34]) * q4_14 + FLOAT_TYPE(data_b[b_offset + y2_idx + 35]) * q4_15);
-        const FLOAT_TYPE smin = FLOAT_TYPE(
-            FLOAT_TYPE(data_b[b_offset + y1_idx    ]) * sc2 + FLOAT_TYPE(data_b[b_offset + y1_idx + 32]) * sc3 + FLOAT_TYPE(data_b[b_offset + y2_idx    ]) * sc6 + FLOAT_TYPE(data_b[b_offset + y2_idx + 32]) * sc7
-          + FLOAT_TYPE(data_b[b_offset + y1_idx + 1]) * sc2 + FLOAT_TYPE(data_b[b_offset + y1_idx + 33]) * sc3 + FLOAT_TYPE(data_b[b_offset + y2_idx + 1]) * sc6 + FLOAT_TYPE(data_b[b_offset + y2_idx + 33]) * sc7
-          + FLOAT_TYPE(data_b[b_offset + y1_idx + 2]) * sc2 + FLOAT_TYPE(data_b[b_offset + y1_idx + 34]) * sc3 + FLOAT_TYPE(data_b[b_offset + y2_idx + 2]) * sc6 + FLOAT_TYPE(data_b[b_offset + y2_idx + 34]) * sc7
-          + FLOAT_TYPE(data_b[b_offset + y1_idx + 3]) * sc2 + FLOAT_TYPE(data_b[b_offset + y1_idx + 35]) * sc3 + FLOAT_TYPE(data_b[b_offset + y2_idx + 3]) * sc6 + FLOAT_TYPE(data_b[b_offset + y2_idx + 35]) * sc7
-        );
-        tmp[16 * ix + tid] += FLOAT_TYPE(dall * (sx * sc0 + sy * sc1 + sz * sc4 + sw * sc5) - dmin * smin);
+        const FLOAT_TYPE sx = fma(FLOAT_TYPE(data_b[b_offset + y1_idx]),      q4_0,  fma(FLOAT_TYPE(data_b[b_offset + y1_idx + 1]),  q4_1,  fma(FLOAT_TYPE(data_b[b_offset + y1_idx + 2]),  q4_2,  FLOAT_TYPE(data_b[b_offset + y1_idx + 3]) *  q4_3)));
+        const FLOAT_TYPE sy = fma(FLOAT_TYPE(data_b[b_offset + y1_idx + 32]), q4_4,  fma(FLOAT_TYPE(data_b[b_offset + y1_idx + 33]), q4_5,  fma(FLOAT_TYPE(data_b[b_offset + y1_idx + 34]), q4_6,  FLOAT_TYPE(data_b[b_offset + y1_idx + 35]) * q4_7)));
+        const FLOAT_TYPE sz = fma(FLOAT_TYPE(data_b[b_offset + y2_idx]),      q4_8,  fma(FLOAT_TYPE(data_b[b_offset + y2_idx + 1]),  q4_9,  fma(FLOAT_TYPE(data_b[b_offset + y2_idx + 2]),  q4_10, FLOAT_TYPE(data_b[b_offset + y2_idx + 3]) *  q4_11)));
+        const FLOAT_TYPE sw = fma(FLOAT_TYPE(data_b[b_offset + y2_idx + 32]), q4_12, fma(FLOAT_TYPE(data_b[b_offset + y2_idx + 33]), q4_13, fma(FLOAT_TYPE(data_b[b_offset + y2_idx + 34]), q4_14, FLOAT_TYPE(data_b[b_offset + y2_idx + 35]) * q4_15)));
+        const FLOAT_TYPE smin =
+            fma(FLOAT_TYPE(data_b[b_offset + y1_idx    ]), sc2, fma(FLOAT_TYPE(data_b[b_offset + y1_idx + 32]), sc3, fma(FLOAT_TYPE(data_b[b_offset + y2_idx    ]), sc6, fma(FLOAT_TYPE(data_b[b_offset + y2_idx + 32]), sc7,
+            fma(FLOAT_TYPE(data_b[b_offset + y1_idx + 1]), sc2, fma(FLOAT_TYPE(data_b[b_offset + y1_idx + 33]), sc3, fma(FLOAT_TYPE(data_b[b_offset + y2_idx + 1]), sc6, fma(FLOAT_TYPE(data_b[b_offset + y2_idx + 33]), sc7,
+            fma(FLOAT_TYPE(data_b[b_offset + y1_idx + 2]), sc2, fma(FLOAT_TYPE(data_b[b_offset + y1_idx + 34]), sc3, fma(FLOAT_TYPE(data_b[b_offset + y2_idx + 2]), sc6, fma(FLOAT_TYPE(data_b[b_offset + y2_idx + 34]), sc7,
+            fma(FLOAT_TYPE(data_b[b_offset + y1_idx + 3]), sc2, fma(FLOAT_TYPE(data_b[b_offset + y1_idx + 35]), sc3, fma(FLOAT_TYPE(data_b[b_offset + y2_idx + 3]), sc6,     FLOAT_TYPE(data_b[b_offset + y2_idx + 35]) * sc7)))))))))))))));
+        const uint tmp_idx = 16 * ix + tid;
+        tmp[tmp_idx] = fma(dall, fma(sx, sc0, fma(sy, sc1, fma(sz, sc4, sw * sc5))), fma(-dmin, smin, tmp[tmp_idx]));
 #else
         const uint8_t q4_0 = uint8_t(data_a[ib0 + i].qs[q_offset     ] & 0xf);
         const uint8_t q4_1 = uint8_t(data_a[ib0 + i].qs[q_offset +  1] & 0xf);
@@ -88,16 +88,19 @@ void main() {
         const uint8_t q4_6 = uint8_t(data_a[ib0 + i].qs[q_offset + 64]  >> 4);
         const uint8_t q4_7 = uint8_t(data_a[ib0 + i].qs[q_offset + 65]  >> 4);
 
-        const FLOAT_TYPE sx = FLOAT_TYPE(FLOAT_TYPE(data_b[b_offset + y1_idx     ]) * q4_0  + FLOAT_TYPE(data_b[b_offset + y1_idx +  1]) * q4_1);
-        const FLOAT_TYPE sy = FLOAT_TYPE(FLOAT_TYPE(data_b[b_offset + y1_idx + 32]) * q4_2  + FLOAT_TYPE(data_b[b_offset + y1_idx + 33]) * q4_3);
-        const FLOAT_TYPE sz = FLOAT_TYPE(FLOAT_TYPE(data_b[b_offset + y2_idx     ]) * q4_4  + FLOAT_TYPE(data_b[b_offset + y2_idx +  1]) * q4_5);
-        const FLOAT_TYPE sw = FLOAT_TYPE(FLOAT_TYPE(data_b[b_offset + y2_idx + 32]) * q4_6 + FLOAT_TYPE(data_b[b_offset + y2_idx + 33]) * q4_7);
-        const FLOAT_TYPE smin = FLOAT_TYPE(
-            FLOAT_TYPE(data_b[b_offset + y1_idx]) * sc2 + FLOAT_TYPE(data_b[b_offset + y1_idx + 32]) * sc3 + FLOAT_TYPE(data_b[b_offset + y2_idx]) * sc6 + FLOAT_TYPE(data_b[b_offset + y2_idx + 32]) * sc7
-          + FLOAT_TYPE(data_b[b_offset + y1_idx + 1]) * sc2 + FLOAT_TYPE(data_b[b_offset + y1_idx + 33]) * sc3 + FLOAT_TYPE(data_b[b_offset + y2_idx + 1]) * sc6 + FLOAT_TYPE(data_b[b_offset + y2_idx + 33]) * sc7
-        );
+        const FLOAT_TYPE sx = fma(FLOAT_TYPE(data_b[b_offset + y1_idx     ]), q4_0, FLOAT_TYPE(data_b[b_offset + y1_idx +  1]) * q4_1);
+        const FLOAT_TYPE sy = fma(FLOAT_TYPE(data_b[b_offset + y1_idx + 32]), q4_2, FLOAT_TYPE(data_b[b_offset + y1_idx + 33]) * q4_3);
+        const FLOAT_TYPE sz = fma(FLOAT_TYPE(data_b[b_offset + y2_idx     ]), q4_4, FLOAT_TYPE(data_b[b_offset + y2_idx +  1]) * q4_5);
+        const FLOAT_TYPE sw = fma(FLOAT_TYPE(data_b[b_offset + y2_idx + 32]), q4_6, FLOAT_TYPE(data_b[b_offset + y2_idx + 33]) * q4_7);
+        const FLOAT_TYPE smin =
+            fma(FLOAT_TYPE(data_b[b_offset + y1_idx    ]), sc2, fma(FLOAT_TYPE(data_b[b_offset + y1_idx + 32]), sc3, fma(FLOAT_TYPE(data_b[b_offset + y2_idx    ]), sc6, fma(FLOAT_TYPE(data_b[b_offset + y2_idx + 32]), sc7,
+          + fma(FLOAT_TYPE(data_b[b_offset + y1_idx + 1]), sc2, fma(FLOAT_TYPE(data_b[b_offset + y1_idx + 33]), sc3, fma(FLOAT_TYPE(data_b[b_offset + y2_idx + 1]), sc6, FLOAT_TYPE(data_b[b_offset + y2_idx + 33]) * sc7)))))));
 
