Merge branch 'ggml-org:master' into power-law-sampler

2025-12-15 09:07:13 -06:00 · 2025-12-15 09:07:13 -06:00 · 6e66095e1f
parent 4e04bd1ce2 165caaf5fb
commit 6e66095e1f
62 changed files with 2942 additions and 1085 deletions
--- a/.github/ISSUE_TEMPLATE/011-bug-results.yml
+++ b/.github/ISSUE_TEMPLATE/011-bug-results.yml
@ -11,7 +11,7 @@ body:
        (i.e. the generated text) are incorrect or llama.cpp crashes during model evaluation.
        If you encountered the issue while using an external UI (e.g. ollama),
        please reproduce your issue using one of the examples/binaries in this repository.
-        The `llama-cli` binary can be used for simple and reproducible model inference.
+        The `llama-completion` binary can be used for simple and reproducible model inference.
  - type: textarea
    id: version
    attributes:
@ -74,9 +74,12 @@ body:
        Please give us a summary of the problem and tell us how to reproduce it.
        If you can narrow down the bug to specific hardware, compile flags, or command line arguments,
        that information would be very much appreciated by us.
+
+        If possible, please try to reproduce the issue using `llama-completion` with `-fit off`.
+        If you can only reproduce the issue with `-fit on`, please provide logs both with and without `--verbose`.
      placeholder: >
-        e.g. when I run llama-cli with -ngl 99 I get garbled outputs.
-        When I use -ngl 0 it works correctly.
+        e.g. when I run llama-completion with `-fa on` I get garbled outputs for very long prompts.
+        With short prompts or `-fa off` it works correctly.
        Here are the exact commands that I used: ...
    validations:
      required: true
--- a/3
+++ b/3
@ -87,7 +87,8 @@
 /tests/                                 @ggerganov
 /tests/test-chat-.*                     @pwilkin
 /tools/batched-bench/                   @ggerganov
-/tools/main/                            @ggerganov
+/tools/cli/                             @ngxson
+/tools/completion/                      @ggerganov
 /tools/mtmd/                            @ngxson
 /tools/perplexity/                      @ggerganov
 /tools/quantize/                        @ggerganov
--- a/README.md
+++ b/README.md
@ -313,7 +313,7 @@ The Hugging Face platform provides a variety of online tools for converting, qua

 To learn more about model quantization, [read this documentation](tools/quantize/README.md)

-## [`llama-cli`](tools/main)
+## [`llama-cli`](tools/cli)

 #### A CLI tool for accessing and experimenting with most of `llama.cpp`'s functionality.

@ -525,7 +525,8 @@ To learn more about model quantization, [read this documentation](tools/quantize

 ## Other documentation

- [main (cli)](tools/main/README.md)
+- [cli](tools/cli/README.md)
+- [completion](tools/completion/README.md)
 - [server](tools/server/README.md)
 - [GBNF grammars](grammars/README.md)

--- a/ci/run.sh
+++ b/ci/run.sh
@ -398,6 +398,8 @@ function gg_run_qwen3_0_6b {
    ./bin/llama-quantize ${model_bf16} ${model_q5_k} q5_k $(nproc)
    ./bin/llama-quantize ${model_bf16} ${model_q6_k} q6_k $(nproc)

+    (time ./bin/llama-fit-params --model ${model_f16} 2>&1 | tee -a $OUT/${ci}-fp-f16.log)
+
    (time ./bin/llama-completion -no-cnv --model ${model_f16}  -ngl 99 -c 1024 -s 1234 -n 64 --ignore-eos -p "I believe the meaning of life is" ) 2>&1 | tee -a $OUT/${ci}-tg-f16.log
    (time ./bin/llama-completion -no-cnv --model ${model_bf16} -ngl 99 -c 1024 -s 1234 -n 64 --ignore-eos -p "I believe the meaning of life is" ) 2>&1 | tee -a $OUT/${ci}-tg-bf16.log
    (time ./bin/llama-completion -no-cnv --model ${model_q8_0} -ngl 99 -c 1024 -s 1234 -n 64 --ignore-eos -p "I believe the meaning of life is" ) 2>&1 | tee -a $OUT/${ci}-tg-q8_0.log
@ -523,6 +525,8 @@ function gg_run_embd_bge_small {

    ./bin/llama-quantize ${model_f16} ${model_q8_0} q8_0

+    (time ./bin/llama-fit-params --model ${model_f16} 2>&1 | tee -a $OUT/${ci}-fp-f16.log)
+
    (time ./bin/llama-embedding --model ${model_f16}  -p "I believe the meaning of life is" -ngl 99 -c 0 --no-op-offload) 2>&1 | tee -a $OUT/${ci}-tg-f16.log
    (time ./bin/llama-embedding --model ${model_q8_0} -p "I believe the meaning of life is" -ngl 99 -c 0 --no-op-offload) 2>&1 | tee -a $OUT/${ci}-tg-q8_0.log

@ -563,6 +567,8 @@ function gg_run_rerank_tiny {

    model_f16="${path_models}/ggml-model-f16.gguf"

+    (time ./bin/llama-fit-params --model ${model_f16} 2>&1 | tee -a $OUT/${ci}-fp-f16.log)
+
    # for this model, the SEP token is "</s>"
    (time ./bin/llama-embedding --model ${model_f16} -p "what is panda?\thi\nwhat is panda?\tit's a bear\nwhat is panda?\tThe giant panda (Ailuropoda melanoleuca), sometimes called a panda bear or simply panda, is a bear species endemic to China." -ngl 99 -c 0 --pooling rank --embd-normalize -1 --no-op-offload --verbose-prompt) 2>&1 | tee -a $OUT/${ci}-rk-f16.log

--- a/common/arg.cpp
+++ b/common/arg.cpp
@ -20,6 +20,7 @@
 #include <nlohmann/json.hpp>

 #include <algorithm>
+#include <cinttypes>
 #include <climits>
 #include <cstdarg>
 #include <fstream>
@ -529,7 +530,9 @@ static bool common_params_parse_ex(int argc, char ** argv, common_params_context
        params.kv_overrides.back().key[0] = 0;
    }

-    if (!params.tensor_buft_overrides.empty()) {
+    // pad tensor_buft_overrides for llama_params_fit:
+    const size_t ntbo = llama_max_tensor_buft_overrides();
+    while (params.tensor_buft_overrides.size() < ntbo) {
        params.tensor_buft_overrides.push_back({nullptr, nullptr});
    }

@ -2171,6 +2174,34 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
            }
        }
    ).set_env("LLAMA_ARG_MAIN_GPU"));
+    add_opt(common_arg(
+        { "-fit", "--fit" }, "[on|off]",
+        string_format("whether to adjust unset arguments to fit in device memory ('on' or 'off', default: '%s')", params.fit_params ? "on" : "off"),
+        [](common_params & params, const std::string & value) {
+            if (is_truthy(value)) {
+                params.fit_params = true;
+            } else if (is_falsey(value)) {
+                params.fit_params = false;
+            } else {
+                throw std::runtime_error(
+                    string_format("error: unkown value for --fit: '%s'\n", value.c_str()));
+            }
+        }
+    ).set_env("LLAMA_ARG_FIT"));
+    add_opt(common_arg(
+        { "-fitt", "--fit-target" }, "MiB",
+        string_format("target margin per device for --fit option, default: %zu", params.fit_params_target/(1024*1024)),
+        [](common_params & params, int value) {
+            params.fit_params_target = value * size_t(1024*1024);
+        }
+    ).set_env("LLAMA_ARG_FIT_TARGET"));
+    add_opt(common_arg(
+        { "-fitc", "--fit-ctx" }, "N",
+        string_format("minimum ctx size that can be set by --fit option, default: %" PRIu32, params.fit_params_min_ctx),
+        [](common_params & params, int value) {
+            params.fit_params_min_ctx = value;
+        }
+    ).set_env("LLAMA_ARG_FIT_CTX"));
    add_opt(common_arg(
        {"--check-tensors"},
        string_format("check model tensor data for invalid values (default: %s)", params.check_tensors ? "true" : "false"),
--- a/common/common.cpp
+++ b/common/common.cpp
@ -1088,7 +1088,15 @@ struct common_init_result::impl {

 common_init_result::common_init_result(common_params & params) :
    pimpl(new impl{}) {
-    const auto mparams = common_model_params_to_llama(params);
+    auto mparams = common_model_params_to_llama(params);
+    auto cparams = common_context_params_to_llama(params);
+
+    if (params.fit_params) {
+        LOG_INF("%s: fitting params to device memory, to report bugs during this step use -fit off (or --verbose if you can't)\n", __func__);
+        llama_params_fit(params.model.path.c_str(), &mparams, &cparams,
+            params.tensor_split, params.tensor_buft_overrides.data(), params.fit_params_target, params.fit_params_min_ctx,
+            params.verbosity >= 4 ? GGML_LOG_LEVEL_DEBUG : GGML_LOG_LEVEL_ERROR);
+    }

    llama_model * model = llama_model_load_from_file(params.model.path.c_str(), mparams);
    if (model == NULL) {
@ -1103,8 +1111,6 @@ common_init_result::common_init_result(common_params & params) :
    // TODO: fix naming
    common_init_sampler_from_model(model, params.sampling);

-    auto cparams = common_context_params_to_llama(params);
-
    if (params.sampling.ignore_eos && llama_vocab_eos(vocab) == LLAMA_TOKEN_NULL) {
        LOG_WRN("%s: warning: vocab does not have an EOS token, ignoring --ignore-eos\n", __func__);
        params.sampling.ignore_eos = false;
@ -1143,8 +1149,7 @@ common_init_result::common_init_result(common_params & params) :

    llama_context * lctx = llama_init_from_model(model, cparams);
    if (lctx == NULL) {
-        LOG_ERR("%s: failed to create context with model '%s', try reducing --n-gpu-layers if you're running out of VRAM\n",
-                __func__, params.model.path.c_str());
+        LOG_ERR("%s: failed to create context with model '%s'\n", __func__, params.model.path.c_str());
        return;
    }

@ -1176,15 +1181,13 @@ common_init_result_ptr common_init_from_params(common_params & params) {

    llama_model * model = res->model();
    if (model == NULL) {
-        LOG_ERR("%s: failed to load model '%s', try reducing --n-gpu-layers if you're running out of VRAM\n",
-            __func__, params.model.path.c_str());
+        LOG_ERR("%s: failed to load model '%s'\n", __func__, params.model.path.c_str());
        return res;
    }

    llama_context * lctx = res->context();
    if (lctx == NULL) {
-        LOG_ERR("%s: failed to create context with model '%s', try reducing --n-gpu-layers if you're running out of VRAM\n",
-            __func__, params.model.path.c_str());
+        LOG_ERR("%s: failed to create context with model '%s'\n", __func__, params.model.path.c_str());
        return res;
    }

--- a/common/common.h
+++ b/common/common.h
@ -99,6 +99,7 @@ enum llama_example {
    LLAMA_EXAMPLE_TTS,
    LLAMA_EXAMPLE_DIFFUSION,
    LLAMA_EXAMPLE_FINETUNE,
+    LLAMA_EXAMPLE_FIT_PARAMS,

    LLAMA_EXAMPLE_COUNT,
 };
@ -309,8 +310,8 @@ struct lr_opt {
 struct ggml_opt_optimizer_params common_opt_lr_pars(void * userdata);

 struct common_params {
-    int32_t n_predict             =    -1; // new tokens to predict
-    int32_t n_ctx                 =  4096; // context size
+    int32_t n_predict             =    -1; // max. number of new tokens to predict, -1 == no limit
+    int32_t n_ctx                 =     0; // context size, 0 == context the model was trained with
    int32_t n_batch               =  2048; // logical batch size for prompt processing (must be >=32 to use BLAS)
    int32_t n_ubatch              =   512; // physical batch size for prompt processing (must be >=32 to use BLAS)
    int32_t n_keep                =     0; // number of tokens to keep from initial prompt
@ -331,9 +332,12 @@ struct common_params {
    // offload params
    std::vector<ggml_backend_dev_t> devices; // devices to use for offloading

-    int32_t n_gpu_layers      = -1;  // number of layers to store in VRAM (-1 - use default)
-    int32_t main_gpu          = 0;   // the GPU that is used for scratch and small tensors
-    float   tensor_split[128] = {0}; // how split tensors should be distributed across GPUs
+    int32_t n_gpu_layers       = -1;               // number of layers to store in VRAM (-1 - use default)
+    int32_t main_gpu           = 0;                // the GPU that is used for scratch and small tensors
+    float   tensor_split[128]  = {0};              // how split tensors should be distributed across GPUs
+    bool    fit_params         = true;             // whether to fit unset model/context parameters to free device memory
+    size_t  fit_params_target  = 1024 * 1024*1024; // margin per device in bytes for fitting parameters to free memory
+    int32_t fit_params_min_ctx = 4096;             // minimum context size to set when trying to reduce memory use

    enum llama_split_mode split_mode = LLAMA_SPLIT_MODE_LAYER; // how to split the model across GPUs

--- a/convert_hf_to_gguf.py
+++ b/convert_hf_to_gguf.py
@ -713,6 +713,9 @@ class ModelBase:
        if "llm_config" in config:
            # rename for InternVL
            config["text_config"] = config["llm_config"]
+        if "lm_config" in config:
+            # rename for GlmASR
+            config["text_config"] = config["lm_config"]
        if "thinker_config" in config:
            # rename for Qwen2.5-Omni
            config["text_config"] = config["thinker_config"]["text_config"]
@ -1529,6 +1532,21 @@ class TextModel(ModelBase):
                raise NotImplementedError("Only MEAN, CLS, and LAST pooling types supported")
            self.gguf_writer.add_pooling_type(pooling_type)

+    def _set_vocab_glmedge(self):
+        from transformers import AutoTokenizer
+        tokenizer = AutoTokenizer.from_pretrained(self.dir_model)
+        special_vocab = gguf.SpecialVocab(self.dir_model, load_merges=True)
+        tokens, toktypes, tokpre = self.get_vocab_base()
+        self.gguf_writer.add_tokenizer_model("gpt2")
+        self.gguf_writer.add_tokenizer_pre(tokpre)
+        self.gguf_writer.add_token_list(tokens)
+        self.gguf_writer.add_token_types(toktypes)
+        special_vocab._set_special_token("eos", tokenizer.get_added_vocab()["<|endoftext|>"])
+        special_vocab._set_special_token("eot", tokenizer.get_added_vocab()["<|user|>"])
+        special_vocab._set_special_token("unk", tokenizer.get_added_vocab()["<|endoftext|>"])
+        special_vocab._set_special_token("bos", tokenizer.get_added_vocab()["<|endoftext|>"])
+        special_vocab.add_to_gguf(self.gguf_writer)
+
    def _set_vocab_interns1(self):
        tokens: list[str] = []
        toktypes: list[int] = []
@ -1658,7 +1676,7 @@ class MmprojModel(ModelBase):
    preprocessor_config: dict[str, Any]
    global_config: dict[str, Any]

-    n_block_keys = ["n_layers", "num_hidden_layers", "n_layer", "num_layers", "depth"]
+    n_block_keys = ["n_layers", "num_hidden_layers", "n_layer", "num_layers", "depth", "encoder_layers"]

    has_vision_encoder: bool = True # by default
    has_audio_encoder: bool = False
@ -1734,7 +1752,8 @@ class MmprojModel(ModelBase):
        return self.global_config.get(config_name)

    def get_audio_config(self) -> dict[str, Any] | None:
-        return self.global_config.get("audio_config")
+        mm_config_key = "whisper_config" if "whisper_config" in self.hparams else "audio_config"
+        return self.global_config.get(mm_config_key)

    def set_type(self):
        self.gguf_writer.add_type(gguf.GGUFType.MMPROJ)
@ -2372,8 +2391,13 @@ class LlamaModel(TextModel):
        # fix for SmolVLM2, missing `num_attention_heads` in config.json
        if self.hf_arch == "VLlama3ForCausalLM":
            self.hparams["num_attention_heads"] = self.hparams.get("num_attention_heads", 32)
+        hparams = ModelBase.load_hparams(self.dir_model, is_mistral_format=False)
+        self.origin_hf_arch = hparams.get('architectures', [None])[0]

    def set_vocab(self):
+        if self.origin_hf_arch == "GlmasrModel":
+            return self._set_vocab_glmedge()
+
        if self.is_mistral_format:
            return self._set_vocab_mistral()

@ -2444,6 +2468,7 @@ class LlamaModel(TextModel):
            "vision_language_adapter.",
            "patch_merger.",
            "pre_mm_projector_norm",
+            "audio_encoder.",
        ]

        is_multimodal_tensor = "vision_tower" in name \
@ -8846,6 +8871,63 @@ class UltravoxModel(TextModel):
        raise NotImplementedError("Ultravox does not have text decoder. Instead, it uses Llama or other models for text. If you want to get the audio encoder, please use --mmproj argument")


+@ModelBase.register("GlmasrModel")
+class GlmASRWhisperEncoderModel(MmprojModel):
+    has_vision_encoder = False
+    has_audio_encoder = True
+
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        if "hidden_size" not in self.hparams and "intermediate_size" not in self.hparams:
+            self.hparams["hidden_size"] = self.hparams["d_model"]
+            self.hparams["intermediate_size"] = self.hparams["encoder_ffn_dim"]
+            self.hparams["num_attention_heads"] = self.hparams["encoder_attention_heads"]
+
+    def set_gguf_parameters(self):
+        super().set_gguf_parameters()
+        self.gguf_writer.add_clip_projector_type(gguf.VisionProjectorType.GLMA)
+        self.gguf_writer.add_audio_num_mel_bins(self.hparams["num_mel_bins"])
+        self.gguf_writer.add_audio_attention_layernorm_eps(self.hparams.get("layer_norm_eps", 1e-5))
+        self.gguf_writer.add_audio_stack_factor(self.global_config["merge_factor"])
+
+    def tensor_force_quant(self, name, new_name, bid, n_dims):
+        if ".conv" in name and ".weight" in name:
+            return gguf.GGMLQuantizationType.F16
+        return super().tensor_force_quant(name, new_name, bid, n_dims)
+
+    def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iterable[tuple[str, Tensor]]:
+        del bid  # unused
+
+        if name.startswith("model.") or name.startswith("lm_head."):
+            # skip language model tensors
+            return []
+
+        if name.startswith("audio_encoder.whisper."):
+            name = name.replace("audio_encoder.whisper.","audio_tower.")
+        if "audio_encoder.layer_norm." in name or "audio_encoder.proj." in name:
+            name = name.replace("audio_encoder.", "audio_encoder.adapting.")
+
+        if name.startswith("audio_encoder.audio_bos_eos_token."):
+            return [(self.map_tensor_name("model.vision.boi"), data_torch[0]), (self.map_tensor_name("model.vision.eoi"), data_torch[1])]
+
+        if name.startswith("audio_encoder.adapting."):
+            name = name.replace("audio_encoder.adapting.","audio.multi_modal_projector.")
+            if ".layer_norm." in name:
+                name = name.replace(".layer_norm.", ".ln_pre.")
+            if ".0." in name:
+                name = name.replace(".0.", ".linear_1.")
+            if ".2." in name:
+                name = name.replace(".2.", ".linear_2.")
+            if ".proj." in name:
+                return []
+
+        if "conv1.bias" in name or "conv2.bias" in name:
+            # transpose conv1 and conv2 bias
+            data_torch = data_torch.unsqueeze(-1)
+
+        return [(self.map_tensor_name(name), data_torch)]
+
+
@ModelBase.register("Qwen2AudioForConditionalGeneration")
 class WhisperEncoderModel(MmprojModel):
    has_vision_encoder = False # no vision encoder
--- a/docs/development/HOWTO-add-model.md
+++ b/docs/development/HOWTO-add-model.md
@ -9,7 +9,8 @@ Adding a model requires few steps:
 After following these steps, you can open PR.

