Upload 6 files

.gitattributes

@@ -33,3 +33,6 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+librkllmrt.so filter=lfs diff=lfs merge=lfs -text
+Qwen3-Embedding-0.6B_f16.rkllm filter=lfs diff=lfs merge=lfs -text
+Qwen3-Embedding-0.6B_w8a8.rkllm filter=lfs diff=lfs merge=lfs -text

Qwen3-Embedding-0.6B_f16.rkllm

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4a383a715837f432a8af06798e236ab70a30b80d55cdbe4d7e4d489de639310b
+size 1524801182

Qwen3-Embedding-0.6B_w8a8.rkllm

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:be4ffc5c46b6246c71459081c5fdb8b8984c52936fa23083e342f4a843945806
+size 931372078

librkllmrt.so

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f6a9c2de93cf94bb524eb071c27190ad4c83401e01b562534f265dff4cb40da2
+size 6710712

rkllm-convert.py

@@ -0,0 +1,74 @@
+import os
+from rkllm.api import RKLLM
+
+def convert_model(model_path, output_name, do_quantization=False):
+    """转换单个模型"""
+    llm = RKLLM()
+    
+    print(f"正在加载模型: {model_path}")
+    ret = llm.load_huggingface(model=model_path, model_lora=None, device='cpu')
+    if ret != 0:
+        print(f'加载模型失败: {model_path}')
+        return ret
+    
+    print(f"正在构建模型: {output_name} (量化: {do_quantization})")
+    qparams = None
+    ret = llm.build(do_quantization=do_quantization, optimization_level=1, quantized_dtype='w8a8',
+                    quantized_algorithm='normal', target_platform='rk3588', num_npu_core=3, extra_qparams=qparams)
+    
+    if ret != 0:
+        print(f'构建模型失败: {output_name}')
+        return ret
+    
+    # 导出rkllm模型
+    print(f"正在导出模型: {output_name}")
+    ret = llm.export_rkllm(output_name)
+    if ret != 0:
+        print(f'导出模型失败: {output_name}')
+        return ret
+    
+    print(f"成功转换: {output_name}")
+    return 0
+
+def main():
+    """主函数：遍历所有子文件夹并转换模型"""
+    current_dir = '.'
+    
+    # 获取所有子文件夹
+    subdirs = [d for d in os.listdir(current_dir) 
+               if os.path.isdir(os.path.join(current_dir, d)) and not d.startswith('.')]
+    
+    print(f"找到 {len(subdirs)} 个模型文件夹: {subdirs}")
+    
+    for subdir in subdirs:
+        model_path = os.path.join(current_dir, subdir)
+        
+        # 生成输出文件名
+        base_name = subdir.replace('/', '_').replace('\\', '_')
+        quantized_output = f"{base_name}_w8a8.rkllm"
+        unquantized_output = f"{base_name}_f16.rkllm"
+        
+        print(f"\n{'='*50}")
+        print(f"处理模型文件夹: {subdir}")
+        print(f"{'='*50}")
+        
+        # 转换非量化版本
+        print(f"\n--- 转换非量化版本 ---")
+        ret = convert_model(model_path, unquantized_output, do_quantization=False)
+        if ret != 0:
+            print(f"非量化版本转换失败: {subdir}")
+            continue
+        
+        # 转换量化版本
+        print(f"\n--- 转换量化版本 ---")
+        ret = convert_model(model_path, quantized_output, do_quantization=True)
+        if ret != 0:
+            print(f"量化版本转换失败: {subdir}")
+            continue
+        
+        print(f"\n✓ {subdir} 模型转换完成!")