-        tmp[16 * ix + tid] += FLOAT_TYPE(dall * (sx * FLOAT_TYPE(data_a[ib0 + i].scales[v_im] & 0x3f) + sy * FLOAT_TYPE(data_a[ib0 + i].scales[v_im + 1] & 0x3f) + sz * FLOAT_TYPE((data_a[ib0 + i].scales[v_im + 4] & 0x0f) | ((data_a[ib0 + i].scales[v_im] & 0xc0) >> 2)) + sw * FLOAT_TYPE((data_a[ib0 + i].scales[v_im + 5] & 0x0f) | ((data_a[ib0 + i].scales[v_im + 1] & 0xc0) >> 2))) - dmin * smin);
+        tmp[16 * ix + tid] += FLOAT_TYPE(dall * (sx * FLOAT_TYPE(data_a[ib0 + i].scales[v_im] & 0x3f) + sy * FLOAT_TYPE(data_a[ib0 + i].scales[v_im + 1] & 0x3f) +
+                        sz * FLOAT_TYPE((data_a[ib0 + i].scales[v_im + 4] & 0x0f) | ((data_a[ib0 + i].scales[v_im] & 0xc0) >> 2)) + sw * FLOAT_TYPE((data_a[ib0 + i].scales[v_im + 5] & 0x0f) | ((data_a[ib0 + i].scales[v_im + 1] & 0xc0) >> 2))) - dmin * smin);
+        const uint tmp_idx = 16 * ix + tid;
+        tmp[tmp_idx] = fma(dall, (fma(sx, FLOAT_TYPE(data_a[ib0 + i].scales[v_im] & 0x3f), fma(sy, FLOAT_TYPE(data_a[ib0 + i].scales[v_im + 1] & 0x3f),
+                       fma(sz, FLOAT_TYPE((data_a[ib0 + i].scales[v_im + 4] & 0x0f) | ((data_a[ib0 + i].scales[v_im] & 0xc0) >> 2)), fma(sw, FLOAT_TYPE((data_a[ib0 + i].scales[v_im + 5] & 0x0f) | ((data_a[ib0 + i].scales[v_im + 1] & 0xc0) >> 2))))))), fma(-dmin, smin, tmp[tmp_idx]));
 #endif
     }
 

ggml/src/vulkan-shaders/mul_mat_vec_q5_k.comp

@@ -66,35 +66,33 @@ void main() {
         const uint8_t q4_14 = uint8_t(data_a[ib0 + i].qs[q_offset + 80]  >> 4);
         const uint8_t q4_15 = uint8_t(data_a[ib0 + i].qs[q_offset + 81]  >> 4);
 
-        const FLOAT_TYPE sx = FLOAT_TYPE(
-            FLOAT_TYPE(data_b[b_offset + y1_idx     ]) * (q4_0 + (((data_a[ib0 + i].qh[l0     ] & hm1) != 0) ? 16 : 0))
-          + FLOAT_TYPE(data_b[b_offset + y1_idx +  1]) * (q4_1 + (((data_a[ib0 + i].qh[l0 +  1] & hm1) != 0) ? 16 : 0))
-          + FLOAT_TYPE(data_b[b_offset + y1_idx + 16]) * (q4_2 + (((data_a[ib0 + i].qh[l0 + 16] & hm1) != 0) ? 16 : 0))
-          + FLOAT_TYPE(data_b[b_offset + y1_idx + 17]) * (q4_3 + (((data_a[ib0 + i].qh[l0 + 17] & hm1) != 0) ? 16 : 0))
-        );
-        const FLOAT_TYPE sy = FLOAT_TYPE(
-            FLOAT_TYPE(data_b[b_offset + y1_idx + 32]) * (q4_4 + (((data_a[ib0 + i].qh[l0     ] & (hm1 << 1)) != 0) ? 16 : 0))
-          + FLOAT_TYPE(data_b[b_offset + y1_idx + 33]) * (q4_5 + (((data_a[ib0 + i].qh[l0 +  1] & (hm1 << 1)) != 0) ? 16 : 0))
-          + FLOAT_TYPE(data_b[b_offset + y1_idx + 48]) * (q4_6 + (((data_a[ib0 + i].qh[l0 + 16] & (hm1 << 1)) != 0) ? 16 : 0))
-          + FLOAT_TYPE(data_b[b_offset + y1_idx + 49]) * (q4_7 + (((data_a[ib0 + i].qh[l0 + 17] & (hm1 << 1)) != 0) ? 16 : 0))
-        );
-        const FLOAT_TYPE sz = FLOAT_TYPE(
-            FLOAT_TYPE(data_b[b_offset + y2_idx     ]) * (q4_8  + (((data_a[ib0 + i].qh[l0     ] & hm2) != 0) ? 16 : 0))
-          + FLOAT_TYPE(data_b[b_offset + y2_idx +  1]) * (q4_9  + (((data_a[ib0 + i].qh[l0 +  1] & hm2) != 0) ? 16 : 0))
-          + FLOAT_TYPE(data_b[b_offset + y2_idx + 16]) * (q4_10 + (((data_a[ib0 + i].qh[l0 + 16] & hm2) != 0) ? 16 : 0))
-          + FLOAT_TYPE(data_b[b_offset + y2_idx + 17]) * (q4_11 + (((data_a[ib0 + i].qh[l0 + 17] & hm2) != 0) ? 16 : 0))
-        );
-        const FLOAT_TYPE sw = FLOAT_TYPE(
-            FLOAT_TYPE(data_b[b_offset + y2_idx + 32]) * (q4_12 + (((data_a[ib0 + i].qh[l0     ] & (hm2 << 1)) != 0) ? 16 : 0))
-          + FLOAT_TYPE(data_b[b_offset + y2_idx + 33]) * (q4_13 + (((data_a[ib0 + i].qh[l0 +  1] & (hm2 << 1)) != 0) ? 16 : 0))
-          + FLOAT_TYPE(data_b[b_offset + y2_idx + 48]) * (q4_14 + (((data_a[ib0 + i].qh[l0 + 16] & (hm2 << 1)) != 0) ? 16 : 0))
-          + FLOAT_TYPE(data_b[b_offset + y2_idx + 49]) * (q4_15 + (((data_a[ib0 + i].qh[l0 + 17] & (hm2 << 1)) != 0) ? 16 : 0))
-        );
-        const FLOAT_TYPE smin = FLOAT_TYPE(
-            (FLOAT_TYPE(data_b[b_offset + y1_idx]) + FLOAT_TYPE(data_b[b_offset + y1_idx + 1]) + FLOAT_TYPE(data_b[b_offset + y1_idx + 16]) + FLOAT_TYPE(data_b[b_offset + y1_idx + 17])) * sc2 + (FLOAT_TYPE(data_b[b_offset + y1_idx + 32]) + FLOAT_TYPE(data_b[b_offset + y1_idx + 33]) + FLOAT_TYPE(data_b[b_offset + y1_idx + 48]) + FLOAT_TYPE(data_b[b_offset + y1_idx + 49])) * sc3
-          + (FLOAT_TYPE(data_b[b_offset + y2_idx]) + FLOAT_TYPE(data_b[b_offset + y2_idx + 1]) + FLOAT_TYPE(data_b[b_offset + y2_idx + 16]) + FLOAT_TYPE(data_b[b_offset + y2_idx + 17])) * sc6 + (FLOAT_TYPE(data_b[b_offset + y2_idx + 32]) + FLOAT_TYPE(data_b[b_offset + y2_idx + 33]) + FLOAT_TYPE(data_b[b_offset + y2_idx + 48]) + FLOAT_TYPE(data_b[b_offset + y2_idx + 49])) * sc7
-        );
-        tmp[16 * ix + tid] += FLOAT_TYPE(dall * (sx * sc0 + sy * sc1 + sz * sc4 + sw * sc5) - dmin * smin);
+        const FLOAT_TYPE sx =
+          fma(FLOAT_TYPE(data_b[b_offset + y1_idx     ]), (q4_0 + (((data_a[ib0 + i].qh[l0     ] & hm1) != 0) ? 16 : 0)),
+          fma(FLOAT_TYPE(data_b[b_offset + y1_idx +  1]), (q4_1 + (((data_a[ib0 + i].qh[l0 +  1] & hm1) != 0) ? 16 : 0)),
+          fma(FLOAT_TYPE(data_b[b_offset + y1_idx + 16]), (q4_2 + (((data_a[ib0 + i].qh[l0 + 16] & hm1) != 0) ? 16 : 0)),
+             FLOAT_TYPE(data_b[b_offset + y1_idx + 17]) * (q4_3 + (((data_a[ib0 + i].qh[l0 + 17] & hm1) != 0) ? 16 : 0)))));
+        const FLOAT_TYPE sy =
+          fma(FLOAT_TYPE(data_b[b_offset + y1_idx + 32]), (q4_4 + (((data_a[ib0 + i].qh[l0     ] & (hm1 << 1)) != 0) ? 16 : 0)),
+          fma(FLOAT_TYPE(data_b[b_offset + y1_idx + 33]), (q4_5 + (((data_a[ib0 + i].qh[l0 +  1] & (hm1 << 1)) != 0) ? 16 : 0)),
+          fma(FLOAT_TYPE(data_b[b_offset + y1_idx + 48]), (q4_6 + (((data_a[ib0 + i].qh[l0 + 16] & (hm1 << 1)) != 0) ? 16 : 0)),
+             FLOAT_TYPE(data_b[b_offset + y1_idx + 49]) * (q4_7 + (((data_a[ib0 + i].qh[l0 + 17] & (hm1 << 1)) != 0) ? 16 : 0)))));
+        const FLOAT_TYPE sz =
+          fma(FLOAT_TYPE(data_b[b_offset + y2_idx     ]), (q4_8  + (((data_a[ib0 + i].qh[l0     ] & hm2) != 0) ? 16 : 0)),
+          fma(FLOAT_TYPE(data_b[b_offset + y2_idx +  1]), (q4_9  + (((data_a[ib0 + i].qh[l0 +  1] & hm2) != 0) ? 16 : 0)),
+          fma(FLOAT_TYPE(data_b[b_offset + y2_idx + 16]), (q4_10 + (((data_a[ib0 + i].qh[l0 + 16] & hm2) != 0) ? 16 : 0)),
+             FLOAT_TYPE(data_b[b_offset + y2_idx + 17]) * (q4_11 + (((data_a[ib0 + i].qh[l0 + 17] & hm2) != 0) ? 16 : 0)))));
+        const FLOAT_TYPE sw =
+          fma(FLOAT_TYPE(data_b[b_offset + y2_idx + 32]), (q4_12 + (((data_a[ib0 + i].qh[l0     ] & (hm2 << 1)) != 0) ? 16 : 0)),
+          fma(FLOAT_TYPE(data_b[b_offset + y2_idx + 33]), (q4_13 + (((data_a[ib0 + i].qh[l0 +  1] & (hm2 << 1)) != 0) ? 16 : 0)),
+          fma(FLOAT_TYPE(data_b[b_offset + y2_idx + 48]), (q4_14 + (((data_a[ib0 + i].qh[l0 + 16] & (hm2 << 1)) != 0) ? 16 : 0)),
+             FLOAT_TYPE(data_b[b_offset + y2_idx + 49]) * (q4_15 + (((data_a[ib0 + i].qh[l0 + 17] & (hm2 << 1)) != 0) ? 16 : 0)))));
+        const FLOAT_TYPE smin =
+          fma(FLOAT_TYPE(data_b[b_offset + y1_idx     ]) + FLOAT_TYPE(data_b[b_offset + y1_idx + 1 ]) + FLOAT_TYPE(data_b[b_offset + y1_idx + 16]) + FLOAT_TYPE(data_b[b_offset + y1_idx + 17]), sc2,
+          fma(FLOAT_TYPE(data_b[b_offset + y1_idx + 32]) + FLOAT_TYPE(data_b[b_offset + y1_idx + 33]) + FLOAT_TYPE(data_b[b_offset + y1_idx + 48]) + FLOAT_TYPE(data_b[b_offset + y1_idx + 49]), sc3,
+          fma(FLOAT_TYPE(data_b[b_offset + y2_idx     ]) + FLOAT_TYPE(data_b[b_offset + y2_idx + 1 ]) + FLOAT_TYPE(data_b[b_offset + y2_idx + 16]) + FLOAT_TYPE(data_b[b_offset + y2_idx + 17]), sc6,
+              (FLOAT_TYPE(data_b[b_offset + y2_idx + 32]) + FLOAT_TYPE(data_b[b_offset + y2_idx + 33]) + FLOAT_TYPE(data_b[b_offset + y2_idx + 48]) + FLOAT_TYPE(data_b[b_offset + y2_idx + 49])) * sc7)));
+        const uint tmp_idx = 16 * ix + tid;
+        tmp[tmp_idx] = fma(dall, fma(sx, sc0, fma(sy, sc1, fma(sz, sc4, sw * sc5))), fma(-dmin, smin, tmp[tmp_idx]));
     }
 