 Also, it is important to check that the examples and main ggml backends (CUDA, METAL, CPU) are working with the new architecture, especially:
- [main](/tools/main/)
+- [cli](/tools/cli/)
+- [completion](/tools/completion/)
 - [imatrix](/tools/imatrix/)
 - [quantize](/tools/quantize/)
 - [server](/tools/server/)
--- a/docs/ops.md
+++ b/docs/ops.md
@ -18,12 +18,12 @@ Legend:
 |                              ACC | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ |
 |                              ADD | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ |
 |                             ADD1 | ❌ | ✅ | ✅ | ✅ | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ |
-|                           ADD_ID | ❌ | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ✅ | ❌ | ❌ | ❌ |
+|                           ADD_ID | ❌ | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ | ❌ |
 |                           ARANGE | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ |
 |                           ARGMAX | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ |
-|                          ARGSORT | ❌ | ✅ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | ❌ | ❌ | ❌ |
+|                          ARGSORT | ❌ | ✅ | ✅ | ✅ | ✅ | 🟡 | 🟡 | ✅ | ❌ | ❌ | ❌ |
 |                             CEIL | ❌ | ❌ | ✅ | 🟡 | ❌ | ❌ | 🟡 | 🟡 | ❌ | ❌ | ❌ |
-|                            CLAMP | ❌ | ✅ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | 🟡 | ❌ | ❌ | ❌ |
+|                            CLAMP | ❌ | ✅ | ✅ | ✅ | 🟡 | 🟡 | ✅ | 🟡 | ❌ | ❌ | ❌ |
 |                           CONCAT | ❌ | ✅ | ✅ | 🟡 | ✅ | 🟡 | ✅ | ✅ | ❌ | ❌ | ❌ |
 |                             CONT | ❌ | 🟡 | ✅ | ✅ | ✅ | 🟡 | 🟡 | ✅ | 🟡 | ❌ | ❌ |
 |                          CONV_2D | ❌ | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ✅ | ❌ | ❌ | ❌ |
@ -31,7 +31,7 @@ Legend:
 |                          CONV_3D | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
 |                CONV_TRANSPOSE_1D | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ |
 |                CONV_TRANSPOSE_2D | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ |
-|                              COS | ❌ | ✅ | ✅ | ✅ | 🟡 | ❌ | 🟡 | 🟡 | ❌ | ❌ | ❌ |
+|                              COS | ❌ | ✅ | ✅ | ✅ | 🟡 | ❌ | ✅ | 🟡 | ❌ | ❌ | ❌ |
 |                      COUNT_EQUAL | ❌ | ✅ | ✅ | ✅ | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ |
 |                              CPY | ❌ | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | ❌ | ❌ |
 |               CROSS_ENTROPY_LOSS | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
@ -64,7 +64,7 @@ Legend:
 |                        IM2COL_3D | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ |
 |                          L2_NORM | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ |
 |                       LEAKY_RELU | ❌ | ✅ | ✅ | ✅ | 🟡 | ❌ | ✅ | 🟡 | ❌ | ❌ | ❌ |
-|                              LOG | ❌ | ✅ | ✅ | ✅ | 🟡 | ❌ | 🟡 | ✅ | ❌ | ❌ | ❌ |
+|                              LOG | ❌ | ✅ | ✅ | ✅ | 🟡 | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ |
 |                             MEAN | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ | ❌ |
 |                              MUL | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ |
 |                          MUL_MAT | 🟡 | 🟡 | 🟡 | 🟡 | ✅ | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 |
@ -98,14 +98,14 @@ Legend:
 |                          SIGMOID | ❌ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | ✅ | 🟡 | ✅ | ❌ | ❌ |
 |                             SILU | ❌ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | ✅ | 🟡 | ✅ | ❌ | ❌ |
 |                        SILU_BACK | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ |
-|                              SIN | ❌ | ✅ | ✅ | ✅ | 🟡 | ❌ | 🟡 | 🟡 | ❌ | ❌ | ❌ |
+|                              SIN | ❌ | ✅ | ✅ | ✅ | 🟡 | ❌ | ✅ | 🟡 | ❌ | ❌ | ❌ |
 |                          SOFTCAP | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
 |                         SOFTPLUS | ❌ | ❌ | ✅ | 🟡 | 🟡 | ❌ | ❌ | 🟡 | ❌ | ❌ | ❌ |
 |                         SOFT_MAX | ❌ | 🟡 | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ |
 |                    SOFT_MAX_BACK | ❌ | ❌ | 🟡 | 🟡 | ❌ | ❌ | 🟡 | ✅ | ❌ | ❌ | ❌ |
 |                        SOLVE_TRI | ❌ | ❌ | ✅ | 🟡 | ❌ | ❌ | ❌ | 🟡 | ❌ | ❌ | ❌ |
-|                              SQR | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | 🟡 | 🟡 | ❌ | ❌ | ❌ |
-|                             SQRT | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | 🟡 | 🟡 | ❌ | ❌ | ❌ |
+|                              SQR | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | 🟡 | ❌ | ❌ | ❌ |
+|                             SQRT | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | 🟡 | ❌ | ❌ | ❌ |
 |                         SSM_CONV | ❌ | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ | ❌ |
 |                         SSM_SCAN | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | ❌ | 🟡 | ❌ | ❌ | ❌ |
 |                             STEP | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | ✅ | 🟡 | ✅ | ❌ | ❌ |
@ -113,7 +113,7 @@ Legend:
 |                              SUM | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | 🟡 | 🟡 | ❌ | ❌ | ❌ |
 |                         SUM_ROWS | ❌ | ✅ | ✅ | 🟡 | ✅ | 🟡 | 🟡 | ✅ | ❌ | ❌ | ❌ |
 |                           SWIGLU | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | 🟡 | ✅ | ❌ | ❌ |
-|                       SWIGLU_OAI | ❌ | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | 🟡 | ✅ | ❌ | ❌ |
+|                       SWIGLU_OAI | ❌ | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | 🟡 | ✅ | ❌ | ❌ |
 |                             TANH | ❌ | ✅ | ✅ | 🟡 | 🟡 | ✅ | ✅ | 🟡 | ✅ | ❌ | ❌ |
 |               TIMESTEP_EMBEDDING | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ | ❌ |
 |                            TOP_K | ❌ | ❌ | ✅ | ❌ | ✅ | ❌ | ❌ | 🟡 | ❌ | ❌ | ❌ |
--- a/docs/ops/SYCL.csv
+++ b/docs/ops/SYCL.csv
--- a/ggml/include/ggml-alloc.h
+++ b/ggml/include/ggml-alloc.h
@ -53,7 +53,14 @@ GGML_API void           ggml_gallocr_free(ggml_gallocr_t galloc);
 // call with a worst-case graph to avoid buffer reallocations
 // not strictly required for single buffer usage: ggml_gallocr_alloc_graph will reallocate the buffers automatically if needed
 // returns false if the buffer allocation failed
+// ggml_gallocr_resrve_n_size writes the buffer sizes per galloc buffer that would be allocated by ggml_gallocr_reserve_n to sizes
 GGML_API bool ggml_gallocr_reserve(ggml_gallocr_t galloc, struct ggml_cgraph * graph);
+GGML_API void ggml_gallocr_reserve_n_size(
+    ggml_gallocr_t galloc,
+    struct ggml_cgraph * graph,
+    const int * node_buffer_ids,
+    const int * leaf_buffer_ids,
+    size_t * sizes);
 GGML_API bool ggml_gallocr_reserve_n(
    ggml_gallocr_t galloc,
    struct ggml_cgraph * graph,
@ -68,6 +75,8 @@ GGML_API size_t ggml_gallocr_get_buffer_size(ggml_gallocr_t galloc, int buffer_i

 // Utils
 // Create a buffer and allocate all the tensors in a ggml_context
+// ggml_backend_alloc_ctx_tensors_from_buft_size returns the size of the buffer that would be allocated by ggml_backend_alloc_ctx_tensors_from_buft
+GGML_API size_t                       ggml_backend_alloc_ctx_tensors_from_buft_size(struct ggml_context * ctx, ggml_backend_buffer_type_t buft);
 GGML_API struct ggml_backend_buffer * ggml_backend_alloc_ctx_tensors_from_buft(struct ggml_context * ctx, ggml_backend_buffer_type_t buft);
 GGML_API struct ggml_backend_buffer * ggml_backend_alloc_ctx_tensors(struct ggml_context * ctx, ggml_backend_t backend);

--- a/ggml/include/ggml-backend.h
+++ b/ggml/include/ggml-backend.h
@ -307,6 +307,7 @@ extern "C" {
    GGML_API void                 ggml_backend_sched_free(ggml_backend_sched_t sched);

    // Initialize backend buffers from a measure graph
+    GGML_API void                 ggml_backend_sched_reserve_size(ggml_backend_sched_t sched, struct ggml_cgraph * measure_graph, size_t * sizes);
    GGML_API bool                 ggml_backend_sched_reserve(ggml_backend_sched_t sched, struct ggml_cgraph * measure_graph); // returns success

    GGML_API int                  ggml_backend_sched_get_n_backends(ggml_backend_sched_t sched);
--- a/ggml/include/ggml.h
+++ b/ggml/include/ggml.h
@ -2615,7 +2615,8 @@ extern "C" {

    // Set callback for all future logging events.
    // If this is not called, or NULL is supplied, everything is output on stderr.
-    GGML_API void ggml_log_set(ggml_log_callback log_callback, void * user_data);
+    GGML_API void ggml_log_get(ggml_log_callback * log_callback, void ** user_data);
+    GGML_API void ggml_log_set(ggml_log_callback   log_callback, void *  user_data);

    GGML_API struct ggml_tensor * ggml_set_zero(struct ggml_tensor * tensor);

--- a/ggml/src/ggml-alloc.c
+++ b/ggml/src/ggml-alloc.c
@ -594,7 +594,9 @@ static bool ggml_gallocr_is_own(ggml_gallocr_t galloc, struct ggml_tensor * t) {
 }

 static bool ggml_gallocr_is_allocated(ggml_gallocr_t galloc, struct ggml_tensor * t) {
-    return t->data != NULL || ggml_gallocr_hash_get(galloc, t)->allocated;
+    return t->data != NULL // tensor data already set externally
+        || t->buffer // tensor on external buffer (but not yet allocated)
+        || ggml_gallocr_is_own(galloc, t); // tensor will be allocated by galloc
 }

 // free the extra space at the end if the new tensor is smaller
@ -823,7 +825,8 @@ static void ggml_gallocr_alloc_graph_impl(ggml_gallocr_t galloc, struct ggml_cgr
    }
 }

-bool ggml_gallocr_reserve_n(ggml_gallocr_t galloc, struct ggml_cgraph * graph, const int * node_buffer_ids, const int * leaf_buffer_ids) {
+static bool ggml_gallocr_reserve_n_impl(
+        ggml_gallocr_t galloc, struct ggml_cgraph * graph, const int * node_buffer_ids, const int * leaf_buffer_ids, bool no_alloc) {
    size_t min_hash_size = graph->n_nodes + graph->n_leafs;
    // add 25% margin to avoid hash collisions
    min_hash_size += min_hash_size / 4;
@ -928,16 +931,19 @@ bool ggml_gallocr_reserve_n(ggml_gallocr_t galloc, struct ggml_cgraph * graph, c
                size_t cur_size = galloc->buffers[i] ? ggml_vbuffer_size(galloc->buffers[i]) : 0;
                if (cur_size > 0) {
                    GGML_LOG_DEBUG("%s: reallocating %s buffer from size %.02f MiB to %.02f MiB\n",
-                        __func__, ggml_backend_buft_name(galloc->bufts[i]),
-                        cur_size / 1024.0 / 1024.0, new_size / 1024.0 / 1024.0);
+                        __func__, ggml_backend_buft_name(galloc->bufts[i]), cur_size / 1024.0 / 1024.0, new_size / 1024.0 / 1024.0);
                }
            }
 #endif
            ggml_vbuffer_free(galloc->buffers[i]);
-            galloc->buffers[i] = ggml_vbuffer_alloc(galloc->bufts[i], galloc->buf_tallocs[i], GGML_BACKEND_BUFFER_USAGE_COMPUTE);
-            if (galloc->buffers[i] == NULL) {
-                GGML_LOG_ERROR("%s: failed to allocate %s buffer of size %zu\n", __func__, ggml_backend_buft_name(galloc->bufts[i]), new_size);
-                return false;
+            if (no_alloc) {
+                galloc->buffers[i] = NULL;
+            } else {
+                galloc->buffers[i] = ggml_vbuffer_alloc(galloc->bufts[i], galloc->buf_tallocs[i], GGML_BACKEND_BUFFER_USAGE_COMPUTE);
+                if (galloc->buffers[i] == NULL) {
+                    GGML_LOG_ERROR("%s: failed to allocate %s buffer of size %zu\n", __func__, ggml_backend_buft_name(galloc->bufts[i]), new_size);
+                    return false;
+                }
            }
        }
    }
@ -945,6 +951,21 @@ bool ggml_gallocr_reserve_n(ggml_gallocr_t galloc, struct ggml_cgraph * graph, c
    return true;
 }

+void ggml_gallocr_reserve_n_size(
+        ggml_gallocr_t galloc, struct ggml_cgraph * graph, const int * node_buffer_ids, const int * leaf_buffer_ids, size_t * sizes) {
+    GGML_ASSERT(ggml_gallocr_reserve_n_impl(galloc, graph, node_buffer_ids, leaf_buffer_ids, /*no_alloc =*/ true));
+    for (int i = 0; i < galloc->n_buffers; i++) {
+        sizes[i] = 0;
+        for (int c = 0; c < galloc->buf_tallocs[i]->n_chunks; c++) {
+            sizes[i] += galloc->buf_tallocs[i]->chunks[c]->max_size;
+        }
+    }
+}
+
+bool ggml_gallocr_reserve_n(ggml_gallocr_t galloc, struct ggml_cgraph * graph, const int * node_buffer_ids, const int * leaf_buffer_ids) {
+    return ggml_gallocr_reserve_n_impl(galloc, graph, node_buffer_ids, leaf_buffer_ids, /*no_alloc =*/ false);
+}
+
 bool ggml_gallocr_reserve(ggml_gallocr_t galloc, struct ggml_cgraph *graph) {
    return ggml_gallocr_reserve_n(galloc, graph, NULL, NULL);
 }
@ -1147,7 +1168,8 @@ static bool alloc_tensor_range(struct ggml_context * ctx,
    return true;
 }

-ggml_backend_buffer_t ggml_backend_alloc_ctx_tensors_from_buft(struct ggml_context * ctx, ggml_backend_buffer_type_t buft) {
+static ggml_backend_buffer_t ggml_backend_alloc_ctx_tensors_from_buft_impl(
+        struct ggml_context * ctx, ggml_backend_buffer_type_t buft, size_t * nbytes_total, bool no_alloc) {
    GGML_ASSERT(ggml_get_no_alloc(ctx) == true);

    size_t alignment = ggml_backend_buft_get_alignment(buft);
@ -1155,6 +1177,7 @@ ggml_backend_buffer_t ggml_backend_alloc_ctx_tensors_from_buft(struct ggml_conte

    ggml_backend_buffer_t * buffers = NULL;
    size_t n_buffers = 0;
+    *nbytes_total = 0;

    size_t cur_buf_size = 0;
    struct ggml_tensor * first = ggml_get_first_tensor(ctx);
@ -1166,10 +1189,11 @@ ggml_backend_buffer_t ggml_backend_alloc_ctx_tensors_from_buft(struct ggml_conte

        if (cur_buf_size > 0 && (cur_buf_size + this_size) > max_size) {
            // allocate tensors in the current buffer
-            if (!alloc_tensor_range(ctx, first, t, buft, cur_buf_size, &buffers, &n_buffers)) {
+            if (!no_alloc && !alloc_tensor_range(ctx, first, t, buft, cur_buf_size, &buffers, &n_buffers)) {
                return NULL;
            }
            first = t;
+            *nbytes_total += cur_buf_size;
            cur_buf_size = this_size;
        } else {
            cur_buf_size += this_size;
@ -1178,15 +1202,21 @@ ggml_backend_buffer_t ggml_backend_alloc_ctx_tensors_from_buft(struct ggml_conte

    // allocate remaining tensors
    if (cur_buf_size > 0) {
-        if (!alloc_tensor_range(ctx, first, NULL, buft, cur_buf_size, &buffers, &n_buffers)) {
+        *nbytes_total += cur_buf_size;
+        if (!no_alloc && !alloc_tensor_range(ctx, first, NULL, buft, cur_buf_size, &buffers, &n_buffers)) {
            return NULL;
        }
    }

+    if (no_alloc) {
+        return NULL;
+    }
+
    if (n_buffers == 0) {
 #ifndef NDEBUG
        GGML_LOG_DEBUG("%s: all tensors in the context are already allocated\n", __func__);
 #endif
+        GGML_ASSERT(!buffers);
        return NULL;
    }

@ -1196,10 +1226,24 @@ ggml_backend_buffer_t ggml_backend_alloc_ctx_tensors_from_buft(struct ggml_conte
    } else {
        buffer = ggml_backend_multi_buffer_alloc_buffer(buffers, n_buffers);
    }
-    free(buffers);
+    if (buffers) {
+        free(buffers); // can be NULL if context is empty or no_alloc
+    }
    return buffer;
 }

+size_t ggml_backend_alloc_ctx_tensors_from_buft_size(struct ggml_context * ctx, ggml_backend_buffer_type_t buft) {
+    size_t nbytes_total = 0;
+    ggml_backend_buffer_t buf = ggml_backend_alloc_ctx_tensors_from_buft_impl(ctx, buft, &nbytes_total, /*no_alloc=*/ true);
+    GGML_ASSERT(!buf);
+    return nbytes_total;
+}
+
+ggml_backend_buffer_t ggml_backend_alloc_ctx_tensors_from_buft(struct ggml_context * ctx, ggml_backend_buffer_type_t buft) {
+    size_t nbytes_total = 0;
+    return ggml_backend_alloc_ctx_tensors_from_buft_impl(ctx, buft, &nbytes_total, /*no_alloc =*/ false);
+}
+
 ggml_backend_buffer_t ggml_backend_alloc_ctx_tensors(struct ggml_context * ctx, ggml_backend_t backend) {
    return ggml_backend_alloc_ctx_tensors_from_buft(ctx, ggml_backend_get_default_buffer_type(backend));
 }
--- a/ggml/src/ggml-backend.cpp
+++ b/ggml/src/ggml-backend.cpp
@ -36,12 +36,11 @@ const char * ggml_backend_buft_name(ggml_backend_buffer_type_t buft) {
 }

 ggml_backend_buffer_t ggml_backend_buft_alloc_buffer(ggml_backend_buffer_type_t buft, size_t size) {
+    GGML_ASSERT(buft);
    if (size == 0) {
        // return a dummy buffer for zero-sized allocations
        return ggml_backend_buffer_init(buft, {}, NULL, 0);
    }
-
-    GGML_ASSERT(buft);
    return buft->iface.alloc_buffer(buft, size);
 }

@ -128,6 +127,12 @@ void * ggml_backend_buffer_get_base(ggml_backend_buffer_t buffer) {
        return NULL;
    }

+    // FIXME JG: a multi_buffer has a non-zero size, according to the above comment get_base is not optional,
+    //     I don't know whether the above comment is correct
+    if (!buffer->iface.get_base) {
+        return NULL;
+    }
+
    void * base = buffer->iface.get_base(buffer);

    GGML_ASSERT(base != NULL && "backend buffer base cannot be NULL");
@ -1727,6 +1732,20 @@ void ggml_backend_sched_reset(ggml_backend_sched_t sched) {
    sched->is_alloc = false;
 }

+void ggml_backend_sched_reserve_size(ggml_backend_sched_t sched, struct ggml_cgraph * measure_graph, size_t * sizes) {
+    GGML_ASSERT(sched);
+    GGML_ASSERT((int)sched->hash_set.size >= measure_graph->n_nodes + measure_graph->n_leafs);
+    GGML_ASSERT(sizes);
+
+    ggml_backend_sched_reset(sched);
+
+    ggml_backend_sched_synchronize(sched);
+
+    ggml_backend_sched_split_graph(sched, measure_graph);
+
+    ggml_gallocr_reserve_n_size(sched->galloc, &sched->graph, sched->node_backend_ids, sched->leaf_backend_ids, sizes);
+}
+
 bool ggml_backend_sched_reserve(ggml_backend_sched_t sched, struct ggml_cgraph * measure_graph) {
    GGML_ASSERT(sched);
    GGML_ASSERT((int)sched->hash_set.size >= measure_graph->n_nodes + measure_graph->n_leafs);
--- a/ggml/src/ggml-metal/ggml-metal-device.m
+++ b/ggml/src/ggml-metal/ggml-metal-device.m
@ -769,9 +769,16 @@ ggml_metal_device_t ggml_metal_device_init(void) {
 #endif

            dev->props.use_shared_buffers = dev->props.has_unified_memory;
+#if TARGET_OS_OSX
+            // In case of eGPU, shared memory may be preferable.
+            dev->props.use_shared_buffers |= [dev->mtl_device location] == MTLDeviceLocationExternal;
+#endif
            if (getenv("GGML_METAL_SHARED_BUFFERS_DISABLE") != NULL) {
                dev->props.use_shared_buffers = false;
            }
+            if (getenv("GGML_METAL_SHARED_BUFFERS_ENABLE") != NULL) {
+                dev->props.use_shared_buffers = true;
+            }

            dev->props.supports_gpu_family_apple7 = [dev->mtl_device supportsFamily:MTLGPUFamilyApple7];

--- a/ggml/src/ggml-sycl/add-id.cpp
+++ b/ggml/src/ggml-sycl/add-id.cpp
@ -0,0 +1,77 @@
+#include <sycl/sycl.hpp>
+#include "common.hpp"
+#include "add-id.hpp"
+
+static void add_id_kernel(
+    const float* src0,
+    const float* src1,
+    const int32_t* src2,
+    float* dst,
+    int64_t ne0,
+    int64_t ne1,
+    size_t nb01,
+    size_t nb02,
+    size_t nb11,
+    size_t nb21,
+    sycl::nd_item<3> item_ct1) {
+  const int64_t i1 = item_ct1.get_group(2);
+  const int64_t i2 = item_ct1.get_group(1);
+
+  const int i11 =
+      *(const int32_t*)((const char*)src2 + i1 * sizeof(int32_t) + i2 * nb21);
+
+  const size_t nb1 = ne0 * sizeof(float);
+  const size_t nb2 = ne1 * nb1;
+
+  float* dst_row = (float*)((char*)dst + i1 * nb1 + i2 * nb2);
+  const float* src0_row =
+      (const float*)((const char*)src0 + i1 * nb01 + i2 * nb02);
+  const float* src1_row = (const float*)((const char*)src1 + i11 * nb11);
+
+  for (int64_t i0 = item_ct1.get_local_id(2); i0 < ne0;
+       i0 += item_ct1.get_local_range(2)) {
+    dst_row[i0] = src0_row[i0] + src1_row[i0];
+  }
+}
+
+void ggml_sycl_add_id(ggml_backend_sycl_context& ctx, ggml_tensor* dst) {
+  const ggml_tensor* src0 = dst->src[0];
+  const ggml_tensor* src1 = dst->src[1];
+  const ggml_tensor* src2 = dst->src[2];
+
+  GGML_TENSOR_TERNARY_OP_LOCALS
+
+  GGML_ASSERT(dst->type == GGML_TYPE_F32);
+  GGML_ASSERT(src0->type == GGML_TYPE_F32);
+  GGML_ASSERT(src1->type == GGML_TYPE_F32);
+  GGML_ASSERT(src2->type == GGML_TYPE_I32);
+
+  GGML_ASSERT(nb00 == sizeof(float));
+  GGML_ASSERT(nb10 == sizeof(float));
+  GGML_ASSERT(nb20 == sizeof(int32_t));
+
+  const float* src0_d = (const float*)src0->data;
+  const float* src1_d = (const float*)src1->data;
+  const int32_t* src2_d = (const int32_t*)src2->data;
+  float* dst_d = (float*)dst->data;
+
+  int threads = std::min((int)ne00, 768);  // cols
+  ctx.stream()->parallel_for(
+      sycl::nd_range<3>(
+          sycl::range<3>(1, ne02, ne01) * sycl::range<3>(1, 1, threads),
+          sycl::range<3>(1, 1, threads)),
+      [=](sycl::nd_item<3> item_ct1) {
+        add_id_kernel(
+            src0_d,
+            src1_d,
+            src2_d,
+            dst_d,
+            ne0,
+            ne1,
+            nb01,
+            nb02,
+            nb11,
+            nb21,
+            item_ct1);
+      });
+}
--- a/ggml/src/ggml-sycl/add-id.hpp
+++ b/ggml/src/ggml-sycl/add-id.hpp
@ -0,0 +1,8 @@
+#ifndef GGML_SYCL_ADD_ID_HPP
+#define GGML_SYCL_ADD_ID_HPP
+
+#include "common.hpp"
+
+void ggml_sycl_add_id(ggml_backend_sycl_context & ctx, ggml_tensor * dst);
+
+#endif // GGML_SYCL_ADD_ID_HPP
--- a/ggml/src/ggml-sycl/common.hpp
+++ b/ggml/src/ggml-sycl/common.hpp
@ -642,5 +642,22 @@ static __dpct_inline__ sycl::uint2 fast_div_modulo(uint32_t n, const sycl::uint3
    return sycl::uint2(div_val, mod_val);
 }