+        print(f"  - 非量化版本: {unquantized_output}")
+        print(f"  - 量化版本: {quantized_output}")
+
+if __name__ == "__main__":
+    main()

rkllm_binding.py

@@ -0,0 +1,658 @@
+import ctypes
+import enum
+import os
+
+# Define constants from the header
+CPU0 = (1 << 0)  # 0x01
+CPU1 = (1 << 1)  # 0x02
+CPU2 = (1 << 2)  # 0x04
+CPU3 = (1 << 3)  # 0x08
+CPU4 = (1 << 4)  # 0x10
+CPU5 = (1 << 5)  # 0x20
+CPU6 = (1 << 6)  # 0x40
+CPU7 = (1 << 7)  # 0x80
+
+# --- Enums ---
+class LLMCallState(enum.IntEnum):
+    RKLLM_RUN_NORMAL = 0
+    RKLLM_RUN_WAITING = 1
+    RKLLM_RUN_FINISH = 2
+    RKLLM_RUN_ERROR = 3
+
+class RKLLMInputType(enum.IntEnum):
+    RKLLM_INPUT_PROMPT = 0
+    RKLLM_INPUT_TOKEN = 1
+    RKLLM_INPUT_EMBED = 2
+    RKLLM_INPUT_MULTIMODAL = 3
+
+class RKLLMInferMode(enum.IntEnum):
+    RKLLM_INFER_GENERATE = 0
+    RKLLM_INFER_GET_LAST_HIDDEN_LAYER = 1
+    RKLLM_INFER_GET_LOGITS = 2
+
+# --- Structures ---
+class RKLLMExtendParam(ctypes.Structure):
+    # 基础iommu domain ID, 对>1b的模型建议设置为1
+    base_domain_id: ctypes.c_int32
+    # 是否使用flash存储Embedding
+    embed_flash: ctypes.c_int8
+    # 启用的cpu核心数
+    enabled_cpus_num: ctypes.c_int8
+    # 启用的cpu核心掩码
+    enabled_cpus_mask: ctypes.c_uint32
+    reserved: ctypes.c_uint8 * 106
+
+    _fields_ = [
+        ("base_domain_id", ctypes.c_int32),
+        ("embed_flash", ctypes.c_int8),
+        ("enabled_cpus_num", ctypes.c_int8),
+        ("enabled_cpus_mask", ctypes.c_uint32),
+        ("reserved", ctypes.c_uint8 * 106)
+    ]
+
+class RKLLMParam(ctypes.Structure):
+    # 模型文件路径
+    model_path: ctypes.c_char_p
+    # 上下文窗口最大token数
+    max_context_len: ctypes.c_int32
+    # 最大生成新token数
+    max_new_tokens: ctypes.c_int32
+    # Top-K采样参数
+    top_k: ctypes.c_int32
+    # 上下文窗口移动时保留的kv缓存数量
+    n_keep: ctypes.c_int32
+    # Top-P采样参数
+    top_p: ctypes.c_float
+    # 采样温度，影响token选择的随机性
+    temperature: ctypes.c_float
+    # 重复token惩罚
+    repeat_penalty: ctypes.c_float
+    # 频繁token惩罚
+    frequency_penalty: ctypes.c_float
+    # 输入中已存在token的惩罚
+    presence_penalty: ctypes.c_float
+    # Mirostat采样策略标志（0表示禁用）
+    mirostat: ctypes.c_int32
+    # Mirostat采样Tau参数
+    mirostat_tau: ctypes.c_float
+    # Mirostat采样Eta参数
+    mirostat_eta: ctypes.c_float
+    # 是否跳过特殊token
+    skip_special_token: ctypes.c_bool
+    # 是否异步推理
+    is_async: ctypes.c_bool
+    # 多模态输入中图像的起始Token
+    img_start: ctypes.c_char_p
+    # 多模态输入中图像的结束Token
+    img_end: ctypes.c_char_p
+    # 图像内容指针
+    img_content: ctypes.c_char_p
+    # 扩展参数
+    extend_param: RKLLMExtendParam
+
+    _fields_ = [
+        ("model_path", ctypes.c_char_p),         # 模型文件路径
+        ("max_context_len", ctypes.c_int32),     # 上下文窗口最大token数
+        ("max_new_tokens", ctypes.c_int32),      # 最大生成新token数
+        ("top_k", ctypes.c_int32),               # Top-K采样参数
+        ("n_keep", ctypes.c_int32),              # 上下文窗口移动时保留的kv缓存数量
+        ("top_p", ctypes.c_float),               # Top-P（nucleus）采样参数
+        ("temperature", ctypes.c_float),         # 采样温度，影响token选择的随机性
+        ("repeat_penalty", ctypes.c_float),      # 重复token惩罚
+        ("frequency_penalty", ctypes.c_float),   # 频繁token惩罚
+        ("presence_penalty", ctypes.c_float),    # 输入中已存在token的惩罚
+        ("mirostat", ctypes.c_int32),            # Mirostat采样策略标志（0表示禁用）
+        ("mirostat_tau", ctypes.c_float),        # Mirostat采样Tau参数
+        ("mirostat_eta", ctypes.c_float),        # Mirostat采样Eta参数
+        ("skip_special_token", ctypes.c_bool),   # 是否跳过特殊token
+        ("is_async", ctypes.c_bool),             # 是否异步推理