     // sum up partial sums and write back result

ggml/src/vulkan-shaders/mul_mat_vec_q6_k.comp

@@ -44,22 +44,22 @@ void main() {
         const FLOAT_TYPE d = FLOAT_TYPE(data_a[ib0 + i].d);
 
 #if K_QUANTS_PER_ITERATION == 1
-        FLOAT_TYPE sum = FLOAT_TYPE(data_b[b_offset + y_idx +  0]) * FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 0]) * d * FLOAT_TYPE(int8_t((data_a[ib0 + i].ql[ql_offset +  0] & 0xF) | ((data_a[ib0 + i].qh[qh_offset +  0] & 0x03) << 4)) - 32)
-                       + FLOAT_TYPE(data_b[b_offset + y_idx + 16]) * FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 1]) * d * FLOAT_TYPE(int8_t((data_a[ib0 + i].ql[ql_offset + 16] & 0xF) | ((data_a[ib0 + i].qh[qh_offset + 16] & 0x03) << 4)) - 32)
-                       + FLOAT_TYPE(data_b[b_offset + y_idx + 32]) * FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 2]) * d * FLOAT_TYPE(int8_t((data_a[ib0 + i].ql[ql_offset + 32] & 0xF) | ((data_a[ib0 + i].qh[qh_offset +  0] & 0x0c) << 2)) - 32)
-                       + FLOAT_TYPE(data_b[b_offset + y_idx + 48]) * FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 3]) * d * FLOAT_TYPE(int8_t((data_a[ib0 + i].ql[ql_offset + 48] & 0xF) | ((data_a[ib0 + i].qh[qh_offset + 16] & 0x0c) << 2)) - 32)
-                       + FLOAT_TYPE(data_b[b_offset + y_idx + 64]) * FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 4]) * d * FLOAT_TYPE(int8_t((data_a[ib0 + i].ql[ql_offset +  0]  >> 4) | ((data_a[ib0 + i].qh[qh_offset +  0] & 0x30) >> 0)) - 32)
-                       + FLOAT_TYPE(data_b[b_offset + y_idx + 80]) * FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 5]) * d * FLOAT_TYPE(int8_t((data_a[ib0 + i].ql[ql_offset + 16]  >> 4) | ((data_a[ib0 + i].qh[qh_offset + 16] & 0x30) >> 0)) - 32)
-                       + FLOAT_TYPE(data_b[b_offset + y_idx + 96]) * FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 6]) * d * FLOAT_TYPE(int8_t((data_a[ib0 + i].ql[ql_offset + 32]  >> 4) | ((data_a[ib0 + i].qh[qh_offset +  0] & 0xc0) >> 2)) - 32)
-                       + FLOAT_TYPE(data_b[b_offset + y_idx +112]) * FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 7]) * d * FLOAT_TYPE(int8_t((data_a[ib0 + i].ql[ql_offset + 48]  >> 4) | ((data_a[ib0 + i].qh[qh_offset + 16] & 0xc0) >> 2)) - 32);
-        tmp[16 * ix + tid] += sum;
+        const uint tmp_idx = 16 * ix + tid;
+        tmp[tmp_idx] = fma(FLOAT_TYPE(data_b[b_offset + y_idx +  0]) * FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 0]) * d, FLOAT_TYPE(int8_t((data_a[ib0 + i].ql[ql_offset +  0] & 0xF) | ((data_a[ib0 + i].qh[qh_offset +  0] & 0x03) << 4)) - 32),
+                       fma(FLOAT_TYPE(data_b[b_offset + y_idx + 16]) * FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 1]) * d, FLOAT_TYPE(int8_t((data_a[ib0 + i].ql[ql_offset + 16] & 0xF) | ((data_a[ib0 + i].qh[qh_offset + 16] & 0x03) << 4)) - 32),
+                       fma(FLOAT_TYPE(data_b[b_offset + y_idx + 32]) * FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 2]) * d, FLOAT_TYPE(int8_t((data_a[ib0 + i].ql[ql_offset + 32] & 0xF) | ((data_a[ib0 + i].qh[qh_offset +  0] & 0x0c) << 2)) - 32),
+                       fma(FLOAT_TYPE(data_b[b_offset + y_idx + 48]) * FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 3]) * d, FLOAT_TYPE(int8_t((data_a[ib0 + i].ql[ql_offset + 48] & 0xF) | ((data_a[ib0 + i].qh[qh_offset + 16] & 0x0c) << 2)) - 32),
+                       fma(FLOAT_TYPE(data_b[b_offset + y_idx + 64]) * FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 4]) * d, FLOAT_TYPE(int8_t((data_a[ib0 + i].ql[ql_offset +  0]  >> 4) | ((data_a[ib0 + i].qh[qh_offset +  0] & 0x30) >> 0)) - 32),
+                       fma(FLOAT_TYPE(data_b[b_offset + y_idx + 80]) * FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 5]) * d, FLOAT_TYPE(int8_t((data_a[ib0 + i].ql[ql_offset + 16]  >> 4) | ((data_a[ib0 + i].qh[qh_offset + 16] & 0x30) >> 0)) - 32),
+                       fma(FLOAT_TYPE(data_b[b_offset + y_idx + 96]) * FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 6]) * d, FLOAT_TYPE(int8_t((data_a[ib0 + i].ql[ql_offset + 32]  >> 4) | ((data_a[ib0 + i].qh[qh_offset +  0] & 0xc0) >> 2)) - 32),
+                       fma(FLOAT_TYPE(data_b[b_offset + y_idx +112]) * FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 7]) * d, FLOAT_TYPE(int8_t((data_a[ib0 + i].ql[ql_offset + 48]  >> 4) | ((data_a[ib0 + i].qh[qh_offset + 16] & 0xc0) >> 2)) - 32), tmp[tmp_idx]))))))));
 #else
         FLOAT_TYPE sum = FLOAT_TYPE(0.0);
         [[unroll]] for (int l = 0; l < 4; ++l) {
-            sum += FLOAT_TYPE(data_b[b_offset + y_idx + l+ 0]) * FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 0]) * d * FLOAT_TYPE(int8_t((data_a[ib0 + i].ql[ql_offset + l+ 0] & 0xF) | (((data_a[ib0 + i].qh[qh_offset + l] >> 0) & 3) << 4)) - 32)
-                 + FLOAT_TYPE(data_b[b_offset + y_idx + l+32]) * FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 2]) * d * FLOAT_TYPE(int8_t((data_a[ib0 + i].ql[ql_offset + l+32] & 0xF) | (((data_a[ib0 + i].qh[qh_offset + l] >> 2) & 3) << 4)) - 32)
-                 + FLOAT_TYPE(data_b[b_offset + y_idx + l+64]) * FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 4]) * d * FLOAT_TYPE(int8_t((data_a[ib0 + i].ql[ql_offset + l+ 0]  >> 4) | (((data_a[ib0 + i].qh[qh_offset + l] >> 4) & 3) << 4)) - 32)
-                 + FLOAT_TYPE(data_b[b_offset + y_idx + l+96]) * FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 6]) * d * FLOAT_TYPE(int8_t((data_a[ib0 + i].ql[ql_offset + l+32]  >> 4) | (((data_a[ib0 + i].qh[qh_offset + l] >> 6) & 3) << 4)) - 32);
+            sum = fma(FLOAT_TYPE(data_b[b_offset + y_idx + l+ 0]) * FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 0]) * d, FLOAT_TYPE(int8_t((data_a[ib0 + i].ql[ql_offset + l+ 0] & 0xF) | (((data_a[ib0 + i].qh[qh_offset + l] >> 0) & 3) << 4)) - 32),
+                  fma(FLOAT_TYPE(data_b[b_offset + y_idx + l+32]) * FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 2]) * d, FLOAT_TYPE(int8_t((data_a[ib0 + i].ql[ql_offset + l+32] & 0xF) | (((data_a[ib0 + i].qh[qh_offset + l] >> 2) & 3) << 4)) - 32),
+                  fma(FLOAT_TYPE(data_b[b_offset + y_idx + l+64]) * FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 4]) * d, FLOAT_TYPE(int8_t((data_a[ib0 + i].ql[ql_offset + l+ 0]  >> 4) | (((data_a[ib0 + i].qh[qh_offset + l] >> 4) & 3) << 4)) - 32),
+                  fma(FLOAT_TYPE(data_b[b_offset + y_idx + l+96]) * FLOAT_TYPE(data_a[ib0 + i].scales[s_offset + 6]) * d, FLOAT_TYPE(int8_t((data_a[ib0 + i].ql[ql_offset + l+32]  >> 4) | (((data_a[ib0 + i].qh[qh_offset + l] >> 6) & 3) << 4)) - 32), sum))));
         }
         tmp[16 * ix + tid] += sum;
 #endif