+static __dpct_inline__ int ggml_sycl_dp4a(const int a, const int b, int c) {
+    return dpct::dp4a(a, b, c);
+}
+
+static __dpct_inline__ float ggml_sycl_e8m0_to_fp32(uint8_t x) {
+    uint32_t bits;
+    if (x == 0) {
+        bits = 0x00400000;
+    } else {
+        bits = (uint32_t) x << 23;
+    }
+
+    float result;
+    memcpy(&result, &bits, sizeof(float));
+    return result;
+}
+

 #endif // GGML_SYCL_COMMON_HPP
--- a/ggml/src/ggml-sycl/convert.cpp
+++ b/ggml/src/ggml-sycl/convert.cpp
@ -472,6 +472,16 @@ static void dequantize_row_iq4_nl_sycl(const void *vx, dst_t *y, const int64_t k
      }
 }

+template <typename dst_t>
+static void dequantize_row_mxfp4_sycl(const void * vx, dst_t * y, const int64_t k, dpct::queue_ptr stream) {
+    const int nb = (k + QK_K - 1) / QK_K;
+    stream->parallel_for(
+        sycl::nd_range<3>(sycl::range<3>(1, 1, nb) * sycl::range<3>(1, 1, 32), sycl::range<3>(1, 1, 32)),
+        [=](sycl::nd_item<3> item_ct1) {
+            dequantize_block_mxfp4(vx, y, item_ct1);
+        });
+}
+
 template <typename src_t, typename dst_t>
 static void convert_unary_nc(const void * __restrict__ vx, dst_t * __restrict__ y, const int64_t ne00, const int64_t ne01,
                          const int64_t ne02, const int64_t s01, const int64_t s02, const int64_t s03,
@ -518,6 +528,7 @@ static void convert_unary_sycl(const void * vx, dst_t * y, const int64_t k, dpct
    convert_unary_nc_sycl<src_t>(vx, y, k, 1, 1, 1, k, k, k, queue);
 }

+
 to_fp16_sycl_t ggml_get_to_fp16_sycl(ggml_type type, ggml_tensor * dst) {
    switch (type) {
        case GGML_TYPE_Q4_0:
@ -571,6 +582,8 @@ to_fp16_sycl_t ggml_get_to_fp16_sycl(ggml_type type, ggml_tensor * dst) {
            return dequantize_row_iq4_xs_sycl;
        case GGML_TYPE_IQ4_NL:
            return dequantize_row_iq4_nl_sycl;
+        case GGML_TYPE_MXFP4:
+            return dequantize_row_mxfp4_sycl;
        case GGML_TYPE_F32:
            return convert_unary_sycl<float>;
 #ifdef GGML_SYCL_HAS_BF16
@ -636,6 +649,8 @@ to_fp32_sycl_t ggml_get_to_fp32_sycl(ggml_type type, ggml_tensor *dst) {
            return dequantize_row_iq4_xs_sycl;
        case GGML_TYPE_IQ4_NL:
            return dequantize_row_iq4_nl_sycl;
+        case GGML_TYPE_MXFP4:
+            return dequantize_row_mxfp4_sycl;
        case GGML_TYPE_F16:
            return convert_unary_sycl<sycl::half>;
 #ifdef GGML_SYCL_HAS_BF16
--- a/ggml/src/ggml-sycl/dequantize.hpp
+++ b/ggml/src/ggml-sycl/dequantize.hpp
@ -819,5 +819,23 @@ dequantize_block_iq4_xs(const void *__restrict__ vx, dst_t *__restrict__ yy,
    }
 }

+template<typename dst_t>
+static void dequantize_block_mxfp4(const void * __restrict__ vx, dst_t * __restrict__ yy,
+                                   const sycl::nd_item<3> &item_ct1) {
+    // auto                item_ct1 = sycl::ext::oneapi::this_work_item::get_nd_item<3>();
+    const int64_t       i        = item_ct1.get_group(2);
+    const block_mxfp4 * x = (const block_mxfp4 *) vx + i*(QK_K/QK_MXFP4);
+
+    const int64_t    tid = item_ct1.get_local_id(2);
+    const int64_t il = tid/8; // 0...3
+    const int64_t ib = tid%8; // 0...7
+    dst_t * y = yy + i*QK_K + 32*ib + 4*il;
+    const uint8_t  * q4 = x[ib].qs + 4*il;
+    const float d = ggml_sycl_e8m0_to_fp32(x[ib].e);
+    for (int j = 0; j < 4; ++j) {
+        y[j+ 0] = d * kvalues_mxfp4[q4[j] & 0xf]*0.5f;
+        y[j+16] = d * kvalues_mxfp4[q4[j] >>  4]*0.5f;
+    }
+}

 #endif // GGML_SYCL_DEQUANTIZE_HPP
--- a/ggml/src/ggml-sycl/dpct/helper.hpp
+++ b/ggml/src/ggml-sycl/dpct/helper.hpp
@ -1860,10 +1860,31 @@ namespace dpct
                                           : id);
    }

+    template <typename T1, typename T2>
+    using dot_product_acc_t = std::conditional_t<
+        std::is_unsigned_v<T1> && std::is_unsigned_v<T2>,
+        uint32_t,
+        int32_t>;
+
+    template <typename T>
+    sycl::vec<T, 4> extract_and_sign_or_zero_extend4(T val) {
+      return sycl::vec<T, 1>(val)
+          .template as<sycl::vec<
+              std::conditional_t<std::is_signed_v<T>, int8_t, uint8_t>,
+              4>>()
+          .template convert<T>();
+    }
+
    template <typename T1, typename T2, typename T3>
-    inline auto dp4a(T1 a, T2 b, T3 c)
-    {
-        return syclcompat::dp4a(a, b, c);
+    inline auto dp4a(T1 a, T2 b, T3 c) {
+      dot_product_acc_t<T1, T2> res = c;
+      auto va = extract_and_sign_or_zero_extend4(a);
+      auto vb = extract_and_sign_or_zero_extend4(b);
+      res += va[0] * vb[0];
+      res += va[1] * vb[1];
+      res += va[2] * vb[2];
+      res += va[3] * vb[3];
+      return res;
    }

    struct sub_sat
@ -2972,6 +2993,38 @@ namespace dpct
    atomic_fetch_add<T1, addressSpace>(addr, operand, memoryOrder);
    }

+    inline unsigned int byte_level_permute(
+        unsigned int a, unsigned int b, unsigned int s) {
+      unsigned int ret;
+      ret = ((((std::uint64_t)b << 32 | a) >> (s & 0x7) * 8) & 0xff) |
+            (((((std::uint64_t)b << 32 | a) >> ((s >> 4) & 0x7) * 8) & 0xff)
+             << 8) |
+            (((((std::uint64_t)b << 32 | a) >> ((s >> 8) & 0x7) * 8) & 0xff)
+             << 16) |
+            (((((std::uint64_t)b << 32 | a) >> ((s >> 12) & 0x7) * 8) & 0xff)
+             << 24);
+      return ret;
+    }
+
+    inline uint32_t byte_level_permute_custom(
+        uint32_t low32, uint32_t high32, uint32_t sel, int mode = 0) {
+      constexpr uint16_t lookup[6][4] = {
+          {0x3210, 0x4321, 0x5432, 0x6543},  // Forward 4-byte extract
+          {0x5670, 0x6701, 0x7012, 0x0123},  // Backward 4-byte extract
+          {0x0000, 0x1111, 0x2222, 0x3333},  // Replicate 8-bit values
+          {0x3210, 0x3211, 0x3222, 0x3333},  // Edge clamp left
+          {0x0000, 0x1110, 0x2210, 0x3210},  // Edge clamp right
+          {0x1010, 0x3232, 0x1010, 0x3232}   // Replicate 16-bit values
+      };
+
+      if (mode >= 1 && mode <= 6) {
+        return byte_level_permute(low32, high32, lookup[mode - 1][sel & 0x3]);
+      } else if (!mode) {
+        return byte_level_permute(low32, high32, sel);
+      }
+      return 0;
+    }
+
 } // COPY from DPCT head files

 #endif // GGML_SYCL_DPCT_HELPER_HPP
--- a/ggml/src/ggml-sycl/element_wise.cpp
+++ b/ggml/src/ggml-sycl/element_wise.cpp
@ -911,6 +911,98 @@ static inline void ggml_sycl_op_swiglu(ggml_backend_sycl_context & ctx, ggml_ten
        });
 }

+__dpct_inline__ float ggml_sycl_op_swiglu_oai_single(float x, float g, float alpha = 1.702f, float limit = 7.0f) {
+    x = sycl::fmin(x, limit);
+    g = sycl::fmax(sycl::fmin(g, limit), -limit);
+
+    float out_glu = x / (1.0f + sycl::native::exp(-x * alpha));
+    out_glu = out_glu * (1.0f + g);
+    return out_glu;
+}
+
+
+template <typename T>
+static void swiglu_oai_kernel(const T * x, const T * g, T * dst, const int64_t k,
+                              const int64_t n, const int64_t o0, const int64_t o1,
+                              float alpha, float limit, sycl::nd_item<3> item_ct1) {
+    const int64_t i = int64_t(item_ct1.get_local_range(2)) * item_ct1.get_group(2) + item_ct1.get_local_id(2);
+
+    if (i >= k) {
+        return;
+    }
+
+    const int64_t j0 = (i / n) * o0 + (i % n);
+    const int64_t j1 = o0 == o1 ? j0 : (i / n) * o1 + (i % n);
+
+    float xi = x[j0];
+    float gi = g[j1];
+
+    dst[i] = ggml_sycl_op_swiglu_oai_single(xi, gi, alpha, limit);
+}
+
+template <typename T>
+static void swiglu_oai_sycl(const T *       x,
+                            const T *       g,
+                            T *             dst,
+                            const int64_t   k,
+                            const int64_t   n,
+                            const int64_t   o0,
+                            const int64_t   o1,
+                            const float     alpha,
+                            const float     limit,
+                            dpct::queue_ptr stream) {
+    const int64_t num_blocks = (k + SYCL_GLU_BLOCK_SIZE - 1) / SYCL_GLU_BLOCK_SIZE;
+    stream->parallel_for(sycl::nd_range<3>(sycl::range<3>(1, 1, num_blocks) * sycl::range<3>(1, 1, SYCL_GLU_BLOCK_SIZE),
+                                           sycl::range<3>(1, 1, SYCL_GLU_BLOCK_SIZE)),
+                         [=](sycl::nd_item<3> item_ct1) {
+                             swiglu_oai_kernel(x, g, dst, k, n, o0, o1, alpha, limit, item_ct1);
+                         });
+}
+
+void ggml_sycl_op_swiglu_oai(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
+    const ggml_tensor * src0 = dst->src[0];
+    const ggml_tensor * src1 = dst->src[1];
+    void * src0_d = src0->data;
+    void * src1_d = src1 ? src1->data : src0->data;
+    const int64_t src0_o = src0->nb[1];
+    const int64_t src1_o = src1 ? src1->nb[1] : src0->nb[1];
+    void * dst_d = dst->data;
+    const int64_t nc = src1 ? src0->ne[0] : src0->ne[0] / 2;
+    dpct::queue_ptr     stream = ctx.stream();
+
+    GGML_ASSERT(ggml_is_contiguous_1(src0));
+    GGML_ASSERT(src0->nb[0] == ggml_element_size(src0));
+    GGML_ASSERT(ggml_is_contiguous(dst));
+
+    GGML_ASSERT(src0->type == GGML_TYPE_F32);
+    GGML_ASSERT( dst->type == GGML_TYPE_F32);
+    GGML_ASSERT(src0->type == dst->type);
+    GGML_ASSERT(dst->ne[0] == nc);
+    GGML_ASSERT(ggml_nrows(dst) == ggml_nrows(src0));
+
+    if (src1) {
+        GGML_ASSERT(ggml_is_contiguous_1(src1));
+        GGML_ASSERT(src1->nb[0] == ggml_element_size(src1));
+        GGML_ASSERT(src1->ne[0] == nc);
+        GGML_ASSERT(src0->type == src1->type);
+    }
+
+    //const int32_t swapped = ((const int32_t *) dst->op_params)[1];
+    const int32_t swapped = ggml_get_op_params_i32(dst, 1);
+    const float alpha = ggml_get_op_params_f32(dst, 2);
+    const float limit = ggml_get_op_params_f32(dst, 3);
+
+    float * src0_p = (float *) src0_d;
+    float * src1_p = (float *) src1_d;
+
+    if (!src1) {
+        src0_p += swapped ? nc : 0;
+        src1_p += swapped ? 0 : nc;
+    }
+
+    swiglu_oai_sycl(src0_p, src1_p, (float *)dst_d, ggml_nelements(dst), nc, src0_o / sizeof(float), src1_o / sizeof(float), alpha, limit, stream);
+}
+
 static inline void ggml_sycl_op_geglu_erf(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
    ggml_sycl_detail::dispatch_ggml_sycl_op_fused_glu(ctx, dst,
        [](const auto* x_ptr, const auto* g_ptr, auto* dst_ptr, uint64_t k, uint64_t n, uint64_t o0, uint64_t o1, queue_ptr main_stream) {
@ -1070,6 +1162,11 @@ void ggml_sycl_swiglu(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
    ggml_sycl_op_swiglu(ctx, dst);
 }

+void ggml_sycl_swiglu_oai(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
+    scope_op_debug_print scope_dbg_print(__func__, dst, /*num_src=*/1);
+    ggml_sycl_op_swiglu_oai(ctx, dst);
+}
+
 void ggml_sycl_geglu_erf(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
    scope_op_debug_print scope_dbg_print(__func__, dst, /*num_src=*/1);
    ggml_sycl_op_geglu_erf(ctx, dst);
--- a/ggml/src/ggml-sycl/element_wise.hpp
+++ b/ggml/src/ggml-sycl/element_wise.hpp
@ -5,6 +5,8 @@
 #include "ggml.h"
 #include <limits> // For std::numeric_limits

+#define SYCL_GLU_BLOCK_SIZE 256
+
 template <typename T>
 T neg_infinity() {
    return -std::numeric_limits<T>::infinity();
@ -41,6 +43,8 @@ void ggml_sycl_silu(ggml_backend_sycl_context & ctx, ggml_tensor * dst);

 void ggml_sycl_gelu_quick(ggml_backend_sycl_context & ctx, ggml_tensor * dst);

+void ggml_sycl_swiglu_oai(ggml_backend_sycl_context & ctx, ggml_tensor * dst);
+
 void ggml_sycl_gelu_erf(ggml_backend_sycl_context & ctx, ggml_tensor * dst);

 void ggml_sycl_tanh(ggml_backend_sycl_context & ctx, ggml_tensor * dst);
--- a/ggml/src/ggml-sycl/ggml-sycl.cpp
+++ b/ggml/src/ggml-sycl/ggml-sycl.cpp
@ -39,6 +39,7 @@
 #include "ggml-impl.h"
 #include "ggml-backend-impl.h"

+#include "ggml-sycl/add-id.hpp"
 #include "ggml-sycl/backend.hpp"
 #include "ggml-sycl/common.hpp"
 #include "ggml-sycl/element_wise.hpp"
@ -3313,6 +3314,7 @@ static void ggml_sycl_mul_mat(ggml_backend_sycl_context & ctx, const ggml_tensor
    bool use_mul_mat_q =  ggml_sycl_supports_mmq(src0->type)
        && src1->type == GGML_TYPE_F32 && dst->type == GGML_TYPE_F32;

+
    // mmvq and mmq need the __dp4a instruction which is available for gen12+
    // Workaround in https://github.com/ggerganov/llama.cpp/commit/95f84d5ce8b449a9b16009434aca800df504a02e
    use_mul_mat_q = use_mul_mat_q && (src0->type != GGML_TYPE_IQ2_XXS);
@ -3320,7 +3322,6 @@ static void ggml_sycl_mul_mat(ggml_backend_sycl_context & ctx, const ggml_tensor
    use_mul_mat_q = use_mul_mat_q && (src1->ne[1] <= MMQ_MAX_BATCH_SIZE);
 #endif // SYCL_USE_XMX

-
    // mmvq path is faster in the CUDA backend.
    if (!g_ggml_sycl_prioritize_dmmv && (ctx.stream()->get_backend() == sycl::backend::ext_oneapi_cuda
        // Dispatch becomes obscure with the reorder, MMVQ when the reorder optimization
@ -3711,6 +3712,9 @@ static bool ggml_sycl_compute_forward(ggml_backend_sycl_context & ctx, struct gg
        case GGML_OP_ADD1: // TODO: more efficient implementation
            ggml_sycl_add(ctx, dst);
            break;
+        case GGML_OP_ADD_ID:
+            ggml_sycl_add_id(ctx, dst);
+            break;
        case GGML_OP_SUB:
            ggml_sycl_sub(ctx, dst);
            break;
@ -3803,6 +3807,9 @@ static bool ggml_sycl_compute_forward(ggml_backend_sycl_context & ctx, struct gg
                case GGML_GLU_OP_SWIGLU:
                    ggml_sycl_swiglu(ctx, dst);
                    break;
+                case GGML_GLU_OP_SWIGLU_OAI:
+                    ggml_sycl_swiglu_oai(ctx, dst);
+                    break;
                case GGML_GLU_OP_GEGLU_ERF:
                    ggml_sycl_geglu_erf(ctx, dst);
                    break;
@ -4397,6 +4404,7 @@ static bool ggml_backend_sycl_device_supports_op(ggml_backend_dev_t dev, const g
                case GGML_GLU_OP_REGLU:
                case GGML_GLU_OP_GEGLU:
                case GGML_GLU_OP_SWIGLU:
+                case GGML_GLU_OP_SWIGLU_OAI:
                case GGML_GLU_OP_GEGLU_ERF:
                case GGML_GLU_OP_GEGLU_QUICK:
                    return ggml_is_contiguous_1(op->src[0]);
@ -4424,15 +4432,18 @@ static bool ggml_backend_sycl_device_supports_op(ggml_backend_dev_t dev, const g
                    }
                }
                ggml_type src0_type = op->src[0]->type;
-                if (src0_type == GGML_TYPE_BF16 || src0_type == GGML_TYPE_MXFP4) {
-                    // TODO: support MXFP4
+                if (src0_type == GGML_TYPE_BF16 ) {
+                    // TODO: support GGML_TYPE_BF16
                    // FIXME: keep a list of supported types to avoid breaking the backend when a new type is added
                    return false;
                }
+
                // TODO: The configuration below needs more work to be supported with oneDNN
-                if (ggml_is_permuted(a) && !ggml_is_contiguous(a) && a->ne[2] > 1 && a->ne[3] > 1) {
-                    return false;
+                if (ggml_is_permuted(a) && !ggml_is_contiguous(a) &&
+                    a->ne[2] > 1 && a->ne[3] > 1 && src0_type == GGML_TYPE_F16) {
+                  return false;
                }
+
                // TODO: This specific configuration can fail with oneDNN and needs more debugging
                if (!ggml_is_permuted(a) && ggml_is_permuted(b) && b->ne[2] > 1 && b->ne[3] > 1 &&
                    a->ne[0] > 128 && a->ne[2] == 1 && src0_type == GGML_TYPE_F16) {
@ -4553,9 +4564,9 @@ static bool ggml_backend_sycl_device_supports_op(ggml_backend_dev_t dev, const g
        case GGML_OP_VIEW:
        case GGML_OP_PERMUTE:
        case GGML_OP_TRANSPOSE:
-            return true;
        case GGML_OP_ADD:
        case GGML_OP_ADD1:
+        case GGML_OP_ADD_ID:
        case GGML_OP_SUB:
        case GGML_OP_COUNT_EQUAL:
        case GGML_OP_MUL:
--- a/ggml/src/ggml-sycl/mmvq.cpp
+++ b/ggml/src/ggml-sycl/mmvq.cpp
@ -595,6 +595,25 @@ static void mul_mat_vec_q4_1_q8_1_sycl(const void *vx, const void *vy,
    }
 }