+        ("img_start", ctypes.c_char_p),          # 多模态输入中图像的起始Token
+        ("img_end", ctypes.c_char_p),            # 多模态输入中图像的结束Token
+        ("img_content", ctypes.c_char_p),        # 图像内容指针
+        ("extend_param", RKLLMExtendParam)       # 扩展参数
+    ]
+
+class RKLLMLoraAdapter(ctypes.Structure):
+    lora_adapter_path: ctypes.c_char_p
+    lora_adapter_name: ctypes.c_char_p
+    scale: ctypes.c_float
+
+    _fields_ = [
+        ("lora_adapter_path", ctypes.c_char_p),
+        ("lora_adapter_name", ctypes.c_char_p),
+        ("scale", ctypes.c_float)
+    ]
+
+class RKLLMEmbedInput(ctypes.Structure):
+    # Shape: [n_tokens, embed_size]
+    embed: ctypes.POINTER(ctypes.c_float)
+    n_tokens: ctypes.c_size_t
+
+    _fields_ = [
+        ("embed", ctypes.POINTER(ctypes.c_float)),  
+        ("n_tokens", ctypes.c_size_t)
+    ]
+
+class RKLLMTokenInput(ctypes.Structure):
+    # Shape: [n_tokens]
+    input_ids: ctypes.POINTER(ctypes.c_int32)
+    n_tokens: ctypes.c_size_t
+
+    _fields_ = [
+        ("input_ids", ctypes.POINTER(ctypes.c_int32)),
+        ("n_tokens", ctypes.c_size_t)
+    ]
+
+class RKLLMMultiModelInput(ctypes.Structure):
+    prompt: ctypes.c_char_p
+    image_embed: ctypes.POINTER(ctypes.c_float)
+    n_image_tokens: ctypes.c_size_t
+    n_image: ctypes.c_size_t
+    image_width: ctypes.c_size_t
+    image_height: ctypes.c_size_t
+
+    _fields_ = [
+        ("prompt", ctypes.c_char_p),
+        ("image_embed", ctypes.POINTER(ctypes.c_float)),
+        ("n_image_tokens", ctypes.c_size_t),
+        ("n_image", ctypes.c_size_t),
+        ("image_width", ctypes.c_size_t),
+        ("image_height", ctypes.c_size_t)
+    ]
+
+class _RKLLMInputUnion(ctypes.Union):
+    prompt_input: ctypes.c_char_p
+    embed_input: RKLLMEmbedInput
+    token_input: RKLLMTokenInput
+    multimodal_input: RKLLMMultiModelInput
+
+    _fields_ = [
+        ("prompt_input", ctypes.c_char_p),
+        ("embed_input", RKLLMEmbedInput),
+        ("token_input", RKLLMTokenInput),
+        ("multimodal_input", RKLLMMultiModelInput)
+    ]
+
+class RKLLMInput(ctypes.Structure):
+    input_type: ctypes.c_int
+    _union_data: _RKLLMInputUnion
+
+    _fields_ = [
+        ("input_type", ctypes.c_int), # Enum will be passed as int, changed RKLLMInputType to ctypes.c_int
+        ("_union_data", _RKLLMInputUnion)
+    ]
+    # Properties to make accessing union members easier
+    @property
+    def prompt_input(self) -> bytes: # Assuming c_char_p maps to bytes
+        if self.input_type == RKLLMInputType.RKLLM_INPUT_PROMPT:
+            return self._union_data.prompt_input
+        raise AttributeError("Not a prompt input")
+    @prompt_input.setter
+    def prompt_input(self, value: bytes): # Assuming c_char_p maps to bytes
+        if self.input_type == RKLLMInputType.RKLLM_INPUT_PROMPT:
+            self._union_data.prompt_input = value
+        else:
+            raise AttributeError("Not a prompt input")
+    @property
+    def embed_input(self) -> RKLLMEmbedInput:
+        if self.input_type == RKLLMInputType.RKLLM_INPUT_EMBED:
+            return self._union_data.embed_input
+        raise AttributeError("Not an embed input")
+    @embed_input.setter
+    def embed_input(self, value: RKLLMEmbedInput):
+        if self.input_type == RKLLMInputType.RKLLM_INPUT_EMBED:
+            self._union_data.embed_input = value
+        else:
+            raise AttributeError("Not an embed input")
+
+    @property
+    def token_input(self) -> RKLLMTokenInput:
+        if self.input_type == RKLLMInputType.RKLLM_INPUT_TOKEN:
+            return self._union_data.token_input
+        raise AttributeError("Not a token input")
+    @token_input.setter
+    def token_input(self, value: RKLLMTokenInput):
+        if self.input_type == RKLLMInputType.RKLLM_INPUT_TOKEN:
+            self._union_data.token_input = value
+        else:
+            raise AttributeError("Not a token input")
+
+    @property
+    def multimodal_input(self) -> RKLLMMultiModelInput:
+        if self.input_type == RKLLMInputType.RKLLM_INPUT_MULTIMODAL:
+            return self._union_data.multimodal_input
+        raise AttributeError("Not a multimodal input")
+    @multimodal_input.setter
+    def multimodal_input(self, value: RKLLMMultiModelInput):
+        if self.input_type == RKLLMInputType.RKLLM_INPUT_MULTIMODAL:
+            self._union_data.multimodal_input = value
+        else:
+            raise AttributeError("Not a multimodal input")
+
+class RKLLMLoraParam(ctypes.Structure): # For inference
+    lora_adapter_name: ctypes.c_char_p
+
+    _fields_ = [
+        ("lora_adapter_name", ctypes.c_char_p)
+    ]
+
+class RKLLMPromptCacheParam(ctypes.Structure): # For inference
+    save_prompt_cache: ctypes.c_int # bool-like
+    prompt_cache_path: ctypes.c_char_p
+
+    _fields_ = [
+        ("save_prompt_cache", ctypes.c_int), # bool-like
+        ("prompt_cache_path", ctypes.c_char_p)
+    ]
+
+class RKLLMInferParam(ctypes.Structure):
+    mode: ctypes.c_int
+    lora_params: ctypes.POINTER(RKLLMLoraParam)
+    prompt_cache_params: ctypes.POINTER(RKLLMPromptCacheParam)
+    keep_history: ctypes.c_int # bool-like
+
+    _fields_ = [
+        ("mode", ctypes.c_int), # Enum will be passed as int, changed RKLLMInferMode to ctypes.c_int
+        ("lora_params", ctypes.POINTER(RKLLMLoraParam)),
+        ("prompt_cache_params", ctypes.POINTER(RKLLMPromptCacheParam)),
+        ("keep_history", ctypes.c_int) # bool-like
+    ]
+
+class RKLLMResultLastHiddenLayer(ctypes.Structure):
+    # Shape: [num_tokens, embd_size]
+    hidden_states: ctypes.POINTER(ctypes.c_float)
+    # 隐藏层大小
+    embd_size: ctypes.c_int
+    # 输出token数
+    num_tokens: ctypes.c_int
+
+    _fields_ = [
+        ("hidden_states", ctypes.POINTER(ctypes.c_float)),
+        ("embd_size", ctypes.c_int),
+        ("num_tokens", ctypes.c_int)
+    ]
+
+class RKLLMResultLogits(ctypes.Structure):
+    # Shape: [num_tokens, vocab_size]
+    logits: ctypes.POINTER(ctypes.c_float)
+    # 词汇表大小
+    vocab_size: ctypes.c_int
+    # 输出token数
+    num_tokens: ctypes.c_int
+
+    _fields_ = [
+        ("logits", ctypes.POINTER(ctypes.c_float)),
+        ("vocab_size", ctypes.c_int),
+        ("num_tokens", ctypes.c_int)
+    ]
+
+class RKLLMResult(ctypes.Structure):
+    text: ctypes.c_char_p
+    token_id: ctypes.c_int32
+    last_hidden_layer: RKLLMResultLastHiddenLayer
+    logits: RKLLMResultLogits
+
+    _fields_ = [
+        ("text", ctypes.c_char_p),
+        ("token_id", ctypes.c_int32),
+        ("last_hidden_layer", RKLLMResultLastHiddenLayer),
+        ("logits", RKLLMResultLogits)
+    ]
+
+# --- Typedefs ---
+LLMHandle = ctypes.c_void_p
+
+# --- Callback Function Type ---
+LLMResultCallback = ctypes.CFUNCTYPE(
+    None,  # return type: void
+    ctypes.POINTER(RKLLMResult),
+    ctypes.c_void_p,  # userdata
+    ctypes.c_int      # enum, will be passed as int. Changed LLMCallState to ctypes.c_int
+)
+
+
+class RKLLMRuntime:
+    def __init__(self, library_path="./librkllmrt.so"):
+        try:
+            self.lib = ctypes.CDLL(library_path)
+        except OSError as e:
+            raise OSError(f"Failed to load RKLLM library from {library_path}. "
+                          f"Ensure it's in your LD_LIBRARY_PATH or provide the full path. Error: {e}")
+        self._setup_functions()
+        self.llm_handle = LLMHandle()
+        self._c_callback = None # To keep the callback object alive
+
+    def _setup_functions(self):
+        # RKLLMParam rkllm_createDefaultParam();
+        self.lib.rkllm_createDefaultParam.restype = RKLLMParam
+        self.lib.rkllm_createDefaultParam.argtypes = []
+
+        # int rkllm_init(LLMHandle* handle, RKLLMParam* param, LLMResultCallback callback);
+        self.lib.rkllm_init.restype = ctypes.c_int
+        self.lib.rkllm_init.argtypes = [
+            ctypes.POINTER(LLMHandle),
+            ctypes.POINTER(RKLLMParam),
+            LLMResultCallback
+        ]
+
+        # int rkllm_load_lora(LLMHandle handle, RKLLMLoraAdapter* lora_adapter);
+        self.lib.rkllm_load_lora.restype = ctypes.c_int
+        self.lib.rkllm_load_lora.argtypes = [LLMHandle, ctypes.POINTER(RKLLMLoraAdapter)]
+
+        # int rkllm_load_prompt_cache(LLMHandle handle, const char* prompt_cache_path);
+        self.lib.rkllm_load_prompt_cache.restype = ctypes.c_int
+        self.lib.rkllm_load_prompt_cache.argtypes = [LLMHandle, ctypes.c_char_p]
+
+        # int rkllm_release_prompt_cache(LLMHandle handle);
+        self.lib.rkllm_release_prompt_cache.restype = ctypes.c_int
+        self.lib.rkllm_release_prompt_cache.argtypes = [LLMHandle]
+
+        # int rkllm_destroy(LLMHandle handle);
+        self.lib.rkllm_destroy.restype = ctypes.c_int
+        self.lib.rkllm_destroy.argtypes = [LLMHandle]
+
+        # int rkllm_run(LLMHandle handle, RKLLMInput* rkllm_input, RKLLMInferParam* rkllm_infer_params, void* userdata);
+        self.lib.rkllm_run.restype = ctypes.c_int
+        self.lib.rkllm_run.argtypes = [
+            LLMHandle,
+            ctypes.POINTER(RKLLMInput),
+            ctypes.POINTER(RKLLMInferParam),
+            ctypes.c_void_p # userdata
+        ]
+
+        # int rkllm_run_async(LLMHandle handle, RKLLMInput* rkllm_input, RKLLMInferParam* rkllm_infer_params, void* userdata);
+        # Assuming async also takes userdata for the callback context
+        self.lib.rkllm_run_async.restype = ctypes.c_int
+        self.lib.rkllm_run_async.argtypes = [
+            LLMHandle,
+            ctypes.POINTER(RKLLMInput),
+            ctypes.POINTER(RKLLMInferParam),
+            ctypes.c_void_p # userdata
+        ]
+
+        # int rkllm_abort(LLMHandle handle);
+        self.lib.rkllm_abort.restype = ctypes.c_int
+        self.lib.rkllm_abort.argtypes = [LLMHandle]
+
+        # int rkllm_is_running(LLMHandle handle);
+        self.lib.rkllm_is_running.restype = ctypes.c_int # 0 if running, non-zero otherwise
+        self.lib.rkllm_is_running.argtypes = [LLMHandle]
+
+        # int rkllm_clear_kv_cache(LLMHandle handle, int keep_system_prompt);
+        self.lib.rkllm_clear_kv_cache.restype = ctypes.c_int
+        self.lib.rkllm_clear_kv_cache.argtypes = [LLMHandle, ctypes.c_int]
+
+        # int rkllm_set_chat_template(LLMHandle handle, const char* system_prompt, const char* prompt_prefix, const char* prompt_postfix);
+        self.lib.rkllm_set_chat_template.restype = ctypes.c_int
+        self.lib.rkllm_set_chat_template.argtypes = [
+            LLMHandle,
+            ctypes.c_char_p,
+            ctypes.c_char_p,
+            ctypes.c_char_p
+        ]
+
+    def create_default_param(self) -> RKLLMParam:
+        """Creates a default RKLLMParam structure."""