ggml/src/vulkan-shaders/mul_mm.comp

@@ -326,10 +326,10 @@ void main() {
                 mbyte = uint8_t((data_a[ib].scales[is + 4] >>  4) | ((data_a[ib].scales[is    ] >> 6) << 4));
             }
             const float d = loadd.x * sc;
-            const float m = loadd.y * mbyte;
+            const float m = -loadd.y * mbyte;
 
-            buf_a[buf_idx    ] = FLOAT_TYPE(d * float((data_a[ib].qs[qsi    ] >> (b * 4)) & 0xF) - m);
-            buf_a[buf_idx + 1] = FLOAT_TYPE(d * float((data_a[ib].qs[qsi + 1] >> (b * 4)) & 0xF) - m);
+            buf_a[buf_idx    ] = FLOAT_TYPE(fma(d, float((data_a[ib].qs[qsi    ] >> (b * 4)) & 0xF), m));
+            buf_a[buf_idx + 1] = FLOAT_TYPE(fma(d, float((data_a[ib].qs[qsi + 1] >> (b * 4)) & 0xF), m));
 #elif defined(DATA_A_Q5_K)
             const uint idx = pos_a + (loadc_a + l) * p.stride_a / LOAD_VEC_A + loadr_a;
             const uint buf_idx = (loadc_a + l) * (BK+1) + loadr_a * LOAD_VEC_A;
@@ -357,10 +357,10 @@ void main() {
                 mbyte = uint8_t((data_a[ib].scales[is + 4] >>  4) | ((data_a[ib].scales[is    ] >> 6) << 4));
             }
             const float d = loadd.x * sc;
-            const float m = loadd.y * mbyte;
+            const float m = -loadd.y * mbyte;
 
-            buf_a[buf_idx    ] = FLOAT_TYPE(d * (float((data_a[ib].qs[qsi    ] >> (b * 4)) & 0xF) + float((data_a[ib].qh[qhi    ] & hm) != 0 ? 16 : 0)) - m);
-            buf_a[buf_idx + 1] = FLOAT_TYPE(d * (float((data_a[ib].qs[qsi + 1] >> (b * 4)) & 0xF) + float((data_a[ib].qh[qhi + 1] & hm) != 0 ? 16 : 0)) - m);
+            buf_a[buf_idx    ] = FLOAT_TYPE(fma(d, float((data_a[ib].qs[qsi    ] >> (b * 4)) & 0xF) + float((data_a[ib].qh[qhi    ] & hm) != 0 ? 16 : 0), m));
+            buf_a[buf_idx + 1] = FLOAT_TYPE(fma(d, float((data_a[ib].qs[qsi + 1] >> (b * 4)) & 0xF) + float((data_a[ib].qh[qhi + 1] & hm) != 0 ? 16 : 0), m));
 #elif defined(DATA_A_Q6_K)
             const uint idx = pos_a + (loadc_a + l) * p.stride_a / LOAD_VEC_A + loadr_a;
             const uint buf_idx = (loadc_a + l) * (BK+1) + loadr_a * LOAD_VEC_A;
@@ -463,7 +463,8 @@ void main() {
                 [[unroll]] for (uint wsir = 0; wsir < WMITER; wsir++) {
                     [[unroll]] for (uint cc = 0; cc < TN; cc++) {
                         [[unroll]] for (uint cr = 0; cr < TM; cr++) {
-                            sums[(wsic * TN + cc) * (WMITER * TM) + wsir * TM + cr] += float(cache_a[wsir * TM + cr]) * float(cache_b[wsic * TN + cc]);
+                            const uint sums_idx = (wsic * TN + cc) * (WMITER * TM) + wsir * TM + cr;
+                            sums[sums_idx] = fma(float(cache_a[wsir * TM + cr]), float(cache_b[wsic * TN + cc]), sums[sums_idx]);
                         }
                     }
                 }

ggml/src/vulkan-shaders/repeat.comp

@@ -0,0 +1,24 @@
+#version 450
+
+#include "types.comp"
+#include "generic_unary_head.comp"
+
+uint src0_idx_mod(uint idx) {
+    const uint i13 = idx / (p.ne12*p.ne11*p.ne10);
+    const uint i13_offset = i13 * p.ne12*p.ne11*p.ne10;
+    const uint i12 = (idx - i13_offset) / (p.ne11*p.ne10);
+    const uint i12_offset = i12*p.ne11*p.ne10;
+    const uint i11 = (idx - i13_offset - i12_offset) / p.ne10;
+    const uint i10 = idx - i13_offset - i12_offset - i11*p.ne10;
+    return (i13 % p.ne03)*p.nb03 + (i12 % p.ne02)*p.nb02 + (i11 % p.ne01)*p.nb01 + (i10 % p.ne00)*p.nb00;
+}
+
+void main() {
+    const uint idx = get_idx();
+
+    if (idx >= p.ne) {
+        return;
+    }
+
+    data_d[p.d_offset + dst_idx(idx)] = D_TYPE(data_a[src0_idx_mod(idx)]);
+}

ggml/src/vulkan-shaders/vulkan-shaders-gen.cpp

@@ -380,6 +380,10 @@ void process_shaders(std::vector<std::future<void>>& tasks) {
         string_to_spv("div_f32", "div.comp", {{"A_TYPE", "float"}, {"B_TYPE", "float"}, {"D_TYPE", "float"}, {"FLOAT_TYPE", "float"}});
     }));
 
+    tasks.push_back(std::async(std::launch::async, [] {
+        string_to_spv("repeat_f32", "repeat.comp", {{"A_TYPE", "float"}, {"D_TYPE", "float"}});
+    }));
+
     tasks.push_back(std::async(std::launch::async, [] {
         string_to_spv("scale_f32", "scale.comp", {{"A_TYPE", "float"}, {"D_TYPE", "float"}, {"FLOAT_TYPE", "float"}});
     }));

gguf-py/gguf/constants.py

@@ -219,6 +219,8 @@ class MODEL_ARCH(IntEnum):
     T5           = auto()
     T5ENCODER    = auto()
     JAIS         = auto()
+    NEMOTRON     = auto()
+    EXAONE       = auto()
 
 
 class MODEL_TENSOR(IntEnum):
@@ -347,6 +349,8 @@ MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = {
     MODEL_ARCH.T5:             "t5",
     MODEL_ARCH.T5ENCODER:      "t5encoder",
     MODEL_ARCH.JAIS:           "jais",
+    MODEL_ARCH.NEMOTRON:       "nemotron",
+    MODEL_ARCH.EXAONE:         "exaone",
 }
 