+static void mul_mat_vec_mxfp4_q8_1_sycl(const void * vx, const void * vy, float * dst, const int ncols, const int nrows,
+                                        dpct::queue_ptr stream) {
+    GGML_ASSERT(ncols % QK_MXFP4 == 0);
+    const int block_num_y = (nrows + GGML_SYCL_MMV_Y - 1) / GGML_SYCL_MMV_Y;
+    const sycl::range<3> block_nums(1, 1, block_num_y);
+    const sycl::range<3> block_dims(1, GGML_SYCL_MMV_Y, WARP_SIZE);
+
+    {
+        stream->submit([&](sycl::handler & cgh) {
+            cgh.parallel_for(sycl::nd_range<3>(block_nums * block_dims, block_dims),
+                             [=](sycl::nd_item<3> item_ct1) [[sycl::reqd_sub_group_size(WARP_SIZE)]] {
+                                 mul_mat_vec_q<QK_MXFP4, QI_MXFP4, block_mxfp4, VDR_MXFP4_Q8_1_MMVQ, vec_dot_mxfp4_q8_1>(
+                                     vx, vy, dst, ncols, nrows, item_ct1);
+                             });
+        });
+    }
+}
+
+
 static void mul_mat_vec_q5_0_q8_1_sycl(const void *vx, const void *vy,
                                       float *dst, const int ncols,
                                       const int nrows,
@ -1123,6 +1142,9 @@ void ggml_sycl_op_mul_mat_vec_q(ggml_backend_sycl_context & ctx, const ggml_tens
            case GGML_TYPE_IQ4_XS:
                mul_mat_vec_iq4_xs_q8_1_sycl(src0_dd_i, src1_ddq_i_bs, dst_dd_i_bs, ne00, row_diff, stream);
                break;
+            case GGML_TYPE_MXFP4:
+                mul_mat_vec_mxfp4_q8_1_sycl(src0_dd_i, src1_ddq_i_bs, dst_dd_i_bs, ne00, row_diff, stream);
+                break;
            default:
                GGML_ABORT("fatal error");
        }
--- a/ggml/src/ggml-sycl/pad.cpp
+++ b/ggml/src/ggml-sycl/pad.cpp
@ -14,10 +14,10 @@
 #include "pad.hpp"

 static void pad_f32(const float * src, float * dst,
-                               const int lp0, const int rp0, const int lp1, const int rp1,
-                               const int lp2, const int rp2, const int lp3, const int rp3,
-                               const int ne0, const int ne1, const int ne2, const int ne3) {
-    auto item_ct1 = sycl::ext::oneapi::this_work_item::get_nd_item<3>();
+                    const int lp0, const int rp0, const int lp1, const int rp1,
+                    const int lp2, const int rp2, const int lp3, const int rp3,
+                    const int ne0, const int ne1, const int ne2, const int ne3,
+                    sycl::nd_item<3> item_ct1) {
    int i0 = item_ct1.get_local_id(2) +
             item_ct1.get_group(2) * item_ct1.get_local_range(2);
    int i1 = item_ct1.get_group(1);
@ -63,7 +63,7 @@ static void pad_f32_sycl(const float *src, float *dst, const int lp0,
                          sycl::range<3>(1, 1, SYCL_PAD_BLOCK_SIZE)),
        [=](sycl::nd_item<3> item_ct1) {
            pad_f32(src, dst, lp0, rp0, lp1, rp1, lp2, rp2, lp3, rp3, ne0, ne1,
-                    ne2, ne3);
+                    ne2, ne3, item_ct1);
        });
 }

--- a/ggml/src/ggml-sycl/ssm_conv.cpp
+++ b/ggml/src/ggml-sycl/ssm_conv.cpp
@ -88,7 +88,7 @@ void ggml_sycl_ssm_conv(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
    GGML_ASSERT(src0->nb[0] == sizeof(float));
    GGML_ASSERT(src1->nb[0] == sizeof(float));

-    GGML_ASSERT(src0->nb[1] == src0->ne[0] * static_cast<int>(sizeof(float)));
+    GGML_ASSERT(src0->nb[1] == src0->ne[0] * sizeof(float));

    const int src_stride_inner = ncs;
    const int src_stride_seq   = ncs * d_inner;
--- a/ggml/src/ggml-sycl/vecdotq.hpp
+++ b/ggml/src/ggml-sycl/vecdotq.hpp
@ -20,6 +20,18 @@
 typedef float (*vec_dot_q_sycl_t)(const void * __restrict__ vbq, const block_q8_1 * __restrict__ bq8_1,
                                  const int & iqs);

+static __dpct_inline__ int get_int_b1(const void * x, const int & i32) {
+    const uint8_t * x8 = (const uint8_t *) x;
+
+    int x32  = x8[4*i32 + 0] <<  0;
+    x32     |= x8[4*i32 + 1] <<  8;
+    x32     |= x8[4*i32 + 2] << 16;
+    x32     |= x8[4*i32 + 3] << 24;
+
+    return x32;
+}
+
+
 static __dpct_inline__ int get_int_from_int8(const int8_t* x8, const int& i32) {
  const uint16_t* x16 =
      (const uint16_t*)(x8 + sizeof(int) * i32); // assume at least 2 byte
@ -75,6 +87,28 @@ static __dpct_inline__ void get_int_from_table_16(const uint32_t &q4,
    val2 = v1 | (v2 << 16);
 }

+static __dpct_inline__ sycl::int2 get_int_from_table_16(
+    const int& q4, const int8_t* table) {
+  const uint32_t* table32 = (const uint32_t*)table;
+  uint32_t tmp[2];
+  const uint32_t low_high_selection_indices =
+      (0x32103210 | ((q4 & 0x88888888) >> 1));
+#pragma unroll
+  for (uint32_t i = 0; i < 2; ++i) {
+    const uint32_t shift = 16 * i;
+
+    const uint32_t low =
+        dpct::byte_level_permute(table32[0], table32[1], q4 >> shift);
+    const uint32_t high =
+        dpct::byte_level_permute(table32[2], table32[3], q4 >> shift);
+    tmp[i] = dpct::byte_level_permute(
+        low, high, low_high_selection_indices >> shift);
+  }
+  return sycl::int2(
+      dpct::byte_level_permute(tmp[0], tmp[1], 0x6420),
+      dpct::byte_level_permute(tmp[0], tmp[1], 0x7531));
+}
+
 #define VDR_Q2_K_Q8_1_MMVQ 1

 // contiguous v/x values
@ -685,6 +719,30 @@ vec_dot_q4_1_q8_1(const void *__restrict__ vbq,
    return vec_dot_q4_1_q8_1_impl<VDR_Q4_1_Q8_1_MMVQ>(v, u, bq4_1->dm, bq8_1->ds);
 }

+#define VDR_MXFP4_Q8_1_MMVQ 2
+#define VDR_MXFP4_Q8_1_MMQ  4
+
+static __dpct_inline__ float vec_dot_mxfp4_q8_1(const void * __restrict__ vbq,
+                                                const block_q8_1 * __restrict__ bq8_1,
+                                                const int & iqs) {
+    const block_mxfp4 * bq4 = (const block_mxfp4 *) vbq;
+
+    const int * q8 = (const int *) bq8_1->qs + iqs;
+
+    int sumi = 0;
+#pragma unroll
+    for (int l = 0; l < VDR_MXFP4_Q8_1_MMVQ; ++l) {
+        const int aux_q4 = get_int_b1(bq4->qs, iqs + l);
+        const sycl::int2 v      = get_int_from_table_16(aux_q4, kvalues_mxfp4);
+        sumi = ggml_sycl_dp4a(v.x(), q8[l + 0], sumi);
+        sumi = ggml_sycl_dp4a(v.y(), q8[l + 4], sumi);
+    }
+
+    const float d = ggml_sycl_e8m0_to_fp32(bq4->e) * 0.5f * (bq8_1->ds)[0];
+    return d * sumi;
+}
+
+
 static __dpct_inline__ float
 vec_dot_q5_0_q8_1(const void *__restrict__ vbq,
                  const block_q8_1 *__restrict__ bq8_1, const int &iqs) {
--- a/ggml/src/ggml.c
+++ b/ggml/src/ggml.c
@ -7566,6 +7566,11 @@ size_t ggml_quantize_chunk(

 ////////////////////////////////////////////////////////////////////////////////

+void ggml_log_get(ggml_log_callback * log_callback, void ** user_data) {
+    *log_callback = g_logger_state.log_callback;
+    *user_data    = g_logger_state.log_callback_user_data;
+}
+
 void ggml_log_set(ggml_log_callback log_callback, void * user_data) {
    g_logger_state.log_callback = log_callback ? log_callback : ggml_log_callback_default;
    g_logger_state.log_callback_user_data = user_data;
--- a/gguf-py/gguf/constants.py
+++ b/gguf-py/gguf/constants.py
@ -3320,6 +3320,7 @@ class VisionProjectorType:
    ULTRAVOX = "ultravox"
    INTERNVL = "internvl"
    QWEN2A = "qwen2a" # audio
+    GLMA = "glma" # audio
    QWEN25O = "qwen2.5o" # omni
    VOXTRAL = "voxtral"
    LFM2 = "lfm2"
--- a/grammars/README.md
+++ b/grammars/README.md
@ -1,6 +1,6 @@
 # GBNF Guide

-GBNF (GGML BNF) is a format for defining [formal grammars](https://en.wikipedia.org/wiki/Formal_grammar) to constrain model outputs in `llama.cpp`. For example, you can use it to force the model to generate valid JSON, or speak only in emojis. GBNF grammars are supported in various ways in `tools/main` and `tools/server`.
+GBNF (GGML BNF) is a format for defining [formal grammars](https://en.wikipedia.org/wiki/Formal_grammar) to constrain model outputs in `llama.cpp`. For example, you can use it to force the model to generate valid JSON, or speak only in emojis. GBNF grammars are supported in various ways in `tools/cli`, `tools/completion` and `tools/server`.

 ## Background

@ -135,7 +135,7 @@ While semantically correct, the syntax `x? x? x?.... x?` (with N repetitions) ma
 You can use GBNF grammars:

 - In [llama-server](../tools/server)'s completion endpoints, passed as the `grammar` body field
- In [llama-cli](../tools/main), passed as the `--grammar` & `--grammar-file` flags
+- In [llama-cli](../tools/cli) and [llama-completion](../tools/completion), passed as the `--grammar` & `--grammar-file` flags
 - With [test-gbnf-validator](../tests/test-gbnf-validator.cpp), to test them against strings.

 ## JSON Schemas → GBNF
@ -145,7 +145,7 @@ You can use GBNF grammars:
 - In [llama-server](../tools/server):
    - For any completion endpoints, passed as the `json_schema` body field
    - For the `/chat/completions` endpoint, passed inside the `response_format` body field (e.g. `{"type", "json_object", "schema": {"items": {}}}` or `{ type: "json_schema", json_schema: {"schema": ...} }`)
- In [llama-cli](../tools/main), passed as the `--json` / `-j` flag
+- In [llama-cli](../tools/cli) and [llama-completion](../tools/completion), passed as the `--json` / `-j` flag
 - To convert to a grammar ahead of time:
    - in CLI, with [examples/json_schema_to_grammar.py](../examples/json_schema_to_grammar.py)
    - in JavaScript with [json-schema-to-grammar.mjs](../tools/server/public_legacy/json-schema-to-grammar.mjs) (this is used by the [server](../tools/server)'s Web UI)
--- a/include/llama.h
+++ b/include/llama.h
@ -313,6 +313,7 @@ extern "C" {
        bool check_tensors;   // validate model tensor data
        bool use_extra_bufts; // use extra buffer types (used for weight repacking)
        bool no_host;         // bypass host buffer allowing extra buffers to be used
+        bool no_alloc;        // only load metadata and simulate memory allocations
    };

    // NOTE: changing the default values of parameters marked as [EXPERIMENTAL] may cause crashes or incorrect results in certain configurations
@ -466,10 +467,24 @@ extern "C" {
    // Frees all allocated memory
    LLAMA_API void llama_free(struct llama_context * ctx);

+    // fits mparams and cparams to free device memory (assumes system memory is unlimited)
+    // returns true if the parameters could be successfully modified to fit device memory
+    // this function is NOT thread safe because it modifies the global llama logger state
+    LLAMA_API bool llama_params_fit(
+                                   const char   * path_model,
+                    struct llama_model_params   * mparams,
+                    struct llama_context_params * cparams,
+                                          float * tensor_split,          // writable buffer for tensor split, needs at least llama_max_devices elements
+        struct llama_model_tensor_buft_override * tensor_buft_overrides, // writable buffer for overrides, needs at least llama_max_tensor_buft_overrides elements
+                                         size_t   margin,                // margin of memory to leave per device in bytes
+                                       uint32_t   n_ctx_min,             // minimum context size to set when trying to reduce memory use
+                            enum ggml_log_level   log_level);            // minimum log level to print during fitting, lower levels go to debug log
+
    LLAMA_API int64_t llama_time_us(void);

    LLAMA_API size_t llama_max_devices(void);
    LLAMA_API size_t llama_max_parallel_sequences(void);
+    LLAMA_API size_t llama_max_tensor_buft_overrides(void);

    LLAMA_API bool llama_supports_mmap       (void);
    LLAMA_API bool llama_supports_mlock      (void);
@ -1377,7 +1392,9 @@ extern "C" {

    // Set callback for all future logging events.
    // If this is not called, or NULL is supplied, everything is output on stderr.
-    LLAMA_API void llama_log_set(ggml_log_callback log_callback, void * user_data);
+    // The logger state is global so these functions are NOT thread safe.
+    LLAMA_API void llama_log_get(ggml_log_callback * log_callback, void ** user_data);
+    LLAMA_API void llama_log_set(ggml_log_callback   log_callback, void *  user_data);

    //
    // Performance utils
--- a/src/llama-context.cpp
+++ b/src/llama-context.cpp
@ -258,6 +258,7 @@ llama_context::llama_context(

        backend_buft.clear();
        backend_ptrs.clear();
+        backend_buf_exp_size.clear();

        for (auto & backend : backends) {
            auto * buft = ggml_backend_get_default_buffer_type(backend.get());
@ -274,6 +275,7 @@ llama_context::llama_context(

            backend_buft.push_back(buft);
            backend_ptrs.push_back(backend.get());
+            backend_buf_exp_size.push_back(0);
        }

        LLAMA_LOG_DEBUG("%s: backend_ptrs.size() = %zu\n", __func__, backend_ptrs.size());
@ -389,7 +391,8 @@ llama_context::llama_context(

        // reserve pp (prompt processing) graph first so that buffers are only allocated once
        {
-            auto * gf = graph_reserve(n_tokens, n_seqs, n_tokens, mctx.get());
+            auto * gf = graph_reserve(n_tokens, n_seqs, n_tokens, mctx.get(),
+                model.hparams.no_alloc, model.hparams.no_alloc ? backend_buf_exp_size.data() : nullptr);
            if (!gf) {
                if (pipeline_parallel) {
                    LLAMA_LOG_WARN("%s: compute buffer allocation failed, retrying without pipeline parallelism\n", __func__);
@ -407,7 +410,7 @@ llama_context::llama_context(

        // reserve with tg (token generation) graph to get the number of splits and nodes
        {
-            auto * gf = graph_reserve(n_seqs, n_seqs, n_seqs, mctx.get());
+            auto * gf = graph_reserve(n_seqs, n_seqs, n_seqs, mctx.get(), model.hparams.no_alloc);
            if (!gf) {
                throw std::runtime_error("failed to allocate compute tg buffers");
            }
@ -422,7 +425,7 @@ llama_context::llama_context(
            //
            // auto * gf = graph_reserve(n_tokens, 1, n_tokens, mctx.get());
            //
-            auto * gf = graph_reserve(n_tokens, n_seqs, n_tokens, mctx.get());
+            auto * gf = graph_reserve(n_tokens, n_seqs, n_tokens, mctx.get(), model.hparams.no_alloc);
            if (!gf) {
                throw std::runtime_error("failed to allocate compute pp buffers");
            }
@ -431,11 +434,13 @@ llama_context::llama_context(
        for (size_t i = 0; i < backend_ptrs.size(); ++i) {
            ggml_backend_t             backend = backend_ptrs[i];
            ggml_backend_buffer_type_t buft    = backend_buft[i];
-            size_t size = ggml_backend_sched_get_buffer_size(sched.get(), backend);
-            if (size > 1) {
+            if (!model.hparams.no_alloc) {
+                backend_buf_exp_size[i] = ggml_backend_sched_get_buffer_size(sched.get(), backend);
+            }
+            if (backend_buf_exp_size[i] > 1) {
                LLAMA_LOG_INFO("%s: %10s compute buffer size = %8.2f MiB\n", __func__,
                        ggml_backend_buft_name(buft),
-                        size / 1024.0 / 1024.0);
+                        backend_buf_exp_size[i] / 1024.0 / 1024.0);
            }
        }

@ -454,6 +459,23 @@ llama_context::llama_context(
 }

 llama_context::~llama_context() {
+    // FIXME this currently results in a use-after-free bug if the model is freed before the context
+    // if (!model.hparams.no_alloc) {
+    //     for (size_t i = 0; i < backend_ptrs.size(); ++i) {
+    //         ggml_backend_t             backend = backend_ptrs[i];
+    //         ggml_backend_buffer_type_t buft    = backend_buft[i];
+
+    //         const size_t size_exp = backend_buf_exp_size[i];
+    //         const size_t size_act = ggml_backend_sched_get_buffer_size(sched.get(), backend);
+    //         if (size_exp == size_act) {
+    //             LLAMA_LOG_DEBUG("%s: %10s compute buffer size is %8.4f MiB, matches expectation of %8.4f MiB\n",
+    //                 __func__, ggml_backend_buft_name(buft), size_act / (1024.0*1024.0), size_exp / (1024.0*1024.0));
+    //         } else {
+    //             LLAMA_LOG_WARN("%s: %10s compute buffer size of %8.4f MiB, does not match expectation of %8.4f MiB\n",
+    //                 __func__, ggml_backend_buft_name(buft), size_act / (1024.0*1024.0), size_exp / (1024.0*1024.0));
+    //         }
+    //     }
+    // }
    ggml_opt_free(opt_ctx);
 }

@ -1428,7 +1450,8 @@ llm_graph_result * llama_context::get_gf_res_reserve() const {
    return static_cast<llm_graph_result *>(gf_res_reserve.get());
 }

-ggml_cgraph * llama_context::graph_reserve(uint32_t n_tokens, uint32_t n_seqs, uint32_t n_outputs, const llama_memory_context_i * mctx, bool split_only) {
+ggml_cgraph * llama_context::graph_reserve(
+        uint32_t n_tokens, uint32_t n_seqs, uint32_t n_outputs, const llama_memory_context_i * mctx, bool split_only, size_t * sizes) {
    LLAMA_LOG_DEBUG("%s: reserving a graph for ubatch with n_tokens = %4u, n_seqs = %2u, n_outputs = %4u\n", __func__, n_tokens, n_seqs, n_outputs);
    GGML_ASSERT(n_outputs >= 1);

@ -1465,8 +1488,13 @@ ggml_cgraph * llama_context::graph_reserve(uint32_t n_tokens, uint32_t n_seqs, u

    // initialize scheduler with the specified graph
    if (split_only) {
-        ggml_backend_sched_split_graph(sched.get(), gf);
+        if (sizes) {
+            ggml_backend_sched_reserve_size(sched.get(), gf, sizes);
+        } else {
+            ggml_backend_sched_split_graph(sched.get(), gf);
+        }
    } else if (!ggml_backend_sched_reserve(sched.get(), gf)) {
+        GGML_ASSERT(!sizes);
        LLAMA_LOG_ERROR("%s: failed to allocate compute buffers\n", __func__);
        return nullptr;
    }
@ -2088,15 +2116,26 @@ void llama_context::perf_reset() {

 std::map<ggml_backend_buffer_type_t, llama_memory_breakdown_data> llama_context::memory_breakdown() const {
    std::map<ggml_backend_buffer_type_t, llama_memory_breakdown_data> ret;
-    for (const auto & buft_size : model.memory_breakdown()) {
-        ret[buft_size.first].model += buft_size.second;
+    for (const auto & [buft, size] : model.memory_breakdown()) {
+        ret[buft].model += size;
    }
-    for (const auto & buft_size : memory->memory_breakdown()) {
-        ret[buft_size.first].context += buft_size.second;
+    if (memory) {
+        for (const auto & [buft, size] : memory->memory_breakdown()) {
+            ret[buft].context += size;
+        }
    }
-    for (const auto & backend_ptr : backends) {
-        ggml_backend_t backend = backend_ptr.get();
-        ret[ggml_backend_sched_get_buffer_type(sched.get(), backend)].compute += ggml_backend_sched_get_buffer_size(sched.get(), backend);
+    if (model.hparams.no_alloc) {
+        for (size_t i = 0; i < backends.size(); ++i) {
+            ggml_backend_t             backend = backends[i].get();
+            ggml_backend_buffer_type_t buft    = ggml_backend_sched_get_buffer_type(sched.get(), backend);
+            ret[buft].compute += backend_buf_exp_size[i];
+        }
+    } else {
+        for (const auto & backend_ptr : backends) {
+            ggml_backend_t             backend = backend_ptr.get();
+            ggml_backend_buffer_type_t buft    = ggml_backend_sched_get_buffer_type(sched.get(), backend);
+            ret[buft].compute += ggml_backend_sched_get_buffer_size(sched.get(), backend);
+        }
    }
    return ret;
 }
--- a/src/llama-context.h
+++ b/src/llama-context.h
@ -26,6 +26,10 @@ struct llama_memory_breakdown_data {
    size_t model   = 0; // memory allocated for the model
    size_t context = 0; // memory allocated for the context
    size_t compute = 0; // memory allocated for temporary compute buffers
+
+    size_t total() const {
+        return model + context + compute;
+    }
 };

 struct llama_context {
@ -206,7 +210,8 @@ public:
    ggml_status graph_compute(ggml_cgraph * gf, bool batched);

    // reserve a graph with a dummy ubatch of the specified size
-    ggml_cgraph * graph_reserve(uint32_t n_tokens, uint32_t n_seqs, uint32_t n_outputs, const llama_memory_context_i * mctx, bool split_only = false);
+    ggml_cgraph * graph_reserve(
+        uint32_t n_tokens, uint32_t n_seqs, uint32_t n_outputs, const llama_memory_context_i * mctx, bool split_only = false, size_t * sizes = nullptr);

 private:
    llm_graph_params graph_params(
@ -281,9 +286,10 @@ private:

    std::vector<std::pair<ggml_backend_t, ggml_backend_set_n_threads_t>> set_n_threads_fns;

-    // buffer types used for the compute buffer of each backend
+    // pointers and buffer types used for the compute buffer of each backend
    std::vector<ggml_backend_t>             backend_ptrs;
    std::vector<ggml_backend_buffer_type_t> backend_buft;
+    std::vector<size_t>                     backend_buf_exp_size; // expected buffer sizes

    llm_graph_result_ptr gf_res_prev;
    llm_graph_result_ptr gf_res_reserve;
--- a/src/llama-hparams.h
+++ b/src/llama-hparams.h
@ -34,6 +34,7 @@ struct llama_hparams_convnext {

 struct llama_hparams {
    bool vocab_only;
+    bool no_alloc;
    bool rope_finetuned;
    bool use_par_res;
    bool swin_norm;
--- a/src/llama-impl.cpp
+++ b/src/llama-impl.cpp
@ -25,6 +25,10 @@ time_meas::~time_meas() {
    }
 }