+        return self.lib.rkllm_createDefaultParam()
+
+    def init(self, param: RKLLMParam, callback_func) -> int:
+        """
+        Initializes the LLM.
+        :param param: RKLLMParam structure.
+        :param callback_func: A Python function that matches the signature:
+                              def my_callback(result_ptr, userdata_ptr, state_enum):
+                                  result = result_ptr.contents # RKLLMResult
+                                  # Process result
+                                  # userdata can be retrieved if passed during run, or ignored
+                                  # state = LLMCallState(state_enum)
+        :return: 0 for success, non-zero for failure.
+        """
+        if not callable(callback_func):
+            raise ValueError("callback_func must be a callable Python function.")
+
+        # Keep a reference to the ctypes callback object to prevent it from being garbage collected
+        self._c_callback = LLMResultCallback(callback_func)
+        
+        ret = self.lib.rkllm_init(ctypes.byref(self.llm_handle), ctypes.byref(param), self._c_callback)
+        if ret != 0:
+            raise RuntimeError(f"rkllm_init failed with error code {ret}")
+        return ret
+
+    def load_lora(self, lora_adapter: RKLLMLoraAdapter) -> int:
+        """Loads a Lora adapter."""
+        ret = self.lib.rkllm_load_lora(self.llm_handle, ctypes.byref(lora_adapter))
+        if ret != 0:
+            raise RuntimeError(f"rkllm_load_lora failed with error code {ret}")
+        return ret
+
+    def load_prompt_cache(self, prompt_cache_path: str) -> int:
+        """Loads a prompt cache from a file."""
+        c_path = prompt_cache_path.encode('utf-8')
+        ret = self.lib.rkllm_load_prompt_cache(self.llm_handle, c_path)
+        if ret != 0:
+            raise RuntimeError(f"rkllm_load_prompt_cache failed for {prompt_cache_path} with error code {ret}")
+        return ret
+
+    def release_prompt_cache(self) -> int:
+        """Releases the prompt cache from memory."""
+        ret = self.lib.rkllm_release_prompt_cache(self.llm_handle)
+        if ret != 0:
+            raise RuntimeError(f"rkllm_release_prompt_cache failed with error code {ret}")
+        return ret
+
+    def destroy(self) -> int:
+        """Destroys the LLM instance and releases resources."""
+        if self.llm_handle and self.llm_handle.value: # Check if handle is not NULL
+            ret = self.lib.rkllm_destroy(self.llm_handle)
+            self.llm_handle = LLMHandle() # Reset handle
+            if ret != 0:
+                # Don't raise here as it might be called in __del__
+                print(f"Warning: rkllm_destroy failed with error code {ret}") 
+            return ret
+        return 0 # Already destroyed or not initialized
+
+    def run(self, rkllm_input: RKLLMInput, rkllm_infer_params: RKLLMInferParam, userdata=None) -> int:
+        """Runs an LLM inference task synchronously."""
+        # userdata can be a ctypes.py_object if you want to pass Python objects,
+        # then cast to c_void_p. Or simply None.
+        if userdata is not None:
+            # Store the userdata object to keep it alive during the call
+            self._userdata_ref = userdata
+            c_userdata = ctypes.cast(ctypes.pointer(ctypes.py_object(userdata)), ctypes.c_void_p)
+        else:
+            c_userdata = None
+        ret = self.lib.rkllm_run(self.llm_handle, ctypes.byref(rkllm_input), ctypes.byref(rkllm_infer_params), c_userdata)
+        if ret != 0:
+            raise RuntimeError(f"rkllm_run failed with error code {ret}")
+        return ret
+
+    def run_async(self, rkllm_input: RKLLMInput, rkllm_infer_params: RKLLMInferParam, userdata=None) -> int:
+        """Runs an LLM inference task asynchronously."""
+        if userdata is not None:
+            # Store the userdata object to keep it alive during the call
+            self._userdata_ref = userdata
+            c_userdata = ctypes.cast(ctypes.pointer(ctypes.py_object(userdata)), ctypes.c_void_p)
+        else:
+            c_userdata = None
+        ret = self.lib.rkllm_run_async(self.llm_handle, ctypes.byref(rkllm_input), ctypes.byref(rkllm_infer_params), c_userdata)
+        if ret != 0:
+            raise RuntimeError(f"rkllm_run_async failed with error code {ret}")
+        return ret
+
+    def abort(self) -> int:
+        """Aborts an ongoing LLM task."""