 TENSOR_NAMES: dict[MODEL_TENSOR, str] = {
@@ -1065,6 +1069,37 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
         MODEL_TENSOR.FFN_GATE,
         MODEL_TENSOR.FFN_UP,
     ],
+    MODEL_ARCH.NEMOTRON: [
+        MODEL_TENSOR.TOKEN_EMBD,
+        MODEL_TENSOR.OUTPUT_NORM,
+        MODEL_TENSOR.OUTPUT,
+        MODEL_TENSOR.ROPE_FREQS,
+        MODEL_TENSOR.ATTN_NORM,
+        MODEL_TENSOR.ATTN_Q,
+        MODEL_TENSOR.ATTN_K,
+        MODEL_TENSOR.ATTN_V,
+        MODEL_TENSOR.ATTN_OUT,
+        MODEL_TENSOR.ATTN_ROT_EMBD,
+        MODEL_TENSOR.FFN_NORM,
+        MODEL_TENSOR.FFN_DOWN,
+        MODEL_TENSOR.FFN_UP,
+    ],
+    MODEL_ARCH.EXAONE: [
+        MODEL_TENSOR.TOKEN_EMBD,
+        MODEL_TENSOR.OUTPUT_NORM,
+        MODEL_TENSOR.OUTPUT,
+        MODEL_TENSOR.ROPE_FREQS,
+        MODEL_TENSOR.ATTN_NORM,
+        MODEL_TENSOR.ATTN_Q,
+        MODEL_TENSOR.ATTN_K,
+        MODEL_TENSOR.ATTN_V,
+        MODEL_TENSOR.ATTN_OUT,
+        MODEL_TENSOR.ATTN_ROT_EMBD,
+        MODEL_TENSOR.FFN_NORM,
+        MODEL_TENSOR.FFN_GATE,
+        MODEL_TENSOR.FFN_DOWN,
+        MODEL_TENSOR.FFN_UP,
+    ],
     # TODO
 }
 
@@ -1105,6 +1140,10 @@ MODEL_TENSOR_SKIP: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
     MODEL_ARCH.CHATGLM: [
         MODEL_TENSOR.ROPE_FREQS,
     ],
+    MODEL_ARCH.NEMOTRON: [
+        MODEL_TENSOR.ROPE_FREQS,
+        MODEL_TENSOR.ATTN_ROT_EMBD,
+    ],
 }
 
 #

gguf-py/gguf/tensor_mapping.py

@@ -10,10 +10,10 @@ class TensorNameMap:
         # Token embeddings
         MODEL_TENSOR.TOKEN_EMBD: (
             "gpt_neox.embed_in",                         # gptneox
-            "transformer.wte",                           # gpt2 gpt-j mpt refact qwen dbrx jais
+            "transformer.wte",                           # gpt2 gpt-j mpt refact qwen dbrx jais exaone
             "transformer.word_embeddings",               # falcon
             "word_embeddings",                           # bloom
-            "model.embed_tokens",                        # llama-hf
+            "model.embed_tokens",                        # llama-hf nemotron
             "tok_embeddings",                            # llama-pth
             "embeddings.word_embeddings",                # bert nomic-bert
             "language_model.embedding.word_embeddings",  # persimmon
@@ -52,7 +52,7 @@ class TensorNameMap:
         # Output
         MODEL_TENSOR.OUTPUT: (
             "embed_out",                 # gptneox
-            "lm_head",                   # gpt2 mpt falcon llama-hf baichuan qwen mamba dbrx jais
+            "lm_head",                   # gpt2 mpt falcon llama-hf baichuan qwen mamba dbrx jais nemotron exaone
             "output",                    # llama-pth bloom internlm2
             "word_embeddings_for_head",  # persimmon
             "lm_head.linear",            # phi2
@@ -62,7 +62,7 @@ class TensorNameMap:
         # Output norm
         MODEL_TENSOR.OUTPUT_NORM: (
             "gpt_neox.final_layer_norm",               # gptneox
-            "transformer.ln_f",                        # gpt2 gpt-j falcon jais
+            "transformer.ln_f",                        # gpt2 gpt-j falcon jais exaone
             "model.norm",                              # llama-hf baichuan internlm2
             "norm",                                    # llama-pth
             "transformer.norm_f",                      # mpt dbrx
@@ -75,6 +75,7 @@ class TensorNameMap:
             "transformer.rms_norm",                    # Grok
             "encoder.final_layernorm",                 # chatglm
             "transformer.norm",                        # openelm
+            "model.norm",                              # nemotron
         ),
 
         # Rope frequencies
@@ -88,12 +89,12 @@ class TensorNameMap:
         # Attention norm
         MODEL_TENSOR.ATTN_NORM: (
             "gpt_neox.layers.{bid}.input_layernorm",                # gptneox
-            "transformer.h.{bid}.ln_1",                             # gpt2 gpt-j refact qwen jais
+            "transformer.h.{bid}.ln_1",                             # gpt2 gpt-j refact qwen jais exaone
             "transformer.blocks.{bid}.norm_1",                      # mpt
             "transformer.h.{bid}.input_layernorm",                  # falcon7b
             "h.{bid}.input_layernorm",                              # bloom
             "transformer.h.{bid}.ln_mlp",                           # falcon40b
-            "model.layers.{bid}.input_layernorm",                   # llama-hf
+            "model.layers.{bid}.input_layernorm",                   # llama-hf nemotron
             "layers.{bid}.attention_norm",                          # llama-pth
             "language_model.encoder.layers.{bid}.input_layernorm",  # persimmon
             "model.layers.{bid}.ln1",                               # yi
@@ -135,18 +136,19 @@ class TensorNameMap:
 
         # Attention query
         MODEL_TENSOR.ATTN_Q: (
-            "model.layers.{bid}.self_attn.q_proj",                       # llama-hf
+            "model.layers.{bid}.self_attn.q_proj",                       # llama-hf nemotron
             "layers.{bid}.attention.wq",                                 # llama-pth
             "encoder.layer.{bid}.attention.self.query",                  # bert
             "transformer.h.{bid}.attn.q_proj",                           # gpt-j
             "model.layers.layers.{bid}.self_attn.q_proj",                # plamo
             "model.layers.{bid}.attention.wq",                           # internlm2
             "transformer.decoder_layer.{bid}.multi_head_attention.query",# Grok
+            "transformer.h.{bid}.attn.attention.q_proj",                 # exaone
         ),
 
         # Attention key
         MODEL_TENSOR.ATTN_K: (
-            "model.layers.{bid}.self_attn.k_proj",                     # llama-hf
+            "model.layers.{bid}.self_attn.k_proj",                     # llama-hf nemotron
             "layers.{bid}.attention.wk",                               # llama-pth
             "encoder.layer.{bid}.attention.self.key",                  # bert
             "transformer.h.{bid}.attn.k_proj",                         # gpt-j
@@ -154,18 +156,20 @@ class TensorNameMap:
             "model.layers.layers.{bid}.self_attn.k_proj",              # plamo
             "model.layers.{bid}.attention.wk",                         # internlm2
             "transformer.decoder_layer.{bid}.multi_head_attention.key",# Grok
+            "transformer.h.{bid}.attn.attention.k_proj",               # exaone
         ),
 
         # Attention value
         MODEL_TENSOR.ATTN_V: (
-            "model.layers.{bid}.self_attn.v_proj",                       # llama-hf
+            "model.layers.{bid}.self_attn.v_proj",                       # llama-hf nemotron
             "layers.{bid}.attention.wv",                                 # llama-pth
             "encoder.layer.{bid}.attention.self.value",                  # bert
             "transformer.h.{bid}.attn.v_proj",                           # gpt-j
             "transformer.h.{bid}.attn.v",                                # refact
             "model.layers.layers.{bid}.self_attn.v_proj",                # plamo
             "model.layers.{bid}.attention.wv",                           # internlm2
-            "transformer.decoder_layer.{bid}.multi_head_attention.value" # Grok
+            "transformer.decoder_layer.{bid}.multi_head_attention.value",# Grok
+            "transformer.h.{bid}.attn.attention.v_proj",                 # exaone
         ),
 
         # Attention output
@@ -175,7 +179,7 @@ class TensorNameMap:
             "transformer.blocks.{bid}.attn.out_proj",                       # mpt
             "transformer.h.{bid}.self_attention.dense",                     # falcon
             "h.{bid}.self_attention.dense",                                 # bloom
-            "model.layers.{bid}.self_attn.o_proj",                          # llama-hf
+            "model.layers.{bid}.self_attn.o_proj",                          # llama-hf nemotron
             "layers.{bid}.attention.wo",                                    # llama-pth
             "encoder.layer.{bid}.attention.output.dense",                   # bert
             "transformer.h.{bid}.attn.out_proj",                            # gpt-j
@@ -190,6 +194,7 @@ class TensorNameMap:
             "transformer.blocks.{bid}.norm_attn_norm.attn.out_proj",        # dbrx
             "encoder.layers.{bid}.self_attention.dense",                    # chatglm
             "transformer.layers.{bid}.attn.out_proj",                       # openelm
+            "transformer.h.{bid}.attn.attention.out_proj",                  # exaone
         ),
 