+void llama_log_get(ggml_log_callback * log_callback, void ** user_data) {
+    ggml_log_get(log_callback, user_data);
+}
+
 void llama_log_set(ggml_log_callback log_callback, void * user_data) {
    ggml_log_set(log_callback, user_data);
    g_logger_state.log_callback = log_callback ? log_callback : llama_log_callback_default;
--- a/src/llama-kv-cache.cpp
+++ b/src/llama-kv-cache.cpp
@ -175,7 +175,15 @@ llama_kv_cache::llama_kv_cache(

    // allocate tensors and initialize the buffers to avoid NaNs in the padding
    for (auto & [buft, ctx] : ctx_map) {
-        ggml_backend_buffer_t buf = ggml_backend_alloc_ctx_tensors_from_buft(ctx.get(), buft);
+        ggml_backend_buffer_t buf;
+        if (model.hparams.no_alloc) {
+            buf = ggml_backend_buft_alloc_buffer(buft, /*size =*/ 0); // dummy buffer
+            for (ggml_tensor * t = ggml_get_first_tensor(ctx.get()); t != nullptr; t = ggml_get_next_tensor(ctx.get(), t)) {
+                t->buffer = buf; // set dummy buffer for KV cache so that the backend scheduler won't try to allocate it
+            }
+        } else {
+            buf = ggml_backend_alloc_ctx_tensors_from_buft(ctx.get(), buft); // real buffer
+        }
        if (!buf) {
            throw std::runtime_error("failed to allocate buffer for kv cache");
        }
@ -482,9 +490,18 @@ llama_pos llama_kv_cache::seq_pos_max(llama_seq_id seq_id) const {

 std::map<ggml_backend_buffer_type_t, size_t> llama_kv_cache::memory_breakdown() const {
    std::map<ggml_backend_buffer_type_t, size_t> ret;
-    for (const auto & [_, buf] : ctxs_bufs) {
-        ret[ggml_backend_buffer_get_type(buf.get())] += ggml_backend_buffer_get_size(buf.get());
+    for (const auto & [ctx, buf] : ctxs_bufs) {
+        ggml_backend_buffer_type_t buft = ggml_backend_buffer_get_type(buf.get());
+
+        if (hparams.no_alloc) {
+            GGML_ASSERT(ggml_backend_buffer_get_base(buf.get()) == nullptr);
+            ret[buft] += ggml_backend_alloc_ctx_tensors_from_buft_size(ctx.get(), buft);
+        } else {
+            // GGML_ASSERT(ggml_backend_buffer_get_base(buf.get()) != nullptr); // multi_buffer does not have a defined base
+            ret[buft] += ggml_backend_buffer_get_size(buf.get());
+        }
    }
+
    return ret;
 }

--- a/src/llama-model-loader.cpp
+++ b/src/llama-model-loader.cpp
@ -473,6 +473,7 @@ llama_model_loader::llama_model_loader(
        std::vector<std::string> & splits,
        bool use_mmap,
        bool check_tensors,
+        bool no_alloc,
        const llama_model_kv_override * param_overrides_p,
        const llama_model_tensor_buft_override * param_tensor_buft_overrides_p) {
    int trace = 0;
@ -716,6 +717,7 @@ llama_model_loader::llama_model_loader(

    this->use_mmap = use_mmap;
    this->check_tensors = check_tensors;
+    this->no_alloc = no_alloc;
 }

 std::string llama_model_loader::get_arch_name() const {
--- a/src/llama-model-loader.h
+++ b/src/llama-model-loader.h
@ -71,6 +71,7 @@ struct llama_model_loader {

    bool use_mmap = false;
    bool check_tensors;
+    bool no_alloc;

    llama_files files;
    llama_ftype ftype;
@ -97,6 +98,7 @@ struct llama_model_loader {
        std::vector<std::string> & splits, // optional, only need if the split does not follow naming scheme
        bool use_mmap,
        bool check_tensors,
+        bool no_alloc,
        const llama_model_kv_override * param_overrides_p,
        const llama_model_tensor_buft_override * param_tensor_buft_overrides_p);

--- a/src/llama-model.cpp
+++ b/src/llama-model.cpp
@ -6606,9 +6606,11 @@ bool llama_model::load_tensors(llama_model_loader & ml) {

        std::vector<ggml_backend_buffer_ptr> bufs;
        if (ml.use_mmap && use_mmap_buffer && buffer_from_host_ptr_supported && is_default_buft) {
+            GGML_ASSERT(!ml.no_alloc);
            for (uint32_t idx = 0; idx < ml.files.size(); idx++) {
                // only the mmap region containing the tensors in the model is mapped to the backend buffer
-                // this is important for metal with apple silicon: if the entire model could be mapped to a metal buffer, then we could just use metal for all layers
+                // this is important for metal with apple silicon: if the entire model could be mapped to a metal buffer,
+                //     then we could just use metal for all layers
                // this allows using partial offloading when the model size exceeds the metal buffer size, but not the RAM size
                void * addr = nullptr;
                size_t first, last; // NOLINT
@ -6624,9 +6626,16 @@ bool llama_model::load_tensors(llama_model_loader & ml) {
                bufs.emplace_back(buf);
                buf_map.emplace(idx, buf);
            }
-        }
-        else {
-            ggml_backend_buffer_t buf = ggml_backend_alloc_ctx_tensors_from_buft(ctx, buft);
+        } else {
+            ggml_backend_buffer_t buf;
+            if (ml.no_alloc) {
+                buf = ggml_backend_buft_alloc_buffer(buft, /*size =*/ 0); // dummy buffer
+                for (ggml_tensor * t = ggml_get_first_tensor(ctx); t != nullptr; t = ggml_get_next_tensor(ctx, t)) {
+                    t->buffer = buf; // set dummy buffer for weights so that the backend scheduler won't try to allocate them
+                }
+            } else {
+                buf = ggml_backend_alloc_ctx_tensors_from_buft(ctx, buft); // real buffer
+            }
            if (buf == nullptr) {
                throw std::runtime_error(format("unable to allocate %s buffer", ggml_backend_buft_name(buft)));
            }
@ -6681,6 +6690,10 @@ bool llama_model::load_tensors(llama_model_loader & ml) {
        }
    }

+    if (ml.no_alloc) {
+        return true;
+    }
+
    // load tensor data
    for (auto & [ctx, buf_map] : ctx_buf_maps) {
        if (!ml.load_all_data(ctx, buf_map, use_mlock ? &pimpl->mlock_mmaps : NULL, params.progress_callback, params.progress_callback_user_data)) {
@ -6723,9 +6736,18 @@ size_t llama_model::n_devices() const {

 std::map<ggml_backend_buffer_type_t, size_t> llama_model::memory_breakdown() const {
    std::map<ggml_backend_buffer_type_t, size_t> ret;
-    for (const auto & [_, bufs] : pimpl->ctxs_bufs) {
-        for (const auto & buf : bufs) {
-            ret[ggml_backend_buffer_get_type(buf.get())] += ggml_backend_buffer_get_size(buf.get());
+    for (const auto & [ctx, bufs] : pimpl->ctxs_bufs) {
+        if (hparams.no_alloc) {
+            GGML_ASSERT(bufs.size() == 1);
+            ggml_backend_buffer_t buf = bufs[0].get();
+            GGML_ASSERT(ggml_backend_buffer_get_base(buf) == nullptr);
+            ggml_backend_buffer_type_t buft = ggml_backend_buffer_get_type(buf);
+            ret[buft] += ggml_backend_alloc_ctx_tensors_from_buft_size(ctx.get(), buft);
+        } else {
+            for (const auto & buf : bufs) {
+                // GGML_ASSERT(ggml_backend_buffer_get_base(buf.get()) != nullptr); // multi_buffer does not have a defined base
+                ret[ggml_backend_buffer_get_type(buf.get())] += ggml_backend_buffer_get_size(buf.get());
+            }
        }
    }
    return ret;
@ -6770,6 +6792,7 @@ void llama_model::print_info() const {
    // hparams
    LLAMA_LOG_INFO("%s: arch             = %s\n",     __func__, arch_name().c_str());
    LLAMA_LOG_INFO("%s: vocab_only       = %d\n",     __func__, hparams.vocab_only);
+    LLAMA_LOG_INFO("%s: no_alloc         = %d\n",     __func__, hparams.no_alloc);

    if (!hparams.vocab_only) {
        LLAMA_LOG_INFO("%s: n_ctx_train      = %u\n",     __func__, hparams.n_ctx_train);
@ -7618,6 +7641,7 @@ llama_model_params llama_model_default_params() {
        /*.check_tensors               =*/ false,
        /*.use_extra_bufts             =*/ true,
        /*.no_host                     =*/ false,
+        /*.no_alloc                    =*/ false,
    };

    return result;
--- a/src/llama-quant.cpp
+++ b/src/llama-quant.cpp
@ -596,7 +596,7 @@ static void llama_model_quantize_impl(const std::string & fname_inp, const std::
    }

    std::vector<std::string> splits = {};
-    llama_model_loader ml(fname_inp, splits, use_mmap, /*check_tensors*/ true, kv_overrides, nullptr);
+    llama_model_loader ml(fname_inp, splits, use_mmap, /*check_tensors*/ true, /*no_alloc*/ false, kv_overrides, nullptr);
    ml.init_mappings(false); // no prefetching

    llama_model model(llama_model_default_params());
--- a/src/llama.cpp
+++ b/src/llama.cpp
@ -1,6 +1,9 @@
+#include "llama.h"
+
 #include "llama-impl.h"

 #include "llama-chat.h"
+#include "llama-context.h"
 #include "llama-mmap.h"
 #include "llama-vocab.h"
 #include "llama-model-loader.h"
@ -11,11 +14,14 @@
 #include "ggml-backend.h"

 #include <algorithm>
+#include <cassert>
+#include <cinttypes>
 #include <cstddef>
 #include <cstdint>
 #include <cstdio>
 #include <cstring>
 #include <ctime>
+#include <stdexcept>

 #if defined(_MSC_VER)
 #pragma warning(disable: 4244 4267) // possible loss of data
@ -37,6 +43,643 @@ const char * llama_flash_attn_type_name(enum llama_flash_attn_type flash_attn_ty
    GGML_ABORT("fatal error");
 }