+        ret = self.lib.rkllm_abort(self.llm_handle)
+        if ret != 0:
+            raise RuntimeError(f"rkllm_abort failed with error code {ret}")
+        return ret
+
+    def is_running(self) -> bool:
+        """Checks if an LLM task is currently running. Returns True if running."""
+        # The C API returns 0 if running, non-zero otherwise.
+        # This is a bit counter-intuitive for a boolean "is_running".
+        return self.lib.rkllm_is_running(self.llm_handle) == 0
+
+    def clear_kv_cache(self, keep_system_prompt: bool) -> int:
+        """Clears the key-value cache."""
+        ret = self.lib.rkllm_clear_kv_cache(self.llm_handle, ctypes.c_int(1 if keep_system_prompt else 0))
+        if ret != 0:
+            raise RuntimeError(f"rkllm_clear_kv_cache failed with error code {ret}")
+        return ret
+
+    def set_chat_template(self, system_prompt: str, prompt_prefix: str, prompt_postfix: str) -> int:
+        """Sets the chat template for the LLM."""
+        c_system = system_prompt.encode('utf-8') if system_prompt else b""
+        c_prefix = prompt_prefix.encode('utf-8') if prompt_prefix else b""
+        c_postfix = prompt_postfix.encode('utf-8') if prompt_postfix else b""
+        
+        ret = self.lib.rkllm_set_chat_template(self.llm_handle, c_system, c_prefix, c_postfix)
+        if ret != 0:
+            raise RuntimeError(f"rkllm_set_chat_template failed with error code {ret}")
+        return ret
+
+    def __enter__(self):
+        return self
+
+    def __exit__(self, exc_type, exc_val, exc_tb):
+        self.destroy()
+
+    def __del__(self):
+        self.destroy() # Ensure resources are freed if object is garbage collected
+
+# --- Example Usage (Illustrative) ---
+if __name__ == "__main__":
+    # This is a placeholder for how you might use it.
+    # You'll need a valid .rkllm model and librkllmrt.so in your path.
+
+    # Global list to store results from callback for demonstration
+    results_buffer = []
+
+    def my_python_callback(result_ptr, userdata_ptr, state_enum):
+        """
+        Callback function to be called by the C library.
+        """
+        global results_buffer
+        state = LLMCallState(state_enum)
+        result = result_ptr.contents
+
+        current_text = ""
+        if result.text: # Check if the char_p is not NULL
+            current_text = result.text.decode('utf-8', errors='ignore')
+        
+        print(f"Callback: State={state.name}, TokenID={result.token_id}, Text='{current_text}'")
+        results_buffer.append(current_text)
+
+        if state == LLMCallState.RKLLM_RUN_FINISH:
+            print("Inference finished.")
+        elif state == LLMCallState.RKLLM_RUN_ERROR:
+            print("Inference error.")
+        
+        # Example: Accessing logits if available (and if mode was set to get logits)
+        # if result.logits.logits and result.logits.vocab_size > 0:
+        #     print(f"  Logits (first 5 of vocab_size {result.logits.vocab_size}):")
+        #     for i in range(min(5, result.logits.vocab_size)):
+        #         print(f"    {result.logits.logits[i]:.4f}", end=" ")
+        #     print()
+
+
+    # --- Attempt to use the wrapper ---
+    try:
+        print("Initializing RKLLMRuntime...")
+        # Adjust library_path if librkllmrt.so is not in default search paths
+        # e.g., library_path="./path/to/librkllmrt.so"
+        rk_llm = RKLLMRuntime() 
+
+        print("Creating default parameters...")
+        params = rk_llm.create_default_param()
+
+        # --- Configure parameters ---
+        # THIS IS CRITICAL: model_path must point to an actual .rkllm file
+        # For this example to run, you need a model file.
+        # Let's assume a dummy path for now, this will fail at init if not valid.
+        model_file = "dummy_model.rkllm" 
+        if not os.path.exists(model_file):
+            print(f"Warning: Model file '{model_file}' does not exist. Init will likely fail.")
+            # Create a dummy file for the example to proceed further, though init will still fail
+            # with a real library unless it's a valid model.