         # Attention output norm
@@ -215,10 +220,10 @@ class TensorNameMap:
         # Feed-forward norm
         MODEL_TENSOR.FFN_NORM: (
             "gpt_neox.layers.{bid}.post_attention_layernorm",                # gptneox
-            "transformer.h.{bid}.ln_2",                                      # gpt2 refact qwen jais
+            "transformer.h.{bid}.ln_2",                                      # gpt2 refact qwen jais exaone
             "h.{bid}.post_attention_layernorm",                              # bloom
             "transformer.blocks.{bid}.norm_2",                               # mpt
-            "model.layers.{bid}.post_attention_layernorm",                   # llama-hf
+            "model.layers.{bid}.post_attention_layernorm",                   # llama-hf nemotron
             "layers.{bid}.ffn_norm",                                         # llama-pth
             "language_model.encoder.layers.{bid}.post_attention_layernorm",  # persimmon
             "model.layers.{bid}.ln2",                                        # yi
@@ -258,7 +263,7 @@ class TensorNameMap:
             "transformer.blocks.{bid}.ffn.up_proj",                   # mpt
             "transformer.h.{bid}.mlp.dense_h_to_4h",                  # falcon
             "h.{bid}.mlp.dense_h_to_4h",                              # bloom
-            "model.layers.{bid}.mlp.up_proj",                         # llama-hf refact
+            "model.layers.{bid}.mlp.up_proj",                         # llama-hf refact nemotron
             "layers.{bid}.feed_forward.w3",                           # llama-pth
             "encoder.layer.{bid}.intermediate.dense",                 # bert
             "transformer.h.{bid}.mlp.fc_in",                          # gpt-j
@@ -277,6 +282,7 @@ class TensorNameMap:
             "encoder.layer.{bid}.mlp.gated_layers_v",                 # jina-bert-v2
             "model.layers.{bid}.residual_mlp.w3",                     # arctic
             "encoder.layers.{bid}.mlp.dense_h_to_4h",                 # chatglm
+            "transformer.h.{bid}.mlp.c_fc_1",                         # exaone
         ),
 
         MODEL_TENSOR.FFN_UP_EXP: (
@@ -308,6 +314,7 @@ class TensorNameMap:
             "encoder.layer.{bid}.mlp.gated_layers_w",     # jina-bert-v2
             "transformer.h.{bid}.mlp.linear_1",           # refact
             "model.layers.{bid}.residual_mlp.w1",         # arctic
+            "transformer.h.{bid}.mlp.c_fc_0",             # exaone
         ),
 
         MODEL_TENSOR.FFN_GATE_EXP: (
@@ -329,7 +336,7 @@ class TensorNameMap:
             "transformer.blocks.{bid}.ffn.down_proj",                 # mpt
             "transformer.h.{bid}.mlp.dense_4h_to_h",                  # falcon
             "h.{bid}.mlp.dense_4h_to_h",                              # bloom
-            "model.layers.{bid}.mlp.down_proj",                       # llama-hf
+            "model.layers.{bid}.mlp.down_proj",                       # llama-hf nemotron
             "layers.{bid}.feed_forward.w2",                           # llama-pth
             "encoder.layer.{bid}.output.dense",                       # bert
             "transformer.h.{bid}.mlp.fc_out",                         # gpt-j
@@ -347,6 +354,7 @@ class TensorNameMap:
             "model.layers.{bid}.residual_mlp.w2",                     # arctic
             "encoder.layer.{bid}.mlp.down_layer",                     # jina-bert-v2
             "encoder.layers.{bid}.mlp.dense_4h_to_h",                 # chatglm
+            "model.layers.h.{bid}.mlp.c_proj",                        # exaone
         ),
 
         MODEL_TENSOR.FFN_DOWN_EXP: (

gguf-py/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "gguf"
-version = "0.9.1"
+version = "0.10.0"
 description = "Read and write ML models in GGUF for GGML"
 authors = ["GGML <ggml@ggml.ai>"]
 packages = [

include/llama.h

@@ -93,15 +93,15 @@ extern "C" {
         LLAMA_VOCAB_PRE_TYPE_TEKKEN         = 20,
         LLAMA_VOCAB_PRE_TYPE_SMOLLM         = 21,
         LLAMA_VOCAB_PRE_TYPE_CODESHELL      = 22,
+        LLAMA_VOCAB_PRE_TYPE_BLOOM          = 23,
+        LLAMA_VOCAB_PRE_TYPE_GPT3_FINNISH   = 24,
+        LLAMA_VOCAB_PRE_TYPE_EXAONE         = 25,
     };
 
-    // note: these values should be synchronized with ggml_rope
-    // TODO: maybe move this enum to ggml.h (ggml_rope_type)
     enum llama_rope_type {
         LLAMA_ROPE_TYPE_NONE = -1,
-        LLAMA_ROPE_TYPE_NORM =  0,
-        LLAMA_ROPE_TYPE_NEOX =  2,
-        LLAMA_ROPE_TYPE_GLM  =  4,
+        LLAMA_ROPE_TYPE_NORM = 0,
+        LLAMA_ROPE_TYPE_NEOX = GGML_ROPE_TYPE_NEOX,
     };
 
     enum llama_token_type { //TODO: remove, required until per token attributes are available from GGUF file
@@ -915,11 +915,8 @@ extern "C" {
     LLAMA_API llama_token llama_token_nl (const struct llama_model * model); // next-line
     LLAMA_API llama_token llama_token_pad(const struct llama_model * model); // padding
 
-    // Returns -1 if unknown, 1 for true or 0 for false.
-    LLAMA_API int32_t llama_add_bos_token(const struct llama_model * model);
-
-    // Returns -1 if unknown, 1 for true or 0 for false.
-    LLAMA_API int32_t llama_add_eos_token(const struct llama_model * model);
+    LLAMA_API bool llama_add_bos_token(const struct llama_model * model);
+    LLAMA_API bool llama_add_eos_token(const struct llama_model * model);
 
     // Codellama infill tokens
     LLAMA_API llama_token llama_token_prefix(const struct llama_model * model); // Beginning of infill prefix

src/llama-vocab.cpp

@@ -388,6 +388,7 @@ struct llm_tokenizer_bpe {
             case LLAMA_VOCAB_PRE_TYPE_COMMAND_R:
             case LLAMA_VOCAB_PRE_TYPE_SMOLLM:
             case LLAMA_VOCAB_PRE_TYPE_CODESHELL:
+            case LLAMA_VOCAB_PRE_TYPE_EXAONE:
                 regex_exprs = {
                     "\\p{N}",
                     "'s|'t|'re|'ve|'m|'ll|'d| ?\\p{L}+| ?\\p{N}+| ?[^\\s\\p{L}\\p{N}]+|\\s+(?!\\S)",
@@ -410,6 +411,8 @@ struct llm_tokenizer_bpe {
                 };
                 break;
             case LLAMA_VOCAB_PRE_TYPE_PORO:
+            case LLAMA_VOCAB_PRE_TYPE_BLOOM:
+            case LLAMA_VOCAB_PRE_TYPE_GPT3_FINNISH:
                 regex_exprs = {
                     " ?[^(\\s|.,!?…。，、।۔،)]+",
                 };
@@ -1466,11 +1469,11 @@ llama_token llama_token_pad_impl(const struct llama_vocab & vocab) {
     return vocab.special_pad_id;
 }
 
-int32_t llama_add_bos_token_impl(const struct llama_vocab & vocab) {
+bool llama_add_bos_token_impl(const struct llama_vocab & vocab) {
     return vocab.tokenizer_add_bos;
 }
 
-int32_t llama_add_eos_token_impl(const struct llama_vocab & vocab) {
+bool llama_add_eos_token_impl(const struct llama_vocab & vocab) {
     return vocab.tokenizer_add_eos;
 }
 

src/llama-vocab.h

@@ -95,8 +95,8 @@ llama_token llama_token_sep_impl(const struct llama_vocab & vocab);
 llama_token llama_token_nl_impl (const struct llama_vocab & vocab);
 llama_token llama_token_pad_impl(const struct llama_vocab & vocab);
 
-int32_t llama_add_bos_token_impl(const struct llama_vocab & vocab);
-int32_t llama_add_eos_token_impl(const struct llama_vocab & vocab);
+bool llama_add_bos_token_impl(const struct llama_vocab & vocab);
+bool llama_add_eos_token_impl(const struct llama_vocab & vocab);
 
 llama_token llama_token_prefix_impl(const struct llama_vocab & vocab);
 llama_token llama_token_middle_impl(const struct llama_vocab & vocab);

src/llama.cpp

@@ -210,6 +210,8 @@ enum llm_arch {
     LLM_ARCH_T5,
     LLM_ARCH_T5ENCODER,
     LLM_ARCH_JAIS,
+    LLM_ARCH_NEMOTRON,
+    LLM_ARCH_EXAONE,
     LLM_ARCH_UNKNOWN,
 };
 
@@ -255,6 +257,8 @@ static const std::map<llm_arch, const char *> LLM_ARCH_NAMES = {
     { LLM_ARCH_T5,              "t5"           },
     { LLM_ARCH_T5ENCODER,       "t5encoder"    },
     { LLM_ARCH_JAIS,            "jais"         },
+    { LLM_ARCH_NEMOTRON,        "nemotron"     },
+    { LLM_ARCH_EXAONE,          "exaone"       },
     { LLM_ARCH_UNKNOWN,         "(unknown)"    },
 };
 