+struct llama_device_memory_data {
+    int64_t total;
+    int64_t free;
+    llama_memory_breakdown_data mb;
+};
+
+static std::vector<llama_device_memory_data> llama_get_device_memory_data(
+        const char * path_model, const llama_model_params * mparams, const llama_context_params * cparams,
+        std::vector<ggml_backend_dev_t> & devs, uint32_t & hp_ngl, uint32_t & hp_n_ctx_train, uint32_t & hp_n_expert,
+        const ggml_log_level log_level) {
+    struct user_data_t {
+        struct {
+            ggml_log_callback callback;
+            void * user_data;
+        } original_logger;
+        ggml_log_level min_level; // prints below this log level go to debug log
+    };
+    user_data_t ud;
+    llama_log_get(&ud.original_logger.callback, &ud.original_logger.user_data);
+    ud.min_level = log_level;
+
+    llama_log_set([](ggml_log_level level, const char * text, void * user_data) {
+        const user_data_t * ud = (const user_data_t *) user_data;
+        const ggml_log_level level_eff = level >= ud->min_level ? level : GGML_LOG_LEVEL_DEBUG;
+        ud->original_logger.callback(level_eff, text, ud->original_logger.user_data);
+    }, &ud);
+
+    llama_model_params mparams_copy = *mparams;
+    mparams_copy.no_alloc = true;
+    mparams_copy.use_mmap = false;
+
+    llama_model * model = llama_model_load_from_file(path_model, mparams_copy);
+    if (model == nullptr) {
+        llama_log_set(ud.original_logger.callback, ud.original_logger.user_data);
+        throw std::runtime_error("failed to load model");
+    }
+
+    llama_context * ctx = llama_init_from_model(model, *cparams);
+    if (ctx == nullptr) {
+        llama_model_free(model);
+        llama_log_set(ud.original_logger.callback, ud.original_logger.user_data);
+        throw std::runtime_error("failed to create llama_context from model");
+    }
+
+    std::vector<llama_device_memory_data> ret(model->devices.size());
+
+    std::map<ggml_backend_buffer_type_t, llama_memory_breakdown_data> memory_breakdown = ctx->memory_breakdown();
+
+    for (const auto & [buft, mb] : memory_breakdown) {
+        if (ggml_backend_buft_is_host(buft)) {
+            continue;
+        }
+
+        ggml_backend_dev_t dev = ggml_backend_buft_get_device(buft);
+        if (!dev) {
+            continue;
+        }
+        for (size_t i = 0; i < ret.size(); i++) {
+            if (model->devices[i] == dev) {
+                ret[i].mb.model   += mb.model;
+                ret[i].mb.context += mb.context;
+                ret[i].mb.compute += mb.compute;
+                break;
+            }
+        }
+    }
+    for (size_t i = 0; i < ret.size(); i++) {
+        size_t free, total;
+        ggml_backend_dev_memory(model->devices[i], &free, &total);
+        ret[i].free  = free;
+        ret[i].total = total;
+    }
+
+    devs           = model->devices;
+    hp_ngl         = model->hparams.n_layer;
+    hp_n_ctx_train = model->hparams.n_ctx_train;
+    hp_n_expert    = model->hparams.n_expert;
+
+    llama_memory_breakdown_print(ctx); // goes to debug log
+
+    llama_free(ctx);
+    llama_model_free(model);
+    llama_log_set(ud.original_logger.callback, ud.original_logger.user_data);
+    return ret;
+}
+
+// enum to identify part of a layer for distributing its tensors:
+enum layer_fraction_t {
+    LAYER_FRACTION_NONE = 0, // nothing
+    LAYER_FRACTION_ATTN = 1, // attention
+    LAYER_FRACTION_UP   = 2, // attention + up
+    LAYER_FRACTION_GATE = 3, // attention + up + gate
+    LAYER_FRACTION_MOE  = 4, // everything but sparse MoE weights
+};
+// this enum is only used in llama_params_fit_impl but needs to be defined outside of it to fix a Windows compilation issue
+
+static void llama_params_fit_impl(
+        const char * path_model, struct llama_model_params * mparams, struct llama_context_params * cparams,
+        float * tensor_split, struct llama_model_tensor_buft_override * tensor_buft_overrides,
+        size_t margin_s, uint32_t n_ctx_min, enum ggml_log_level log_level) {
+    constexpr int64_t MiB = 1024*1024;
+    const int64_t margin = margin_s; // this function uses int64_t rather than size_t for memory sizes to more conveniently handle deficits
+    typedef std::vector<llama_device_memory_data> dmds_t;
+    const llama_model_params default_mparams = llama_model_default_params();
+
+    std::vector<ggml_backend_dev_t> devs;
+    uint32_t hp_ngl = 0; // hparams.n_gpu_layers
+    uint32_t hp_nct = 0; // hparams.n_ctx_train
+    uint32_t hp_nex = 0; // hparams.n_expert
+
+    // step 1: get data for default parameters and check whether any changes are necessary in the first place
+
+    LLAMA_LOG_DEBUG("%s: getting device memory data for initial parameters:\n", __func__);
+    const dmds_t dmds_full = llama_get_device_memory_data(path_model, mparams, cparams, devs, hp_ngl, hp_nct, hp_nex, log_level);
+    const size_t nd = devs.size(); // number of devices
+    if (nd == 0) {
+        LLAMA_LOG_INFO("%s: no devices with dedicated memory found\n", __func__);
+        return;
+    }
+
+    std::vector<std::string> dev_names;
+    {
+        dev_names.reserve(nd);
+        size_t max_length = 0;
+        for (ggml_backend_dev_t dev : devs) {
+            std::string name = ggml_backend_dev_name(dev);
+            name += " (";
+            name += ggml_backend_dev_description(dev);
+            name += ")";
+            dev_names.push_back(name);
+            max_length = std::max(max_length, name.length());
+        }
+        for (std::string & dn : dev_names) {
+            dn.insert(dn.end(), max_length - dn.length(), ' ');
+        }
+    }
+
+    int64_t sum_total          = 0;
+    int64_t sum_projected_free = 0;
+    int64_t min_projected_free = INT64_MAX;
+    int64_t sum_projected_used = 0;
+    int64_t sum_projected_ctx  = 0;
+
+    if (nd > 1) {
+        LLAMA_LOG_INFO("%s: projected memory use with initial parameters [MiB]:\n", __func__);
+    }
+    for (size_t id = 0; id < nd; id++) {
+        const llama_device_memory_data & dmd = dmds_full[id];
+
+        const int64_t projected_used = dmd.mb.total();
+        const int64_t projected_free = dmd.free - projected_used;
+
+        sum_total          += dmd.total;
+        sum_projected_used += projected_used;
+        sum_projected_free += projected_free;
+        min_projected_free  = std::min(min_projected_free, projected_free);
+        sum_projected_ctx  += dmd.mb.context;
+
+        if (nd > 1) {
+            LLAMA_LOG_INFO("%s:   - %s: %6" PRId64 " total, %6" PRId64 " used, %6" PRId64 " %s\n",
+                __func__, dev_names[id].c_str(), dmd.total/MiB, projected_used/MiB, std::abs(projected_free)/MiB,
+                projected_free >= 0 ? "surplus" : "deficit");
+        }
+    }
+    assert(sum_total >= 0 && sum_projected_used >= 0 && sum_projected_ctx >= 0);
+    assert(sum_projected_used >= sum_projected_ctx);
+    LLAMA_LOG_INFO("%s: projected to use %" PRId64 " MiB of device memory vs. %" PRId64 " MiB of free device memory\n",
+        __func__, sum_projected_used/MiB, sum_total/MiB);
+    if (min_projected_free >= margin) {
+        if (nd == 1) {
+            LLAMA_LOG_INFO("%s: will leave %" PRId64 " >= %" PRId64 " MiB of free device memory, no changes needed\n",
+                __func__, min_projected_free/MiB, margin/MiB);
+            return;
+        }
+        LLAMA_LOG_INFO("%s: will leave at least %" PRId64 " >= %" PRId64 " MiB of free memory on all devices, no changes needed\n",
+            __func__, min_projected_free/MiB, margin/MiB);
+        return;
+    }
+
+    // step 2: try reducing memory use by reducing the context size
+
+    {
+        int64_t global_surplus = sum_projected_free - int64_t(nd)*margin;
+        if (global_surplus < 0) {
+            LLAMA_LOG_INFO(nd == 1 ?
+                "%s: cannot fulfill margin of %" PRId64 " MiB, need to reduce device memory by %" PRId64 " MiB\n" :
+                "%s: cannot fulfill margin of %" PRId64 " MiB on all devices, need to use %" PRId64 " MiB less in total\n",
+                __func__, margin/MiB, -global_surplus/MiB);
+            if (cparams->n_ctx == 0) {
+                if (hp_nct > n_ctx_min) {
+                    const int64_t bytes_per_ctx = sum_projected_ctx / hp_nct;
+                    const uint32_t ctx_reduction = std::min(
+                        uint32_t((-global_surplus + bytes_per_ctx - 1) / bytes_per_ctx), hp_nct - n_ctx_min);
+                    cparams->n_ctx = hp_nct - ctx_reduction;
+                    const int64_t memory_reduction = ctx_reduction * bytes_per_ctx;
+                    global_surplus += memory_reduction;
+                    LLAMA_LOG_INFO("%s: context size reduced from %" PRIu32 " to %" PRIu32 " -> need %" PRId64 " MiB less memory in total\n",
+                        __func__, hp_nct, cparams->n_ctx, memory_reduction/MiB);
+                } else {
+                    LLAMA_LOG_INFO("%s: default model context size is %" PRIu32 " which is <= the min. context size of %" PRIu32 " -> no change\n",
+                        __func__, hp_nct, n_ctx_min);
+                }
+            } else {
+                LLAMA_LOG_INFO("%s: context size set by user to %" PRIu32 " -> no change\n", __func__, cparams->n_ctx);
+            }
+        }
+        if (global_surplus >= 0) {
+            LLAMA_LOG_INFO("%s: entire model can be fit across devices by reducing context\n", __func__);
+            return;
+        }
+    }
+
+    if (mparams->n_gpu_layers != default_mparams.n_gpu_layers) {
+        throw std::runtime_error("n_gpu_layers already set by user to " + std::to_string(mparams->n_gpu_layers) + ", abort");
+    }
+    if (nd > 1) {
+        if (!tensor_split) {
+            throw std::runtime_error("did not provide a buffer to write the tensor_split to, abort");
+        }
+        if (mparams->tensor_split) {
+            for (size_t id = 0; id < nd; id++) {
+                if (mparams->tensor_split[id] != 0.0f) {
+                    throw std::runtime_error("model_params::tensor_split already set by user, abort");
+                }
+            }
+        }
+        if (mparams->split_mode == LLAMA_SPLIT_MODE_ROW) {
+            throw std::runtime_error("changing weight allocation for LLAMA_SPLIT_MODE_ROW not implemented, abort");
+        }
+        if (hp_ngl < 2*nd) {
+            throw std::runtime_error("model has only " + std::to_string(hp_ngl) + " layers but need at least "
+                + std::to_string(2*nd) + " to fit memory for " + std::to_string(nd) + " devices, abort");
+        }
+    }
+    if (!tensor_buft_overrides) {
+        throw std::runtime_error("did not provide buffer to set tensor_buft_overrides, abort");
+    }
+    if (mparams->tensor_buft_overrides && (mparams->tensor_buft_overrides->pattern || mparams->tensor_buft_overrides->buft)) {
+        throw std::runtime_error("model_params::tensor_buft_overrides already set by user, abort");
+    }
+
+    // step 3: iteratively fill the back to front with "dense" layers
+    //   - for a dense model simply fill full layers, giving each device a contiguous slice of the model
+    //   - for a MoE model, same as dense model but with all MoE tensors in system memory
+
+    // utility function that returns a static C string matching the tensors for a specific layer index and layer fraction:
+    auto get_overflow_pattern = [&](const size_t il, const layer_fraction_t lf) -> const char * {
+        constexpr size_t n_strings = 1000;
+        if (il >= n_strings) {
+            throw std::runtime_error("at most " + std::to_string(n_strings) + " model layers are supported");
+        }
+        switch (lf) {
+            case LAYER_FRACTION_ATTN: {
+                static std::array<std::string, n_strings> patterns;
+                if (patterns[il].empty()) {
+                    patterns[il] = "blk\\." + std::to_string(il) + "\\.ffn_(up|gate|down).*";
+                }
+                return patterns[il].c_str();
+            }
+            case LAYER_FRACTION_UP: {
+                static std::array<std::string, n_strings> patterns;
+                if (patterns[il].empty()) {
+                    patterns[il] = "blk\\." + std::to_string(il) + "\\.ffn_(gate|down).*";
+                }
+                return patterns[il].c_str();
+            }
+            case LAYER_FRACTION_GATE: {
+                static std::array<std::string, n_strings> patterns;
+                if (patterns[il].empty()) {
+                    patterns[il] = "blk\\." + std::to_string(il) + "\\.ffn_down.*";
+                }
+                return patterns[il].c_str();
+            }
+            case LAYER_FRACTION_MOE: {
+                static std::array<std::string, n_strings> patterns;
+                if (patterns[il].empty()) {
+                    patterns[il] = "blk\\." + std::to_string(il) + "\\.ffn_(up|down|gate)_(ch|)exps";
+                }
+                return patterns[il].c_str();
+            }
+            default:
+                GGML_ABORT("fatal error");
+        }
+    };
+
+    struct ngl_t {
+        uint32_t n_layer = 0; // number of total layers
+        uint32_t n_part  = 0; // number of partial layers, <= n_layer
+
+        // for the first partial layer varying parts can overflow, all further layers use LAYER_FRACTION_MOE:
+        layer_fraction_t overflow_type = LAYER_FRACTION_MOE;
+    };
+
+    const size_t ntbo = llama_max_tensor_buft_overrides();
+
+    // utility function to set n_gpu_layers and tensor_split
+    auto set_ngl_tensor_split_tbo = [&](
+            const std::vector<ngl_t> & ngl_per_device,
+            const std::vector<ggml_backend_buffer_type_t> & overflow_bufts,
+            llama_model_params & mparams,
+            const bool add_nonrepeating) {
+        mparams.n_gpu_layers = 0;
+        for (size_t id = 0; id < nd; id++) {
+            mparams.n_gpu_layers += ngl_per_device[id].n_layer;
+            if (nd > 1) {
+                tensor_split[id] = ngl_per_device[id].n_layer;
+            }
+        }
+        assert(uint32_t(mparams.n_gpu_layers) <= hp_ngl);
+        uint32_t il0 = hp_ngl - mparams.n_gpu_layers; // start index for tensor buft overrides
+
+        if (add_nonrepeating) {
+            mparams.n_gpu_layers += 1;
+            tensor_split[nd - 1] += 1;
+        }
+        mparams.tensor_split = tensor_split;
+
+        size_t itbo = 0;
+        for (size_t id = 0; id < nd; id++) {
+            il0 += ngl_per_device[id].n_layer - ngl_per_device[id].n_part;
+            for (uint32_t il = il0; il < il0 + ngl_per_device[id].n_part; il++) {
+                if (itbo + 1 >= ntbo) {
+                    tensor_buft_overrides[itbo].pattern = nullptr;
+                    tensor_buft_overrides[itbo].buft    = nullptr;
+                    itbo++;
+                    mparams.tensor_buft_overrides = tensor_buft_overrides;
+                    throw std::runtime_error("llama_params_fit_n_tensor_buft_overrides() == "
+                        + std::to_string(ntbo) + " is insufficient for model\n");
+                }
+                tensor_buft_overrides[itbo].pattern = get_overflow_pattern(il, il == il0 ? ngl_per_device[id].overflow_type : LAYER_FRACTION_MOE);
+                tensor_buft_overrides[itbo].buft = overflow_bufts[id];
+                itbo++;
+            }
+            il0 += ngl_per_device[id].n_part;
+        }
+        tensor_buft_overrides[itbo].pattern = nullptr;
+        tensor_buft_overrides[itbo].buft    = nullptr;
+        itbo++;
+        mparams.tensor_buft_overrides = tensor_buft_overrides;
+    };
+
+    // utility function that returns the memory use per device for given numbers of layers per device
+    auto get_memory_for_layers = [&](
+            const char * func_name,
+            const std::vector<ngl_t> & ngl_per_device,
+            const std::vector<ggml_backend_buffer_type_t> & overflow_bufts,
+            const bool add_nonrepeating) -> std::vector<int64_t> {
+        llama_model_params mparams_copy = *mparams;
+        set_ngl_tensor_split_tbo(ngl_per_device, overflow_bufts, mparams_copy, add_nonrepeating);
+
+        const dmds_t dmd_nl = llama_get_device_memory_data(
+            path_model, &mparams_copy, cparams, devs, hp_ngl, hp_nct, hp_nex, log_level);
+
+        LLAMA_LOG_DEBUG("%s: memory for test allocation by device:\n", func_name);
+        for (size_t id = 0; id < nd; id++) {
+            const ngl_t & n = ngl_per_device[id];
+            LLAMA_LOG_DEBUG(
+                "%s: id=%zu, n_layer=%2" PRIu32 ", n_part=%2" PRIu32 ", overflow_type=%d, mem=%6" PRId64 " MiB\n",
+                func_name, id, n.n_layer, n.n_part, int(n.overflow_type), dmd_nl[id].mb.total()/MiB);
+        }
+
+        std::vector<int64_t> ret;
+        ret.reserve(nd);
+        for (const llama_device_memory_data & dmd : dmd_nl) {
+            ret.push_back(dmd.mb.total());
+        }
+        return ret;
+    };
+
+    int64_t global_surplus_cpu_moe = 0;
+    if (hp_nex > 0) {
+        const static std::string pattern_moe_all = "blk\\.\\d+\\.ffn_(up|down|gate)_(ch|)exps"; // matches all MoE tensors
+        ggml_backend_buffer_type_t cpu_buft = ggml_backend_cpu_buffer_type();
+        tensor_buft_overrides[0] = {pattern_moe_all.c_str(), cpu_buft};
+        tensor_buft_overrides[1] = {nullptr, nullptr};
+        mparams->tensor_buft_overrides = tensor_buft_overrides;
+
+        LLAMA_LOG_DEBUG("%s: getting device memory data with all MoE tensors moved to system memory:\n", __func__);
+        const dmds_t dmds_cpu_moe = llama_get_device_memory_data(
+            path_model, mparams, cparams, devs, hp_ngl, hp_nct, hp_nex, log_level);
+
+        for (const llama_device_memory_data & dmd : dmds_cpu_moe) {
+            global_surplus_cpu_moe += dmd.free;
+            global_surplus_cpu_moe -= int64_t(dmd.mb.total()) + margin;
+        }
+
+        if (global_surplus_cpu_moe > 0) {
+            LLAMA_LOG_INFO("%s: with only dense weights in device memory there is a total surplus of %" PRId64 " MiB\n",
+                __func__, global_surplus_cpu_moe/MiB);
+        } else {
+            LLAMA_LOG_INFO("%s: with only dense weights in device memory there is still a total deficit of %" PRId64 " MiB\n",
+                __func__, -global_surplus_cpu_moe/MiB);
+        }
+
+        // reset
+        tensor_buft_overrides[0] = {nullptr, nullptr};
+        mparams->tensor_buft_overrides = tensor_buft_overrides;
+    }
+
+    std::vector<int64_t> targets; // maximum acceptable memory use per device
+    targets.reserve(nd);
+    for (size_t id = 0; id < nd; id++) {
+        targets.push_back(dmds_full[id].free - margin);
+        LLAMA_LOG_DEBUG("%s: id=%zu, target=%" PRId64 " MiB\n", __func__, id, targets[id]/MiB);
+    }
+
+    // whether for the optimal memory use we expect to load at least some MoE tensors:
+    const bool partial_moe = hp_nex > 0 && global_surplus_cpu_moe > 0;
+
+    std::vector<ggml_backend_buffer_type_t> overflow_bufts; // which bufts the partial layers of a device overflow to:
+    overflow_bufts.reserve(nd);
+    for (size_t id = 0; id < nd - 1; ++id) {
+        overflow_bufts.push_back(ggml_backend_dev_buffer_type(devs[id + 1]));
+    }
+    overflow_bufts.push_back(ggml_backend_cpu_buffer_type());
+
+    std::vector<ngl_t> ngl_per_device(nd);
+    std::vector<int64_t> mem = get_memory_for_layers(__func__, ngl_per_device, overflow_bufts, partial_moe);
+    if (hp_nex > 0) {
+        for (size_t id = 0; id < nd; id++) {
+            ngl_per_device[id].overflow_type = LAYER_FRACTION_MOE;
+        }
+    }
+
+    // optimize the number of layers per device using the method of false position:
+    //   - ngl_per_device has 0 layers for each device, lower bound
+    //   - try a "high" configuration where a device is given all unassigned layers
+    //   - interpolate the memory use / layer between low and high linearly to get a guess where it meets our target
+    //   - check memory use of our guess, replace either the low or high bound
+    //   - once we only have a difference of a single layer, stop and return the lower bound that just barely still fits
+    if (hp_nex == 0) {
+        LLAMA_LOG_INFO("%s: filling dense layers back-to-front:\n", __func__);
+    } else {
+        LLAMA_LOG_INFO("%s: filling dense-only layers back-to-front:\n", __func__);
+    }
+    uint32_t n_unassigned = hp_ngl;
+    for (int id = nd - 1; id >= 0; id--) {
+        std::vector<ngl_t> ngl_per_device_high = ngl_per_device;
+        ngl_per_device_high[id].n_layer = n_unassigned;
+        if (hp_nex > 0) {
+            ngl_per_device_high[id].n_part = ngl_per_device_high[id].n_layer;
+        }
+        if (ngl_per_device_high[id].n_layer > 0) {
+            std::vector<int64_t> mem_high = get_memory_for_layers(__func__, ngl_per_device_high, overflow_bufts, partial_moe);
+            if (mem_high[id] > targets[id]) {
+                uint32_t delta = ngl_per_device_high[id].n_layer - ngl_per_device[id].n_layer;
+                while (delta > 1) {
+                    uint32_t step_size = int64_t(delta) * (targets[id] - mem[id]) / (mem_high[id] - mem[id]);
+                    step_size = std::max(step_size, uint32_t(1));
+                    step_size = std::min(step_size, delta - 1);
+
+                    std::vector<ngl_t> ngl_per_device_test = ngl_per_device;
+                    ngl_per_device_test[id].n_layer += step_size;
+                    if (hp_nex) {
+                        ngl_per_device_test[id].n_part += step_size;
+                    }
+                    const std::vector<int64_t> mem_test = get_memory_for_layers(__func__, ngl_per_device_test, overflow_bufts, partial_moe);
+
+                    if (mem_test[id] <= targets[id]) {
+                        ngl_per_device  = ngl_per_device_test;
+                        mem             = mem_test;
+                        n_unassigned   -= ngl_per_device[id].n_layer;
+                        LLAMA_LOG_DEBUG("%s: set ngl_per_device[%d].n_layer=%" PRIu32 "\n", __func__, id, ngl_per_device[id].n_layer);
+                    } else {
+                        ngl_per_device_high = ngl_per_device_test;
+                        mem_high            = mem_test;
+                        LLAMA_LOG_DEBUG("%s: set ngl_per_device_high[%d].n_layer=%" PRIu32 "\n", __func__, id, ngl_per_device[id].n_layer);
+                    }
+                    delta = ngl_per_device_high[id].n_layer - ngl_per_device[id].n_layer;
+                }
+            } else {
+                ngl_per_device  = ngl_per_device_high;
+                n_unassigned   -= ngl_per_device[id].n_layer;
+                LLAMA_LOG_DEBUG("%s: set ngl_per_device[%d].n_layer=%" PRIu32 "\n", __func__, id, ngl_per_device[id].n_layer);
+            }
+        }
+
+        const int64_t projected_margin = dmds_full[id].free - mem[id];
+        LLAMA_LOG_INFO(
+            "%s:   - %s: %2" PRIu32 " layers, %6" PRId64 " MiB used, %6" PRId64 " MiB free\n",
+            __func__, dev_names[id].c_str(), ngl_per_device[id].n_layer, mem[id]/MiB, projected_margin/MiB);
+    }
+    if (hp_nex == 0 || global_surplus_cpu_moe <= 0) {
+        set_ngl_tensor_split_tbo(ngl_per_device, overflow_bufts, *mparams, partial_moe);
+        return;
+    }
+
+    // step 4: for a MoE model where all dense tensors fit,
+    //     convert the dense-only layers in the back to full layers in the front until all devices are full
+    // essentially the same procedure as for the dense-only layers except front-to-back
+    // also, try fitting at least part of one more layer to reduce waste for "small" GPUs with e.g. 24 GiB VRAM
+
+    size_t id_dense_start = nd;
+    for (int id = nd - 1; id >= 0; id--) {
+        if (ngl_per_device[id].n_layer > 0) {
+            id_dense_start = id;
+            continue;
+        }
+        break;
+    }
+    assert(id_dense_start < nd);
+
+    LLAMA_LOG_INFO("%s: converting dense-only layers to full layers and filling them front-to-back with overflow to next device/system memory:\n", __func__);
+    for (size_t id = 0; id <= id_dense_start; id++) {
+        std::vector<ngl_t> ngl_per_device_high = ngl_per_device;
+        for (size_t jd = id_dense_start; jd < nd; jd++) {
+            const uint32_t n_layer_move = ngl_per_device_high[jd].n_layer;
+            ngl_per_device_high[id].n_layer += n_layer_move;
+            ngl_per_device_high[jd].n_layer -= n_layer_move;
+            ngl_per_device_high[jd].n_part = 0;
+        }
+        size_t id_dense_start_high = nd - 1;
+        std::vector<int64_t> mem_high = get_memory_for_layers(__func__, ngl_per_device_high, overflow_bufts, partial_moe);
+
+        if (mem_high[id] > targets[id]) {
+            assert(ngl_per_device_high[id].n_layer >= ngl_per_device_high[id].n_part);
+            assert(ngl_per_device[id].n_layer >= ngl_per_device[id].n_part);
+            assert((ngl_per_device_high[id].n_layer - ngl_per_device_high[id].n_part)
+                   >= ngl_per_device[id].n_layer - ngl_per_device[id].n_part);
+            uint32_t delta = (ngl_per_device_high[id].n_layer - ngl_per_device_high[id].n_part)
+                - (ngl_per_device[id].n_layer - ngl_per_device[id].n_part);
+            while (delta > 1) {
+                uint32_t step_size = int64_t(delta) * (targets[id] - mem[id]) / (mem_high[id] - mem[id]);
+                step_size = std::max(step_size, uint32_t(1));
+                step_size = std::min(step_size, delta - 1);
+
+                std::vector<ngl_t> ngl_per_device_test = ngl_per_device;
+                size_t id_dense_start_test = id_dense_start;
+                uint32_t n_converted_test = 0;
+                for (;id_dense_start_test < nd; id_dense_start_test++) {
+                    const uint32_t n_convert_jd = std::min(step_size - n_converted_test, ngl_per_device_test[id_dense_start_test].n_part);
+                    ngl_per_device_test[id_dense_start_test].n_layer -= n_convert_jd;
+                    ngl_per_device_test[id_dense_start_test].n_part -= n_convert_jd;
+                    ngl_per_device_test[id].n_layer += n_convert_jd;
+                    n_converted_test += n_convert_jd;
+
+                    if (ngl_per_device_test[id_dense_start_test].n_layer > 0) {
+                        break;
+                    }
+                }
+                const std::vector<int64_t> mem_test = get_memory_for_layers(__func__, ngl_per_device_test, overflow_bufts, partial_moe);
+
+                if (mem_test[id] <= targets[id]) {
+                    ngl_per_device = ngl_per_device_test;
+                    mem            = mem_test;
+                    id_dense_start = id_dense_start_test;
+                    LLAMA_LOG_DEBUG("%s: set ngl_per_device[%zu].(n_layer, n_part)=(%" PRIu32 ", %" PRIu32 "), id_dense_start=%zu\n",
+                        __func__, id, ngl_per_device[id].n_layer, ngl_per_device[id].n_part, id_dense_start);
+                } else {
+                    ngl_per_device_high = ngl_per_device_test;
+                    mem_high            = mem_test;
+                    id_dense_start_high = id_dense_start_test;
+                    LLAMA_LOG_DEBUG("%s: set ngl_per_device_high[%zu].(n_layer, n_part)=(%" PRIu32 ", %" PRIu32 "), id_dense_start_high=%zu\n",
+                        __func__, id, ngl_per_device_high[id].n_layer, ngl_per_device_high[id].n_part, id_dense_start_high);
+                }
+                delta = (ngl_per_device_high[id].n_layer - ngl_per_device_high[id].n_part)
+                    - (ngl_per_device[id].n_layer - ngl_per_device[id].n_part);
+            }
+        } else {
+            ngl_per_device = ngl_per_device_high;
+            id_dense_start = id_dense_start_high;
+            LLAMA_LOG_DEBUG("%s: set ngl_per_device[%zu].(n_layer, n_part)=(%" PRIu32 ", %" PRIu32 "), id_dense_start=%zu\n",
+                __func__, id, ngl_per_device[id].n_layer, ngl_per_device[id].n_part, id_dense_start);
+        }
+
+        // try to fit at least part of one more layer
+        if (ngl_per_device[id_dense_start].n_layer > 0) {
+            std::vector<ngl_t> ngl_per_device_test = ngl_per_device;
+            size_t id_dense_start_test = id_dense_start;
+            ngl_per_device_test[id_dense_start_test].n_layer--;
+            ngl_per_device_test[id_dense_start_test].n_part--;
+            ngl_per_device_test[id].n_layer++;
+            ngl_per_device_test[id].n_part++;
+            if (ngl_per_device_test[id_dense_start_test].n_layer == 0) {
+                id_dense_start_test++;
+            }
+            ngl_per_device_test[id].overflow_type = LAYER_FRACTION_UP;
+            LLAMA_LOG_DEBUG("%s: trying to fit one extra layer with overflow_type=LAYER_FRACTION_UP\n", __func__);
+            std::vector<int64_t> mem_test = get_memory_for_layers(__func__, ngl_per_device_test, overflow_bufts, partial_moe);
+            if (mem_test[id] < targets[id]) {
+                ngl_per_device = ngl_per_device_test;
+                mem            = mem_test;
+                id_dense_start = id_dense_start_test;
+                LLAMA_LOG_DEBUG("%s: set ngl_per_device[%zu].(n_layer, n_part, overflow_type)=(%" PRIu32 ", %" PRIu32 ", UP), id_dense_start=%zu\n",
+                    __func__, id, ngl_per_device[id].n_layer, ngl_per_device[id].n_part, id_dense_start);
+
+                ngl_per_device_test[id].overflow_type = LAYER_FRACTION_GATE;
+                LLAMA_LOG_DEBUG("%s: trying to fit one extra layer with overflow_type=LAYER_FRACTION_GATE\n", __func__);
+                mem_test = get_memory_for_layers(__func__, ngl_per_device_test, overflow_bufts, partial_moe);
+                if (mem_test[id] < targets[id]) {
+                    ngl_per_device = ngl_per_device_test;
+                    mem            = mem_test;
+                    id_dense_start = id_dense_start_test;
+                    LLAMA_LOG_DEBUG("%s: set ngl_per_device[%zu].(n_layer, n_part, overflow_type)=(%" PRIu32 ", %" PRIu32 ", GATE), id_dense_start=%zu\n",
+                        __func__, id, ngl_per_device[id].n_layer, ngl_per_device[id].n_part, id_dense_start);
+                }
+            } else {
+                ngl_per_device_test[id].overflow_type = LAYER_FRACTION_ATTN;
+                LLAMA_LOG_DEBUG("%s: trying to fit one extra layer with overflow_type=LAYER_FRACTION_ATTN\n", __func__);
+                mem_test = get_memory_for_layers(__func__, ngl_per_device_test, overflow_bufts, partial_moe);
+                if (mem_test[id] < targets[id]) {
+                    ngl_per_device = ngl_per_device_test;
+                    mem            = mem_test;
+                    id_dense_start = id_dense_start_test;
+                    LLAMA_LOG_DEBUG("%s: set ngl_per_device[%zu].(n_layer, n_part, overflow_type)=(%" PRIu32 ", %" PRIu32 ", ATTN), id_dense_start=%zu\n",
+                        __func__, id, ngl_per_device[id].n_layer, ngl_per_device[id].n_part, id_dense_start);
+                }
+            }
+        }
+
+        const int64_t projected_margin = dmds_full[id].free - mem[id];
+        LLAMA_LOG_INFO(
+            "%s:   - %s: %2" PRIu32 " layers (%2" PRIu32 " overflowing), %6" PRId64 " MiB used, %6" PRId64 " MiB free\n",
+            __func__, dev_names[id].c_str(), ngl_per_device[id].n_layer, ngl_per_device[id].n_part, mem[id]/MiB, projected_margin/MiB);
+    }
+
+    set_ngl_tensor_split_tbo(ngl_per_device, overflow_bufts, *mparams, partial_moe);
+}
+
+bool llama_params_fit(
+        const char * path_model, struct llama_model_params * mparams, struct llama_context_params * cparams,
+        float * tensor_split, struct llama_model_tensor_buft_override * tensor_buft_overrides,
+        size_t margin_s, uint32_t n_ctx_min, enum ggml_log_level log_level) {
+    const int64_t t0_us = llama_time_us();
+    bool ok = true;
+    try {
+        llama_params_fit_impl(path_model, mparams, cparams, tensor_split, tensor_buft_overrides, margin_s, n_ctx_min, log_level);
+        LLAMA_LOG_INFO("%s: successfully fit params to free device memory\n", __func__);
+    } catch (const std::runtime_error & e) {
+        LLAMA_LOG_WARN("%s: failed to fit params to free device memory: %s\n", __func__, e.what());
+        ok = false;
+    }
+    const int64_t t1_us = llama_time_us();
+    LLAMA_LOG_INFO("%s: fitting params to free memory took %.2f seconds\n", __func__, (t1_us - t0_us) * 1e-6);
+    return ok;
+}
+
 struct llama_sampler_chain_params llama_sampler_chain_default_params() {
    struct llama_sampler_chain_params result = {
        /*.no_perf                     =*/ true,
@ -49,6 +692,10 @@ size_t llama_max_devices(void) {
    return 16;
 }

+size_t llama_max_tensor_buft_overrides() {
+    return 4096;
+}
+
 bool llama_supports_mmap(void) {
    return llama_mmap::SUPPORTED;
 }
@ -108,11 +755,12 @@ static int llama_model_load(const std::string & fname, std::vector<std::string>
    model.t_start_us = tm.t_start_us;

    try {
-        llama_model_loader ml(fname, splits, params.use_mmap, params.check_tensors, params.kv_overrides, params.tensor_buft_overrides);
+        llama_model_loader ml(fname, splits, params.use_mmap, params.check_tensors, params.no_alloc, params.kv_overrides, params.tensor_buft_overrides);

        ml.print_info();

        model.hparams.vocab_only = params.vocab_only;
+        model.hparams.no_alloc   = params.no_alloc;

        try {
            model.load_arch(ml);
--- a/tools/CMakeLists.txt
+++ b/tools/CMakeLists.txt
@ -37,4 +37,5 @@ else()
        add_subdirectory(cvector-generator)
        add_subdirectory(export-lora)
    endif()
+    add_subdirectory(fit-params)
 endif()
--- a/tools/cli/README.md
+++ b/tools/cli/README.md
@ -0,0 +1 @@
+TODO
--- a/tools/completion/README.md
+++ b/tools/completion/README.md
@ -1,4 +1,4 @@
-# llama.cpp/tools/main
+# llama.cpp/tools/completion

 This example program allows you to use various LLaMA language models easily and efficiently. It is specifically designed to work with the [llama.cpp](https://github.com/ggml-org/llama.cpp) project, which provides a plain C/C++ implementation with optional 4-bit quantization support for faster, lower memory inference, and is optimized for desktop CPUs. This program can be used to perform various inference tasks with LLaMA models, including generating text based on user-provided prompts and chat-like interactions with reverse prompts.