@@ -1296,6 +1300,43 @@ static const std::map<llm_arch, std::map<llm_tensor, std::string>> LLM_TENSOR_NA
             { LLM_TENSOR_FFN_DOWN,        "blk.%d.ffn_down" },
         },
     },
+    {
+        LLM_ARCH_NEMOTRON,
+        {
+            { LLM_TENSOR_TOKEN_EMBD,      "token_embd" },
+            { LLM_TENSOR_OUTPUT_NORM,     "output_norm" },
+            { LLM_TENSOR_OUTPUT,          "output" },
+            { LLM_TENSOR_ROPE_FREQS,      "rope_freqs" },
+            { LLM_TENSOR_ATTN_NORM,       "blk.%d.attn_norm" },
+            { LLM_TENSOR_ATTN_Q,          "blk.%d.attn_q" },
+            { LLM_TENSOR_ATTN_K,          "blk.%d.attn_k" },
+            { LLM_TENSOR_ATTN_V,          "blk.%d.attn_v" },
+            { LLM_TENSOR_ATTN_OUT,        "blk.%d.attn_output" },
+            { LLM_TENSOR_ATTN_ROT_EMBD,   "blk.%d.attn_rot_embd" },
+            { LLM_TENSOR_FFN_NORM,        "blk.%d.ffn_norm" },
+            { LLM_TENSOR_FFN_DOWN,        "blk.%d.ffn_down" },
+            { LLM_TENSOR_FFN_UP,          "blk.%d.ffn_up" },
+        },
+    },
+    {
+        LLM_ARCH_EXAONE,
+        {
+            { LLM_TENSOR_TOKEN_EMBD,      "token_embd" },
+            { LLM_TENSOR_OUTPUT_NORM,     "output_norm" },
+            { LLM_TENSOR_OUTPUT,          "output" },
+            { LLM_TENSOR_ROPE_FREQS,      "rope_freqs" },
+            { LLM_TENSOR_ATTN_NORM,       "blk.%d.attn_norm" },
+            { LLM_TENSOR_ATTN_Q,          "blk.%d.attn_q" },
+            { LLM_TENSOR_ATTN_K,          "blk.%d.attn_k" },
+            { LLM_TENSOR_ATTN_V,          "blk.%d.attn_v" },
+            { LLM_TENSOR_ATTN_OUT,        "blk.%d.attn_output" },
+            { LLM_TENSOR_ATTN_ROT_EMBD,   "blk.%d.attn_rot_embd" },
+            { LLM_TENSOR_FFN_NORM,        "blk.%d.ffn_norm" },
+            { LLM_TENSOR_FFN_GATE,        "blk.%d.ffn_gate" },
+            { LLM_TENSOR_FFN_DOWN,        "blk.%d.ffn_down" },
+            { LLM_TENSOR_FFN_UP,          "blk.%d.ffn_up" },
+        },
+    },
     {
         LLM_ARCH_UNKNOWN,
         {
@@ -5235,6 +5276,23 @@ static void llm_load_hparams(
                     default: model.type = e_model::MODEL_UNKNOWN;
                 }
             } break;
+        case LLM_ARCH_NEMOTRON:
+            {
+                ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
+                switch (hparams.n_layer) {
+                    case 32: model.type = e_model::MODEL_4B; break;
+                    default: model.type = e_model::MODEL_UNKNOWN;
+                }
+            } break;
+        case LLM_ARCH_EXAONE:
+            {
+                ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+                switch (hparams.n_layer) {
+                    case 32: model.type = e_model::MODEL_8B; break;
+                    default: model.type = e_model::MODEL_UNKNOWN;
+                }
+            } break;
         default: (void)0;
     }
 
@@ -5467,6 +5525,15 @@ static void llm_load_vocab(
             } else if (
                 tokenizer_pre == "codeshell") {
                 vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_CODESHELL;
+            } else if (
+                tokenizer_pre == "bloom") {
+                vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_BLOOM;
+            } else if (
+                tokenizer_pre == "gpt3-finnish") {
+                vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_GPT3_FINNISH;
+            } else if (
+                tokenizer_pre == "exaone") {
+                vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_EXAONE;
             } else {
                 throw std::runtime_error(format("unknown pre-tokenizer type: '%s'", tokenizer_pre.c_str()));
             }
@@ -7562,6 +7629,78 @@ static bool llm_load_tensors(
                         layer.ffn_down   = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd});
                     }
                 } break;
+            case LLM_ARCH_NEMOTRON:
+                {
+                    model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
+
+                    // output
+                    {
+                        model.output_norm   = ml.create_tensor(ctx_output,   tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd});
+                        model.output_norm_b = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "bias"), {n_embd});
+                        model.output        = ml.create_tensor(ctx_output_split,  tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab});
+                    }
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        ggml_context * ctx_layer = ctx_for_layer(i);
+                        ggml_context * ctx_split = ctx_for_layer_split(i);
+
+                        auto & layer = model.layers[i];
+
+                        layer.attn_norm   = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd});
+                        layer.attn_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "bias", i), {n_embd});
+
+                        layer.wq = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q,   "weight", i), {n_embd, n_embd});
+                        layer.wk = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_K,   "weight", i), {n_embd, n_embd_gqa});
+                        layer.wv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_V,   "weight", i), {n_embd, n_embd_gqa});
+                        layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd});
+
+                        // optional bias tensors
+                        layer.bq = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_Q,   "bias", i), {n_embd},     llama_model_loader::TENSOR_NOT_REQUIRED);
+                        layer.bk = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_K,   "bias", i), {n_embd_gqa}, llama_model_loader::TENSOR_NOT_REQUIRED);
+                        layer.bv = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_V,   "bias", i), {n_embd_gqa}, llama_model_loader::TENSOR_NOT_REQUIRED);
+                        layer.bo = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd},     llama_model_loader::TENSOR_NOT_REQUIRED);
+
+                        layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd});
+                        layer.ffn_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "bias", i), {n_embd});
+
+                        layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd});
+                        layer.ffn_up   = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff});
+
+                        // optional MLP bias
+                        layer.ffn_down_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd}, llama_model_loader::TENSOR_NOT_REQUIRED);
+                        layer.ffn_up_b   = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_UP,   "bias", i), {n_ff}, llama_model_loader::TENSOR_NOT_REQUIRED);
+                    }
+                } break;
+            case LLM_ARCH_EXAONE:
+                {
+                    model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
+
+                    // output
+                    {
+                        model.output_norm = ml.create_tensor(ctx_output,       tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd});
+                        model.output      = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab});
+                    }
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        ggml_context * ctx_layer = ctx_for_layer(i);
+                        ggml_context * ctx_split = ctx_for_layer_split(i);
+
+                        auto & layer = model.layers[i];
+
+                        layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd});
+
+                        layer.wq = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q,   "weight", i), {n_embd, n_embd_head_k * n_head});
+                        layer.wk = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_K,   "weight", i), {n_embd, n_embd_k_gqa});
+                        layer.wv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_V,   "weight", i), {n_embd, n_embd_v_gqa});
+                        layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd});
+
+                        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));
+                        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});
+                        layer.ffn_up   = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff});
+                    }
+                } break;
             default:
                 throw std::runtime_error("unknown architecture");
         }
@@ -8248,7 +8387,7 @@ 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 || model.arch == LLM_ARCH_QWEN2) {
+        if (model.arch == LLM_ARCH_PHI2 || model.arch == LLM_ARCH_PHI3 || model.arch == LLM_ARCH_GPTNEOX || model.arch == LLM_ARCH_QWEN2 || model.arch == LLM_ARCH_NEMOTRON) {
             // 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);
@@ -13749,6 +13888,254 @@ struct llm_build_context {
 