@ -27,64 +27,64 @@ Once downloaded, place your model in the models folder in llama.cpp.
 ##### Input prompt (One-and-done)

 ```bash
-./llama-cli -m models/gemma-1.1-7b-it.Q4_K_M.gguf -no-cnv --prompt "Once upon a time"
+./llama-completion -m models/gemma-1.1-7b-it.Q4_K_M.gguf -no-cnv --prompt "Once upon a time"
 ```
 ##### Conversation mode (Allow for continuous interaction with the model)

 ```bash
-./llama-cli -m models/gemma-1.1-7b-it.Q4_K_M.gguf --chat-template gemma
+./llama-completion -m models/gemma-1.1-7b-it.Q4_K_M.gguf --chat-template gemma
 ```

 ##### Conversation mode using built-in jinja chat template

 ```bash
-./llama-cli -m models/gemma-1.1-7b-it.Q4_K_M.gguf --jinja
+./llama-completion -m models/gemma-1.1-7b-it.Q4_K_M.gguf --jinja
 ```

 ##### One-and-done query using jinja with custom system prompt and a starting prompt

 ```bash
-./llama-cli -m models/gemma-1.1-7b-it.Q4_K_M.gguf --jinja --single-turn -sys "You are a helpful assistant" -p "Hello"
+./llama-completion -m models/gemma-1.1-7b-it.Q4_K_M.gguf --jinja --single-turn -sys "You are a helpful assistant" -p "Hello"
 ```

 ##### Infinite text from a starting prompt (you can use `Ctrl-C` to stop it):
 ```bash
-./llama-cli -m models/gemma-1.1-7b-it.Q4_K_M.gguf --ignore-eos -n -1
+./llama-completion -m models/gemma-1.1-7b-it.Q4_K_M.gguf --ignore-eos -n -1
 ```

 ### Windows:

 ##### Input prompt (One-and-done)
 ```powershell
-./llama-cli.exe -m models\gemma-1.1-7b-it.Q4_K_M.gguf -no-cnv --prompt "Once upon a time"
+./llama-completion.exe -m models\gemma-1.1-7b-it.Q4_K_M.gguf -no-cnv --prompt "Once upon a time"
 ```
 ##### Conversation mode (Allow for continuous interaction with the model)

 ```powershell
-./llama-cli.exe -m models\gemma-1.1-7b-it.Q4_K_M.gguf --chat-template gemma
+./llama-completion.exe -m models\gemma-1.1-7b-it.Q4_K_M.gguf --chat-template gemma
 ```

 ##### Conversation mode using built-in jinja chat template

 ```powershell
-./llama-cli.exe -m models\gemma-1.1-7b-it.Q4_K_M.gguf --jinja
+./llama-completion.exe -m models\gemma-1.1-7b-it.Q4_K_M.gguf --jinja
 ```

 ##### One-and-done query using jinja with custom system prompt and a starting prompt

 ```powershell
-./llama-cli.exe -m models\gemma-1.1-7b-it.Q4_K_M.gguf --jinja --single-turn -sys "You are a helpful assistant" -p "Hello"
+./llama-completion.exe -m models\gemma-1.1-7b-it.Q4_K_M.gguf --jinja --single-turn -sys "You are a helpful assistant" -p "Hello"
 ```

 #### Infinite text from a starting prompt (you can use `Ctrl-C` to stop it):

 ```powershell
-llama-cli.exe -m models\gemma-1.1-7b-it.Q4_K_M.gguf --ignore-eos -n -1
+llama-completion.exe -m models\gemma-1.1-7b-it.Q4_K_M.gguf --ignore-eos -n -1
 ```

 ## Common Options

-In this section, we cover the most commonly used options for running the `llama-cli` program with the LLaMA models:
+In this section, we cover the most commonly used options for running the `llama-completion` program with the LLaMA models:

 -   `-m FNAME, --model FNAME`: Specify the path to the LLaMA model file (e.g., `models/gemma-1.1-7b-it.Q4_K_M.gguf`; inferred from `--model-url` if set).
 -   `-mu MODEL_URL --model-url MODEL_URL`: Specify a remote http url to download the file (e.g [https://huggingface.co/ggml-org/gemma-1.1-7b-it-Q4_K_M-GGUF/resolve/main/gemma-1.1-7b-it.Q4_K_M.gguf?download=true](https://huggingface.co/ggml-org/gemma-1.1-7b-it-Q4_K_M-GGUF/resolve/main/gemma-1.1-7b-it.Q4_K_M.gguf?download=true)).
@ -97,7 +97,7 @@ In this section, we cover the most commonly used options for running the `llama-

 ## Input Prompts

-The `llama-cli` program provides several ways to interact with the LLaMA models using input prompts:
+The `llama-completion` program provides several ways to interact with the LLaMA models using input prompts:

 -   `--prompt PROMPT`: Provide a prompt directly as a command-line option.
 -   `--file FNAME`: Provide a file containing a prompt or multiple prompts.
@ -107,7 +107,7 @@ The `llama-cli` program provides several ways to interact with the LLaMA models

 ## Interaction

-The `llama-cli` program offers a seamless way to interact with LLaMA models, allowing users to engage in real-time conversations or provide instructions for specific tasks. The interactive mode can be triggered using various options, including `--interactive` and `--interactive-first`.
+The `llama-completion` program offers a seamless way to interact with LLaMA models, allowing users to engage in real-time conversations or provide instructions for specific tasks. The interactive mode can be triggered using various options, including `--interactive` and `--interactive-first`.

 In interactive mode, users can participate in text generation by injecting their input during the process. Users can press `Ctrl+C` at any time to interject and type their input, followed by pressing `Return` to submit it to the LLaMA model. To submit additional lines without finalizing input, users can end the current line with a backslash (`\`) and continue typing.

@ -136,7 +136,7 @@ To overcome this limitation, you can use the `--in-prefix` flag to add a space o
 The `--in-prefix` flag is used to add a prefix to your input, primarily, this is used to insert a space after the reverse prompt. Here's an example of how to use the `--in-prefix` flag in conjunction with the `--reverse-prompt` flag:

 ```sh
-./llama-cli -r "User:" --in-prefix " "
+./llama-completion -r "User:" --in-prefix " "
 ```

 ### In-Suffix
@ -144,7 +144,7 @@ The `--in-prefix` flag is used to add a prefix to your input, primarily, this is
 The `--in-suffix` flag is used to add a suffix after your input. This is useful for adding an "Assistant:" prompt after the user's input. It's added after the new-line character (`\n`) that's automatically added to the end of the user's input. Here's an example of how to use the `--in-suffix` flag in conjunction with the `--reverse-prompt` flag:

 ```sh
-./llama-cli -r "User:" --in-prefix " " --in-suffix "Assistant:"
+./llama-completion -r "User:" --in-prefix " " --in-suffix "Assistant:"
 ```
 When --in-prefix or --in-suffix options are enabled the chat template ( --chat-template ) is disabled

--- a/tools/fit-params/CMakeLists.txt
+++ b/tools/fit-params/CMakeLists.txt
@ -0,0 +1,8 @@
+set(TARGET llama-fit-params)
+add_executable(${TARGET} fit-params.cpp)
+target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})
+target_compile_features(${TARGET} PRIVATE cxx_std_17)
+
+if(LLAMA_TOOLS_INSTALL)
+    install(TARGETS ${TARGET} RUNTIME)
+endif()
--- a/tools/fit-params/README.md
+++ b/tools/fit-params/README.md
@ -0,0 +1,55 @@
+# fit-params
+
+llama.cpp binaries can automatically fit the projected memory use of a model to the free device memory available at runtime,
+this is controlled using the CLI arguments starting with `-fit`/`--fit`.
+Internally the code is calling `llama_params_fit` to adjust the `llama_model_params` and `llama_context_params` structs.
+`llama-fit-params` is a simple utility that prints the CLI arguments corresponding to these adjustments to stdout.
+Example usage:
+
+``` bash
+# First, run llama-fit-params and store the results in a file:
+> ./build/bin/llama-fit-params --model /opt/models/qwen_3-30b3a-f16.gguf | tee args.txt
+ggml_cuda_init: GGML_CUDA_FORCE_MMQ:    no
+ggml_cuda_init: GGML_CUDA_FORCE_CUBLAS: no
+ggml_cuda_init: found 1 CUDA devices:
+  Device 0: NVIDIA GeForce RTX 4090, compute capability 8.9, VMM: yes
+build: 6895 (4341dc8bc) with cc (GCC) 15.2.1 20250813 for x86_64-pc-linux-gnu
+llama_params_fit_impl: projected to use 61807 MiB of device memory vs. 24077 MiB of free device memory
+llama_params_fit_impl: cannot fulfill margin of 1024 MiB, need to reduce device memory by 42444 MiB
+llama_params_fit_impl: context size reduced from 40960 to 4096 -> need 3456 MiB less memory in total
+llama_params_fit_impl: with only dense weights in device memory there is a total surplus of 16164 MiB
+llama_params_fit_impl: distributing layers across devices with overflow to next device/system memory:
+llama_params_fit_impl:   - CUDA0 (NVIDIA GeForce RTX 4090): 48 layers (34 overflowing),  19187 MiB used,   1199 MiB free
+llama_params_fit: successfully fit params to free device memory
+llama_params_fit: fitting params to free memory took 1.15 seconds
+Printing fitted CLI arguments to stdout...
+-c 4096 -ngl 48 -ot blk\.14\.ffn_(up|down|gate)_(ch|)exps=CPU,blk\.15\.ffn_(up|down|gate)_(ch|)exps=CPU,blk\.16\.ffn_(up|down|gate)_(ch|)exps=CPU,blk\.17\.ffn_(up|down|gate)_(ch|)exps=CPU,blk\.18\.ffn_(up|down|gate)_(ch|)exps=CPU,blk\.19\.ffn_(up|down|gate)_(ch|)exps=CPU,blk\.20\.ffn_(up|down|gate)_(ch|)exps=CPU,blk\.21\.ffn_(up|down|gate)_(ch|)exps=CPU,blk\.22\.ffn_(up|down|gate)_(ch|)exps=CPU,blk\.23\.ffn_(up|down|gate)_(ch|)exps=CPU,blk\.24\.ffn_(up|down|gate)_(ch|)exps=CPU,blk\.25\.ffn_(up|down|gate)_(ch|)exps=CPU,blk\.26\.ffn_(up|down|gate)_(ch|)exps=CPU,blk\.27\.ffn_(up|down|gate)_(ch|)exps=CPU,blk\.28\.ffn_(up|down|gate)_(ch|)exps=CPU,blk\.29\.ffn_(up|down|gate)_(ch|)exps=CPU,blk\.30\.ffn_(up|down|gate)_(ch|)exps=CPU,blk\.31\.ffn_(up|down|gate)_(ch|)exps=CPU,blk\.32\.ffn_(up|down|gate)_(ch|)exps=CPU,blk\.33\.ffn_(up|down|gate)_(ch|)exps=CPU,blk\.34\.ffn_(up|down|gate)_(ch|)exps=CPU,blk\.35\.ffn_(up|down|gate)_(ch|)exps=CPU,blk\.36\.ffn_(up|down|gate)_(ch|)exps=CPU,blk\.37\.ffn_(up|down|gate)_(ch|)exps=CPU,blk\.38\.ffn_(up|down|gate)_(ch|)exps=CPU,blk\.39\.ffn_(up|down|gate)_(ch|)exps=CPU,blk\.40\.ffn_(up|down|gate)_(ch|)exps=CPU,blk\.41\.ffn_(up|down|gate)_(ch|)exps=CPU,blk\.42\.ffn_(up|down|gate)_(ch|)exps=CPU,blk\.43\.ffn_(up|down|gate)_(ch|)exps=CPU,blk\.44\.ffn_(up|down|gate)_(ch|)exps=CPU,blk\.45\.ffn_(up|down|gate)_(ch|)exps=CPU,blk\.46\.ffn_(up|down|gate)_(ch|)exps=CPU,blk\.47\.ffn_(up|down|gate)_(ch|)exps=CPU
+
+# Next, use those results for a llama.cpp binary:
+> cat args.txt | xargs ./build/bin/llama-server --model /opt/models/qwen_3-30b3a-f16.gguf
+ggml_cuda_init: GGML_CUDA_FORCE_MMQ:    no
+ggml_cuda_init: GGML_CUDA_FORCE_CUBLAS: no
+ggml_cuda_init: found 1 CUDA devices:
+  Device 0: NVIDIA GeForce RTX 4090, compute capability 8.9, VMM: yes
+build: 6895 (4341dc8bc) with cc (GCC) 15.2.1 20250813 for x86_64-pc-linux-gnu
+system info: n_threads = 16, n_threads_batch = 16, total_threads = 32
+
+system_info: n_threads = 16 (n_threads_batch = 16) / 32 | CUDA : ARCHS = 890 | USE_GRAPHS = 1 | PEER_MAX_BATCH_SIZE = 128 | CPU : SSE3 = 1 | SSSE3 = 1 | AVX = 1 | AVX_VNNI = 1 | AVX2 = 1 | F16C = 1 | FMA = 1 | BMI2 = 1 | AVX512 = 1 | AVX512_VBMI = 1 | AVX512_VNNI = 1 | AVX512_BF16 = 1 | LLAMAFILE = 1 | OPENMP = 1 | REPACK = 1 |
+
+main: binding port with default address family
+main: HTTP server is listening, hostname: 127.0.0.1, port: 8080, http threads: 31
+main: loading model
+srv    load_model: loading model '/opt/models/qwen_3-30b3a-f16.gguf'
+llama_params_fit_impl: projected to use 19187 MiB of device memory vs. 24077 MiB of free device memory
+llama_params_fit_impl: will leave 1199 >= 1024 MiB of free device memory, no changes needed
+llama_params_fit: successfully fit params to free device memory
+llama_params_fit: fitting params to free memory took 0.28 seconds
+[...]
+main: server is listening on http://127.0.0.1:8080 - starting the main loop
+srv  update_slots: all slots are idle
+^Csrv    operator(): operator(): cleaning up before exit...
+
+llama_memory_breakdown_print: | memory breakdown [MiB] | total   free     self   model   context   compute    unaccounted |
+llama_memory_breakdown_print: |   - CUDA0 (RTX 4090)   | 24077 =  945 + (19187 = 17904 +     384 +     898) +        3945 |
+llama_memory_breakdown_print: |   - Host               |                 58271 = 58259 +       0 +      12                |
+```
--- a/tools/fit-params/fit-params.cpp
+++ b/tools/fit-params/fit-params.cpp
@ -0,0 +1,62 @@
+#include "llama.h"
+
+#include "arg.h"
+#include "common.h"
+#include "log.h"
+
+#include <iostream>
+
+#if defined(_MSC_VER)
+#pragma warning(disable: 4244 4267) // possible loss of data
+#endif
+
+int main(int argc, char ** argv) {
+    common_params params;
+
+    if (!common_params_parse(argc, argv, params, LLAMA_EXAMPLE_COMMON)) {
+        return 1;
+    }
+
+    common_init();
+    llama_backend_init();
+    llama_numa_init(params.numa);
+    auto mparams = common_model_params_to_llama(params);
+    auto cparams = common_context_params_to_llama(params);
+    llama_params_fit(params.model.path.c_str(), &mparams, &cparams,
+        params.tensor_split, params.tensor_buft_overrides.data(), params.fit_params_target, params.fit_params_min_ctx,
+        params.verbosity >= 4 ? GGML_LOG_LEVEL_DEBUG : GGML_LOG_LEVEL_ERROR);
+
+    LOG_INF("Printing fitted CLI arguments to stdout...\n");
+    std::cout << "-c "    << cparams.n_ctx;
+    std::cout << " -ngl " << mparams.n_gpu_layers;
+
+    size_t nd = llama_max_devices();
+    while (nd > 1 && mparams.tensor_split[nd - 1] == 0.0f) {
+        nd--;
+    }
+    if (nd > 1) {
+        for (size_t id = 0; id < nd; id++) {
+            if (id == 0) {
+                std::cout << " -ts ";
+            }
+            if (id > 0) {
+                std::cout << ",";
+            }
+            std::cout << mparams.tensor_split[id];
+        }
+    }
+
+    const size_t ntbo = llama_max_tensor_buft_overrides();
+    for (size_t itbo = 0; itbo < ntbo && mparams.tensor_buft_overrides[itbo].pattern != nullptr; itbo++) {
+        if (itbo == 0) {
+            std::cout << " -ot ";
+        }
+        if (itbo > 0) {
+            std::cout << ",";
+        }
+        std::cout << mparams.tensor_buft_overrides[itbo].pattern << "=" << ggml_backend_buft_name(mparams.tensor_buft_overrides[itbo].buft);
+    }
+    std::cout << "\n";
+
+    return 0;
+}
--- a/tools/llama-bench/README.md
+++ b/tools/llama-bench/README.md
@ -80,7 +80,7 @@ Each test is repeated the number of times given by `-r`, and the results are ave

 Using the `-d <n>` option, each test can be run at a specified context depth, prefilling the KV cache with `<n>` tokens.

-For a description of the other options, see the [main example](../main/README.md).
+For a description of the other options, see the [completion example](../completion/README.md).