         return gf;
     }
+
+    struct ggml_cgraph * build_nemotron() {
+        struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, llama_model_max_nodes(model), false);
+
+        const int64_t n_embd_head = hparams.n_embd_head_v;
+        GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
+        //GGML_ASSERT(n_embd_head == hparams.n_rot);
+
+        struct ggml_tensor * cur;
+        struct ggml_tensor * inpL;
+
+        inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
+
+        // inp_pos - contains the positions
+        struct ggml_tensor * inp_pos = build_inp_pos();
+
+        // KQ_mask (mask for 1 head, it will be broadcasted to all heads)
+        struct ggml_tensor * KQ_mask = build_inp_KQ_mask();
+
+        for (int il = 0; il < n_layer; ++il) {
+            struct ggml_tensor * inpSA = inpL;
+
+            // norm
+            cur = llm_build_norm(ctx0, inpL, hparams,
+                    model.layers[il].attn_norm,
+                    model.layers[il].attn_norm_b,
+                    LLM_NORM, cb, il);
+            cb(cur, "attn_norm", il);
+
+            // self-attention
+            {
+                // 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);
+                if (model.layers[il].bq) {
+                    Qcur = ggml_add(ctx0, Qcur, model.layers[il].bq);
+                    cb(Qcur, "Qcur", il);
+                }
+
+                struct ggml_tensor * Kcur = llm_build_lora_mm(lctx, ctx0, model.layers[il].wk, cur);
+                cb(Kcur, "Kcur", il);
+                if (model.layers[il].bk) {
+                    Kcur = ggml_add(ctx0, Kcur, model.layers[il].bk);
+                    cb(Kcur, "Kcur", il);
+                }
+
+                struct ggml_tensor * Vcur = llm_build_lora_mm(lctx, ctx0, model.layers[il].wv, cur);
+                cb(Vcur, "Vcur", il);
+                if (model.layers[il].bv) {
+                    Vcur = ggml_add(ctx0, Vcur, model.layers[il].bv);
+                    cb(Vcur, "Vcur", il);
+                }
+
+                Qcur = ggml_rope_ext(
+                    ctx0, ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens), inp_pos, nullptr,
+                    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,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                );
+                cb(Kcur, "Kcur", il);
+
+                cur = llm_build_kv(ctx0, lctx, kv_self, gf,
+                        model.layers[il].wo, model.layers[il].bo,
+                        Kcur, Vcur, Qcur, KQ_mask, n_tokens, kv_head, n_kv, 1.0f/sqrtf(float(n_embd_head)), cb, il);
+            }
+
+            if (il == n_layer - 1) {
+                // skip computing output for unused tokens
+                struct ggml_tensor * inp_out_ids = build_inp_out_ids();
+                cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+                inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+            }
+
+            struct ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+            cb(ffn_inp, "ffn_inp", il);
+
+            // feed-forward network
+            cur = llm_build_norm(ctx0, ffn_inp, hparams,
+                    model.layers[il].ffn_norm,
+                    model.layers[il].ffn_norm_b,
+                    LLM_NORM, cb, il);
+            cb(cur, "ffn_norm", il);
+
+            cur = llm_build_ffn(ctx0, lctx, cur,
+                    model.layers[il].ffn_up,   model.layers[il].ffn_up_b,   NULL,
+                    NULL,                      NULL,                        NULL,
+                    model.layers[il].ffn_down, model.layers[il].ffn_down_b, NULL,
+                    NULL,
+                    LLM_FFN_RELU_SQR, LLM_FFN_SEQ, cb, il);
+
+            cur = ggml_add(ctx0, cur, ffn_inp);
+            cb(cur, "ffn_out", il);
+
+            cur = lctx.cvec.apply_to(ctx0, cur, il);
+            cb(cur, "l_out", il);
+
+            // input for next layer
+            inpL = cur;
+        }
+
+        cur = inpL;
+
+        cur = llm_build_norm(ctx0, cur, hparams,
+                model.output_norm, model.output_norm_b,
+                LLM_NORM, cb, -1);
+        cb(cur, "result_norm", -1);
+
+        // lm_head
+        cur = llm_build_lora_mm(lctx, ctx0, model.output, cur);
+        cb(cur, "result_output", -1);
+
+        ggml_build_forward_expand(gf, cur);
+
+        return gf;
+    }
+
+    struct ggml_cgraph * build_exaone() {
+        struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, llama_model_max_nodes(model), false);
+
+        // mutable variable, needed during the last layer of the computation to skip unused tokens
+        int32_t n_tokens = this->n_tokens;
+
+        const int64_t n_embd_head = hparams.n_embd_head_v;
+        GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
+        GGML_ASSERT(n_embd_head == hparams.n_rot);
+
+        struct ggml_tensor * cur;
+        struct ggml_tensor * inpL;
+
+        inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
+
+        // inp_pos - contains the positions
+        struct ggml_tensor * inp_pos = build_inp_pos();
+
+        // KQ_mask (mask for 1 head, it will be broadcasted to all heads)
+        struct ggml_tensor * KQ_mask = build_inp_KQ_mask();
+
+        for (int il = 0; il < n_layer; ++il) {
+            struct ggml_tensor * inpSA = inpL;
+
+            // norm
+            cur = llm_build_norm(ctx0, inpL, hparams,
+                    model.layers[il].attn_norm, NULL,
+                    LLM_NORM_RMS, cb, il);
+            cb(cur, "attn_norm", il);
+
+            // 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);
+                if (model.layers[il].bq) {
+                    Qcur = ggml_add(ctx0, Qcur, model.layers[il].bq);
+                    cb(Qcur, "Qcur", il);
+                }
+
+                struct ggml_tensor * Kcur = llm_build_lora_mm(lctx, ctx0, model.layers[il].wk, cur);
+                cb(Kcur, "Kcur", il);
+                if (model.layers[il].bk) {
+                    Kcur = ggml_add(ctx0, Kcur, model.layers[il].bk);
+                    cb(Kcur, "Kcur", il);
+                }
+
+                struct ggml_tensor * Vcur = llm_build_lora_mm(lctx, ctx0, model.layers[il].wv, cur);
+                cb(Vcur, "Vcur", il);
+                if (model.layers[il].bv) {
+                    Vcur = ggml_add(ctx0, Vcur, model.layers[il].bv);
+                    cb(Vcur, "Vcur", il);
+                }
+
+                Qcur = ggml_rope_ext(
+                    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, rope_factors,
+                    n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                );
+                cb(Kcur, "Kcur", il);
+
+                cur = llm_build_kv(ctx0, lctx, kv_self, gf,
+                        model.layers[il].wo, model.layers[il].bo,
+                        Kcur, Vcur, Qcur, KQ_mask, n_tokens, kv_head, n_kv, 1.0f/sqrtf(float(n_embd_head)), cb, il);
+            }
+
+            if (il == n_layer - 1) {
+                // skip computing output for unused tokens
+                struct ggml_tensor * inp_out_ids = build_inp_out_ids();
+                n_tokens = n_outputs;
+                cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+                inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+            }
+
+            struct ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+            cb(ffn_inp, "ffn_inp", il);
+
+            // feed-forward network
+            cur = llm_build_norm(ctx0, ffn_inp, hparams,
+                    model.layers[il].ffn_norm, NULL,
+                    LLM_NORM_RMS, cb, il);
+            cb(cur, "ffn_norm", il);
+
+            cur = llm_build_ffn(ctx0, lctx, cur,
+                    model.layers[il].ffn_up,   NULL, NULL,
+                    model.layers[il].ffn_gate, NULL, NULL,
+                    model.layers[il].ffn_down, NULL, NULL,
+                    NULL,
+                    LLM_FFN_SILU, LLM_FFN_PAR, cb, il);
+            cb(cur, "ffn_out", il);
+
+            cur = ggml_add(ctx0, cur, ffn_inp);
+            cb(cur, "ffn_out", il);
+
+            cur = lctx.cvec.apply_to(ctx0, cur, il);
+            cb(cur, "l_out", il);
+
+            // input for next layer
+            inpL = cur;
+        }
+
+        cur = inpL;
+
+        cur = llm_build_norm(ctx0, cur, hparams,
+                model.output_norm, NULL,
+                LLM_NORM_RMS, cb, -1);
+        cb(cur, "result_norm", -1);
+
+        // lm_head
+        cur = llm_build_lora_mm(lctx, ctx0, model.output, cur);
+        cb(cur, "result_output", -1);
+
+        ggml_build_forward_expand(gf, cur);
+
+        return gf;
+    }
 };
 
 static struct ggml_cgraph * llama_build_graph_defrag(llama_context & lctx, const std::vector<uint32_t> & ids) {
@@ -14004,6 +14391,14 @@ static struct ggml_cgraph * llama_build_graph(
             {
                 result = llm.build_jais();
             } break;
+        case LLM_ARCH_NEMOTRON:
+            {
+                result = llm.build_nemotron();
+            } break;
+        case LLM_ARCH_EXAONE:
+            {
+                result = llm.build_exaone();
+            } break;
         default:
             GGML_ABORT("fatal error");
     }
@@ -17074,6 +17469,8 @@ enum llama_rope_type llama_rope_type(const struct llama_model * model) {
         case LLM_ARCH_OPENELM:
         case LLM_ARCH_GPTNEOX:
         case LLM_ARCH_CODESHELL:
+        case LLM_ARCH_NEMOTRON:
+        case LLM_ARCH_EXAONE:
             return LLAMA_ROPE_TYPE_NEOX;
 
         // all model arches should be listed explicitly here
@@ -18699,11 +19096,11 @@ llama_token llama_token_pad(const struct llama_model * model) {
     return llama_token_pad_impl(model->vocab);
 }
 
-int32_t llama_add_bos_token(const struct llama_model * model) {
+bool llama_add_bos_token(const struct llama_model * model) {
     return llama_add_bos_token_impl(model->vocab);
 }
 
-int32_t llama_add_eos_token(const struct llama_model * model) {
+bool llama_add_eos_token(const struct llama_model * model) {
     return llama_add_eos_token_impl(model->vocab);
 }
 
@@ -19004,6 +19401,22 @@ static int32_t llama_chat_apply_template_internal(
         if (add_ass) {
             ss << "Assistant:";
         }
+    } else if (tmpl == "exaone3" || (tmpl_contains("[|system|]") && tmpl_contains("[|assistant|]") && tmpl_contains("[|endofturn|]"))) {
+        // ref: https://huggingface.co/LGAI-EXAONE/EXAONE-3.0-7.8B-Instruct/discussions/8#66bae61b1893d14ee8ed85bb
+        // EXAONE-3.0-7.8B-Instruct
+        for (auto message : chat) {
+            std::string role(message->role);
+            if (role == "system") {
+                ss << "[|system|]" << trim(message->content) << "[|endofturn|]\n";
+            } else if (role == "user") {
+                ss << "[|user|]" << trim(message->content) << "\n";
+            } else if (role == "assistant") {
+                ss << "[|assistant|]" << trim(message->content) << "[|endofturn|]\n";
+            }
+        }
+        if (add_ass) {
+            ss << "[|assistant|]";
+        }
     } else {
         // template not supported
         return -1;