 > [!NOTE]
 > The measurements with `llama-bench` do not include the times for tokenization and for sampling.
--- a/tools/mtmd/clip-graph.h
+++ b/tools/mtmd/clip-graph.h
@ -112,4 +112,8 @@ struct clip_graph {
    // aka pixel_shuffle / pixel_unshuffle / patch_merger (Kimi-VL)
    // support dynamic resolution
    ggml_tensor * build_patch_merge_permute(ggml_tensor * cur, int scale_factor);
+
+    // Generic function to stack frames for audio processing
+    // Abstracts out the StackAudioFrames logic used by ultravox
+    ggml_tensor * build_stack(ggml_tensor * cur, int32_t stack_factor, int32_t n_embed);
 };
--- a/tools/mtmd/clip-impl.h
+++ b/tools/mtmd/clip-impl.h
@ -157,6 +157,7 @@ enum projector_type {
    PROJECTOR_TYPE_INTERNVL,
    PROJECTOR_TYPE_LLAMA4,
    PROJECTOR_TYPE_QWEN2A,
+    PROJECTOR_TYPE_GLMA,
    PROJECTOR_TYPE_QWEN25O, // will be replaced by QWEN2A or QWEN25VL depending on clip_ctx
    PROJECTOR_TYPE_VOXTRAL,
    PROJECTOR_TYPE_LFM2,
@ -183,6 +184,7 @@ static std::map<projector_type, std::string> PROJECTOR_TYPE_NAMES = {
    { PROJECTOR_TYPE_INTERNVL,  "internvl"},
    { PROJECTOR_TYPE_LLAMA4,    "llama4"},
    { PROJECTOR_TYPE_QWEN2A,    "qwen2a"},
+    { PROJECTOR_TYPE_GLMA,      "glma"},
    { PROJECTOR_TYPE_QWEN25O,   "qwen2.5o"},
    { PROJECTOR_TYPE_VOXTRAL,   "voxtral"},
    { PROJECTOR_TYPE_LFM2,      "lfm2"},
--- a/tools/mtmd/clip-model.h
+++ b/tools/mtmd/clip-model.h
@ -65,6 +65,13 @@ struct clip_hparams {
    int32_t n_mel_bins = 0; // whisper preprocessor
    int32_t proj_stack_factor = 0; // ultravox

+    // audio-to-mel preprocessor params
+    int32_t audio_chunk_len   = -1; // in seconds
+    int32_t audio_sample_rate = -1;
+    int32_t audio_n_fft       = -1;
+    int32_t audio_window_len  = -1;
+    int32_t audio_hop_len     = -1;
+
    // legacy
    bool has_llava_projector = false;
    int minicpmv_version = 0;
@ -256,6 +263,7 @@ struct clip_model {
    ggml_tensor * conv1d_2_w = nullptr;
    ggml_tensor * conv1d_2_b = nullptr;
    ggml_tensor * mm_norm_pre_w = nullptr;
+    ggml_tensor * mm_norm_pre_b = nullptr;
    ggml_tensor * mm_norm_mid_w = nullptr;

    // cogvlm
@ -277,3 +285,5 @@ struct clip_model {
            || proj_type == PROJECTOR_TYPE_VOXTRAL;
    }
 };
+
+const clip_hparams * clip_get_hparams(const struct clip_ctx * ctx);
--- a/tools/mtmd/clip.cpp
+++ b/tools/mtmd/clip.cpp
@ -720,6 +720,32 @@ ggml_tensor * clip_graph::build_rope_2d(
    return cur;
 }

+// Generic function to stack frames for audio processing
+// Abstracts out the StackAudioFrames logic used by ultravox
+ggml_tensor * clip_graph::build_stack(ggml_tensor * cur, int32_t stack_factor, int32_t n_embed) {
+    if (stack_factor <= 1) {
+        return cur;
+    }
+
+    int64_t total_elements = ggml_nelements(cur);
+    int64_t stride = n_embed * stack_factor;
+
+    // Calculate padded length
+    int64_t padded_len = GGML_PAD(total_elements, stride);
+    int64_t pad = padded_len - total_elements;
+
+    if (pad > 0) {
+        // Pad the tensor to make it divisible by stride
+        cur = ggml_view_1d(ctx0, cur, total_elements, 0);
+        cur = ggml_pad(ctx0, cur, pad, 0, 0, 0);
+    }
+
+    // Reshape to [stride, padded_len / stride]
+    cur = ggml_view_2d(ctx0, cur, stride, padded_len / stride,
+                        ggml_row_size(cur->type, stride), 0);
+    return cur;
+}
+
 // aka pixel_shuffle / pixel_unshuffle / patch_merger (Kimi-VL)
 // support dynamic resolution
 ggml_tensor * clip_graph::build_patch_merge_permute(ggml_tensor * cur, int scale_factor) {
@ -796,6 +822,7 @@ static ggml_cgraph * clip_image_build_graph(clip_ctx * ctx, const clip_image_f32
        case PROJECTOR_TYPE_ULTRAVOX:
        case PROJECTOR_TYPE_VOXTRAL:
        case PROJECTOR_TYPE_QWEN2A:
+        case PROJECTOR_TYPE_GLMA:
            {
                builder = std::make_unique<clip_graph_whisper_enc>(ctx, img);
            } break;
@ -1136,16 +1163,22 @@ struct clip_model_loader {
                    } break;
                case PROJECTOR_TYPE_ULTRAVOX:
                case PROJECTOR_TYPE_QWEN2A:
+                case PROJECTOR_TYPE_GLMA:
                case PROJECTOR_TYPE_VOXTRAL:
                    {
                        bool require_stack = model.proj_type == PROJECTOR_TYPE_ULTRAVOX ||
-                                             model.proj_type == PROJECTOR_TYPE_VOXTRAL;
+                                             model.proj_type == PROJECTOR_TYPE_VOXTRAL ||
+                                             model.proj_type == PROJECTOR_TYPE_GLMA;
                        get_u32(KEY_A_PROJ_STACK_FACTOR, hparams.proj_stack_factor, require_stack);
-                        if (hparams.n_mel_bins != 128) {
-                            throw std::runtime_error(string_format("%s: only 128 mel bins are supported for ultravox\n", __func__));
-                        }
                        hparams.ffn_op = FFN_GELU_ERF;
                        log_ffn_op = "gelu_erf"; // temporary solution for logging
+
+                        // audio preprocessing params
+                        hparams.audio_chunk_len    = 30; // in seconds
+                        hparams.audio_sample_rate  = 16000;
+                        hparams.audio_n_fft        = 400;
+                        hparams.audio_window_len   = 400;
+                        hparams.audio_hop_len      = 160;
                    } break;
                default:
                    break;
@ -1183,6 +1216,11 @@ struct clip_model_loader {
                LOG_INF("\n--- audio hparams ---\n");
                LOG_INF("%s: n_mel_bins:         %d\n", __func__, hparams.n_mel_bins);
                LOG_INF("%s: proj_stack_factor:  %d\n", __func__, hparams.proj_stack_factor);
+                LOG_INF("%s: audio_chunk_len:    %d\n", __func__, hparams.audio_chunk_len);
+                LOG_INF("%s: audio_sample_rate:  %d\n", __func__, hparams.audio_sample_rate);
+                LOG_INF("%s: audio_n_fft:        %d\n", __func__, hparams.audio_n_fft);
+                LOG_INF("%s: audio_window_len:   %d\n", __func__, hparams.audio_window_len);
+                LOG_INF("%s: audio_hop_len:      %d\n", __func__, hparams.audio_hop_len);
            }
            LOG_INF("\n");
            LOG_INF("%s: model size:         %.2f MiB\n", __func__, model_size / 1024.0 / 1024.0);
@ -1510,6 +1548,21 @@ struct clip_model_loader {
                    model.mm_3_w = get_tensor(string_format(TN_MVLM_PROJ_MLP, 3, "weight"));
                    model.mm_3_b = get_tensor(string_format(TN_MVLM_PROJ_MLP, 3, "bias"));
                } break;
+            case PROJECTOR_TYPE_GLMA:
+                {
+                    model.conv1d_1_w = get_tensor(string_format(TN_CONV1D, 1, "weight"));
+                    model.conv1d_1_b = get_tensor(string_format(TN_CONV1D, 1, "bias"));
+                    model.conv1d_2_w = get_tensor(string_format(TN_CONV1D, 2, "weight"));
+                    model.conv1d_2_b = get_tensor(string_format(TN_CONV1D, 2, "bias"));
+                    model.mm_1_w = get_tensor(string_format(TN_MM_AUDIO_MLP, 1, "weight"));
+                    model.mm_1_b = get_tensor(string_format(TN_MM_AUDIO_MLP, 1, "bias"));
+                    model.mm_2_w = get_tensor(string_format(TN_MM_AUDIO_MLP, 2, "weight"));
+                    model.mm_2_b = get_tensor(string_format(TN_MM_AUDIO_MLP, 2, "bias"));
+                    model.mm_norm_pre_w = get_tensor(string_format(TN_MM_NORM_PRE, "weight"));
+                    model.mm_norm_pre_b = get_tensor(string_format(TN_MM_NORM_PRE, "bias"));
+                    model.mm_boi = get_tensor(string_format(TN_TOK_BOI, "weight"));
+                    model.mm_eoi = get_tensor(string_format(TN_TOK_EOI, "weight"));
+                } break;
            case PROJECTOR_TYPE_LLAMA4:
                {
                    model.mm_model_proj    = get_tensor(TN_MM_PROJECTOR);
@ -2895,6 +2948,16 @@ int clip_n_output_tokens(const struct clip_ctx * ctx, struct clip_image_f32 * im
                    n_patches /= 2;
                }
            } break;
+        case PROJECTOR_TYPE_GLMA:
+            {
+                n_patches = img->nx;
+                // whisper downscales input token by half after conv1d
+                n_patches /= 2;
+                // reshape by merge_factor
+                n_patches /= ctx->model.hparams.proj_stack_factor;
+                // for BOI and EOI token embeddings
+                n_patches += 2;
+            } break;
        case PROJECTOR_TYPE_COGVLM:
            {
                n_patches += 2; // for BOI and EOI token embeddings
@ -3230,6 +3293,7 @@ bool clip_image_batch_encode(clip_ctx * ctx, const int n_threads, const clip_ima
        case PROJECTOR_TYPE_IDEFICS3:
        case PROJECTOR_TYPE_INTERNVL:
        case PROJECTOR_TYPE_QWEN2A:
+        case PROJECTOR_TYPE_GLMA:
        case PROJECTOR_TYPE_ULTRAVOX:
        case PROJECTOR_TYPE_LFM2:
        case PROJECTOR_TYPE_VOXTRAL:
@ -3340,6 +3404,8 @@ int clip_n_mmproj_embd(const struct clip_ctx * ctx) {
            return ctx->model.mm_model_proj->ne[1];
        case PROJECTOR_TYPE_QWEN2A:
            return ctx->model.mm_fc_w->ne[1];
+        case PROJECTOR_TYPE_GLMA:
+            return ctx->model.mm_2_w->ne[1];
        case PROJECTOR_TYPE_LFM2:
        case PROJECTOR_TYPE_KIMIVL:
            return ctx->model.mm_2_w->ne[1];
@ -3386,6 +3452,7 @@ bool clip_has_audio_encoder(const struct clip_ctx * ctx) {
 bool clip_has_whisper_encoder(const struct clip_ctx * ctx) {
    return ctx->proj_type() == PROJECTOR_TYPE_ULTRAVOX
        || ctx->proj_type() == PROJECTOR_TYPE_QWEN2A
+        || ctx->proj_type() == PROJECTOR_TYPE_GLMA
        || ctx->proj_type() == PROJECTOR_TYPE_VOXTRAL;
 }

@ -3420,3 +3487,7 @@ void clip_image_f32_batch_add_mel(struct clip_image_f32_batch * batch, int n_mel
    batch->entries.push_back(clip_image_f32_ptr(audio));
    batch->is_audio = true;
 }
+
+const clip_hparams * clip_get_hparams(const struct clip_ctx * ctx) {
+    return &ctx->model.hparams;
+}
--- a/tools/mtmd/models/whisper-enc.cpp
+++ b/tools/mtmd/models/whisper-enc.cpp
@ -30,7 +30,6 @@ ggml_cgraph * clip_graph_whisper_enc::build() {
    GGML_ASSERT(model.layers[0].q_b);
    GGML_ASSERT(model.layers[0].v_b);
    GGML_ASSERT(!model.layers[0].k_b); // no bias for k
-    GGML_ASSERT(model.post_ln_w && model.post_ln_b);

    ggml_tensor * pos_embd_selected = ggml_view_2d(
        ctx0, model.position_embeddings,
@ -49,15 +48,7 @@ ggml_cgraph * clip_graph_whisper_enc::build() {
    if (model.audio_has_stack_frames()) {
        // StackAudioFrames
        // https://huggingface.co/fixie-ai/ultravox-v0_5-llama-3_2-1b/blob/main/ultravox_model.py
-        int64_t stride = n_embd * hparams.proj_stack_factor;
-        int64_t padded_len = GGML_PAD(ggml_nelements(cur), stride);
-        int64_t pad = padded_len - ggml_nelements(cur);
-        if (pad > 0) {
-            cur = ggml_view_1d(ctx0, cur, ggml_nelements(cur), 0);
-            cur = ggml_pad(ctx0, cur, pad, 0, 0, 0);
-        }
-        cur = ggml_view_2d(ctx0, cur, stride, padded_len / stride,
-                            ggml_row_size(cur->type, stride), 0);
+        cur = build_stack(cur, hparams.proj_stack_factor, n_embd);
        cb(cur, "after_stacked", -1);
    }

@ -95,6 +86,14 @@ ggml_cgraph * clip_graph_whisper_enc::build() {
            FFN_GELU_ERF,
            -1);

+    } else if (proj_type == PROJECTOR_TYPE_GLMA) {
+            cur = ggml_norm(ctx0, cur, hparams.eps);
+            cur = ggml_mul(ctx0, cur, model.mm_norm_pre_w);
+            cur = ggml_add(ctx0, cur, model.mm_norm_pre_b);
+            cur = build_stack(cur, hparams.proj_stack_factor, n_embd);
+            cur = build_ffn(cur, model.mm_1_w, model.mm_1_b, nullptr, nullptr, model.mm_2_w, model.mm_2_b, hparams.ffn_op, 0);
+            cur = ggml_concat(ctx0, model.mm_boi, cur, 1);
+            cur = ggml_concat(ctx0, cur, model.mm_eoi, 1);
    } else {
        GGML_ABORT("%s: unknown projector type", __func__);
    }
--- a/tools/mtmd/mtmd-audio.cpp
+++ b/tools/mtmd/mtmd-audio.cpp
--- a/tools/mtmd/mtmd-audio.h
+++ b/tools/mtmd/mtmd-audio.h
@ -1,6 +1,7 @@
 #pragma once

 #include "ggml.h"
+#include "clip-model.h"

 #include <cstdint>
 #include <vector>
@ -8,18 +9,7 @@

 #define MTMD_INTERNAL_HEADER

-#define WHISPER_ASSERT GGML_ASSERT
-
-#define WHISPER_SAMPLE_RATE 16000
-#define WHISPER_N_FFT       400
-#define WHISPER_HOP_LENGTH  160
-#define WHISPER_CHUNK_SIZE  30
-
-#define COMMON_SAMPLE_RATE 16000
-
-namespace whisper_preprocessor {
-
-struct whisper_mel {
+struct mtmd_audio_mel {
    int n_len;
    int n_len_org;
    int n_mel;
@ -27,23 +17,18 @@ struct whisper_mel {
    std::vector<float> data;
 };

-struct whisper_filters {
-    int32_t n_mel;
-    int32_t n_fft;
+struct mtmd_audio_preprocessor {
+    const clip_hparams & hparams;

-    std::vector<float> data;
+    mtmd_audio_preprocessor(const clip_ctx * ctx): hparams(*clip_get_hparams(ctx)) {}
+
+    virtual ~mtmd_audio_preprocessor() = default;
+    virtual void initialize() = 0; // NOT thread-safe
+    virtual bool preprocess(const float * samples, size_t n_samples, std::vector<mtmd_audio_mel> & output) = 0;
 };

-bool preprocess_audio(
-        const float * samples,
-        size_t n_samples,
-        const whisper_filters & filters,
-        std::vector<whisper_mel> & output);
-
-} // namespace whisper_preprocessor
-
-namespace whisper_precalc_filters {
-
-whisper_preprocessor::whisper_filters get_128_bins();
-
-} // namespace whisper_precalc_filters
+struct mtmd_audio_preprocessor_whisper : mtmd_audio_preprocessor {
+    mtmd_audio_preprocessor_whisper(const clip_ctx * ctx) : mtmd_audio_preprocessor(ctx) {}
+    void initialize() override;
+    bool preprocess(const float * samples, size_t n_samples, std::vector<mtmd_audio_mel> & output) override;
+};
--- a/tools/mtmd/mtmd.cpp
+++ b/tools/mtmd/mtmd.cpp
@ -151,8 +151,7 @@ struct mtmd_context {
    // string template for slice image delimiters with row/col (idefics3)
    std::string sli_img_start_tmpl;

-    // for whisper, we pre-calculate the mel filter bank
-    whisper_preprocessor::whisper_filters w_filters;
+    std::unique_ptr<mtmd_audio_preprocessor> audio_preproc;

    // TODO @ngxson : add timings

@ -317,14 +316,25 @@ struct mtmd_context {
        GGML_ASSERT(ctx_a != nullptr);
        projector_type proj = clip_get_projector_type(ctx_a);

-        if (clip_has_whisper_encoder(ctx_a)) {
-            // TODO @ngxson : check if model n_mel is 128 or 80
-            w_filters = whisper_precalc_filters::get_128_bins();
-        }
-
        LOG_WRN("%s: audio input is in experimental stage and may have reduced quality:\n"
                "    https://github.com/ggml-org/llama.cpp/discussions/13759\n", __func__);

+        // set preprocessor
+        switch (proj) {
+            case PROJECTOR_TYPE_QWEN2A:
+            case PROJECTOR_TYPE_QWEN25O:
+            case PROJECTOR_TYPE_ULTRAVOX:
+            case PROJECTOR_TYPE_VOXTRAL:
+                audio_preproc = std::make_unique<mtmd_audio_preprocessor_whisper>(ctx_a);
+                break;
+            default:
+                GGML_ABORT("unsupported audio projector type");
+        }
+
+        // initialize audio preprocessor
+        audio_preproc->initialize();
+
+        // set special tokens
        if (proj == PROJECTOR_TYPE_QWEN2A) {
            // <|audio_bos|> ... (embeddings) ... <|audio_eos|>
            aud_beg = "<|audio_bos|>";
@ -653,11 +663,10 @@ struct mtmd_tokenizer {
            }

            // preprocess audio
-            GGML_ASSERT(ctx->w_filters.n_mel); // make sure we have filter preloaded
-            std::vector<whisper_preprocessor::whisper_mel> mel_spec_chunks;
+            std::vector<mtmd_audio_mel> mel_spec_chunks;
            const float * samples = (const float *)bitmap->data.data();
            size_t n_samples = bitmap->data.size() / sizeof(float);
-            bool ok = whisper_preprocessor::preprocess_audio(samples, n_samples, ctx->w_filters, mel_spec_chunks);
+            bool ok = ctx->audio_preproc->preprocess(samples, n_samples, mel_spec_chunks);
            if (!ok) {
                LOG_ERR("Unable to preprocess audio\n");
                return 2;
@ -863,8 +872,7 @@ int mtmd_get_audio_bitrate(mtmd_context * ctx) {
    if (!ctx->ctx_a) {
        return -1;
    }
-    // for now, we assume that all audio models have the same bitrate
-    return 16000; // 16kHz
+    return clip_get_hparams(ctx->ctx_a)->audio_sample_rate;
 }

 //
--- a/tools/server/public/index.html.gz
+++ b/tools/server/public/index.html.gz
--- a/tools/server/webui/src/lib/components/app/chat/ChatSettings/ChatSettingsImportExportTab.svelte
+++ b/tools/server/webui/src/lib/components/app/chat/ChatSettings/ChatSettingsImportExportTab.svelte
@ -1,9 +1,11 @@
 <script lang="ts">
-	import { Download, Upload } from '@lucide/svelte';
+	import { Download, Upload, Trash2 } from '@lucide/svelte';
 	import { Button } from '$lib/components/ui/button';
 	import { DialogConversationSelection } from '$lib/components/app';
 	import { createMessageCountMap } from '$lib/utils';
 	import { conversationsStore, conversations } from '$lib/stores/conversations.svelte';
+	import { toast } from 'svelte-sonner';
+	import DialogConfirmation from '$lib/components/app/dialogs/DialogConfirmation.svelte';

 	let exportedConversations = $state<DatabaseConversation[]>([]);
 	let importedConversations = $state<DatabaseConversation[]>([]);
@ -18,11 +20,14 @@
 		[]
 	);

+	// Delete functionality state
+	let showDeleteDialog = $state(false);
+
 	async function handleExportClick() {
 		try {
 			const allConversations = conversations();
 			if (allConversations.length === 0) {
-				alert('No conversations to export');
+				toast.info('No conversations to export');
 				return;
 			}

@ -145,6 +150,36 @@
 			alert('Failed to import conversations. Please check the file format.');
 		}
 	}
+
+	async function handleDeleteAllClick() {
+		try {
+			const allConversations = conversations();
+
+			if (allConversations.length === 0) {
+				toast.info('No conversations to delete');
+				return;
+			}
+
+			showDeleteDialog = true;
+		} catch (err) {
+			console.error('Failed to load conversations for deletion:', err);
+			toast.error('Failed to load conversations');
+		}
+	}
+
+	async function handleDeleteAllConfirm() {
+		try {
+			await conversationsStore.deleteAll();
+
+			showDeleteDialog = false;
+		} catch (err) {
+			console.error('Failed to delete conversations:', err);
+		}
+	}
+
+	function handleDeleteAllCancel() {
+		showDeleteDialog = false;
+	}
 </script>

 <div class="space-y-6">
@ -229,6 +264,25 @@
 				</div>
 			{/if}
 		</div>
+
+		<div class="grid border-t border-border/30 pt-4">
+			<h4 class="mb-2 text-sm font-medium text-destructive">Delete All Conversations</h4>
+
+			<p class="mb-4 text-sm text-muted-foreground">
+				Permanently delete all conversations and their messages. This action cannot be undone.
+				Consider exporting your conversations first if you want to keep a backup.
+			</p>
+
+			<Button
+				class="text-destructive-foreground w-full justify-start justify-self-start bg-destructive hover:bg-destructive/80 md:w-auto"
+				onclick={handleDeleteAllClick}
+				variant="destructive"
+			>
+				<Trash2 class="mr-2 h-4 w-4" />
+
+				Delete all conversations
+			</Button>
+		</div>
 	</div>
 </div>

@ -249,3 +303,15 @@
 	onCancel={() => (showImportDialog = false)}
 	onConfirm={handleImportConfirm}
 />
+
+<DialogConfirmation
+	bind:open={showDeleteDialog}
+	title="Delete all conversations"
+	description="Are you sure you want to delete all conversations? This action cannot be undone and will permanently remove all your conversations and messages."
+	confirmText="Delete All"
+	cancelText="Cancel"
+	variant="destructive"
+	icon={Trash2}
+	onConfirm={handleDeleteAllConfirm}
+	onCancel={handleDeleteAllCancel}
+/>
--- a/tools/server/webui/src/lib/stores/conversations.svelte.ts
+++ b/tools/server/webui/src/lib/stores/conversations.svelte.ts
@ -385,8 +385,7 @@ class ConversationsStore {
 			this.conversations = this.conversations.filter((c) => c.id !== convId);

 			if (this.activeConversation?.id === convId) {
-				this.activeConversation = null;
-				this.activeMessages = [];
+				this.clearActiveConversation();
 				await goto(`?new_chat=true#/`);
 			}
 		} catch (error) {
@ -394,6 +393,29 @@ class ConversationsStore {
 		}
 	}

+	/**
+	 * Deletes all conversations and their messages
+	 */
+	async deleteAll(): Promise<void> {
+		try {
+			const allConversations = await DatabaseService.getAllConversations();
+
+			for (const conv of allConversations) {
+				await DatabaseService.deleteConversation(conv.id);
+			}
+
+			this.clearActiveConversation();
+			this.conversations = [];
+
+			toast.success('All conversations deleted');
+
+			await goto(`?new_chat=true#/`);
+		} catch (error) {
+			console.error('Failed to delete all conversations:', error);
+			toast.error('Failed to delete conversations');
+		}
+	}
+
 	// ─────────────────────────────────────────────────────────────────────────────
 	// Import/Export
 	// ─────────────────────────────────────────────────────────────────────────────