+            with open(model_file, "w") as f:
+                f.write("dummy content")
+
+        params.model_path = model_file.encode('utf-8')
+        params.max_context_len = 512
+        params.max_new_tokens = 128
+        params.top_k = 1 # Greedy
+        params.temperature = 0.7
+        params.repeat_penalty = 1.1
+        # ... set other params as needed
+
+        print(f"Initializing LLM with model: {params.model_path.decode()}...")
+        # This will likely fail if dummy_model.rkllm is not a valid model recognized by the library
+        try:
+            rk_llm.init(params, my_python_callback)
+            print("LLM Initialized.")
+        except RuntimeError as e:
+            print(f"Error during LLM initialization: {e}")
+            print("This is expected if 'dummy_model.rkllm' is not a valid model.")
+            print("Replace 'dummy_model.rkllm' with a real model path to test further.")
+            exit()
+
+
+        # --- Prepare input ---
+        print("Preparing input...")
+        rk_input = RKLLMInput()
+        rk_input.input_type = RKLLMInputType.RKLLM_INPUT_PROMPT
+        
+        prompt_text = "Translate the following English text to French: 'Hello, world!'"
+        c_prompt = prompt_text.encode('utf-8')
+        rk_input._union_data.prompt_input = c_prompt # Accessing union member directly
+
+        # --- Prepare inference parameters ---
+        print("Preparing inference parameters...")
+        infer_params = RKLLMInferParam()
+        infer_params.mode = RKLLMInferMode.RKLLM_INFER_GENERATE
+        infer_params.keep_history = 1 # True
+        # infer_params.lora_params = None # or set up RKLLMLoraParam if using LoRA
+        # infer_params.prompt_cache_params = None # or set up RKLLMPromptCacheParam
+
+        # --- Run inference ---
+        print(f"Running inference with prompt: '{prompt_text}'")
+        results_buffer.clear()
+        try:
+            rk_llm.run(rk_input, infer_params) # Userdata is None by default
+            print("\n--- Full Response ---")
+            print("".join(results_buffer))
+            print("---------------------\n")
+        except RuntimeError as e:
+            print(f"Error during LLM run: {e}")
+
+
+        # --- Example: Set chat template (if model supports it) ---
+        # print("Setting chat template...")
+        # try:
+        #     rk_llm.set_chat_template("You are a helpful assistant.", "<user>: ", "<assistant>: ")
+        #     print("Chat template set.")
+        # except RuntimeError as e:
+        #     print(f"Error setting chat template: {e}")
+
+        # --- Example: Clear KV Cache ---
+        # print("Clearing KV cache (keeping system prompt if any)...")
+        # try:
+        #     rk_llm.clear_kv_cache(keep_system_prompt=True)
+        #     print("KV cache cleared.")
+        # except RuntimeError as e:
+        #     print(f"Error clearing KV cache: {e}")
+
+    except OSError as e:
+        print(f"OSError: {e}. Could not load the RKLLM library.")
+        print("Please ensure 'librkllmrt.so' is in your LD_LIBRARY_PATH or provide the full path.")
+    except Exception as e:
+        print(f"An unexpected error occurred: {e}")
+    finally:
+        if 'rk_llm' in locals() and rk_llm.llm_handle and rk_llm.llm_handle.value:
+            print("Destroying LLM instance...")
+            rk_llm.destroy()
+            print("LLM instance destroyed.")
+        if os.path.exists(model_file) and model_file == "dummy_model.rkllm":
+             os.remove(model_file) # Clean up dummy file
+
+    print("Example finished.")

test_embedding.py

@@ -0,0 +1,396 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+Qwen3-Embedding-0.6B 推理测试代码
+使用 RKLLM API 进行文本嵌入推理
+"""
+import faulthandler
+faulthandler.enable()
+import os
+os.environ["RKLLM_LOG_LEVEL"] = "1"
+import numpy as np
+import time
+from typing import List, Dict, Any
+from rkllm_binding import *
+
+
+class Qwen3EmbeddingTester:
+    def __init__(self, model_path: str, library_path: str = "./librkllmrt.so"):
+        """
+        初始化 Qwen3 嵌入模型测试器
+        
+        Args:
+            model_path: 模型文件路径（.rkllm 格式）
+            library_path: RKLLM 库文件路径
+        """
+        self.model_path = model_path
+        self.library_path = library_path
+        self.runtime = None
+        self.embeddings_buffer = []
+        self.current_result = None
+        
+    def callback_function(self, result_ptr, userdata_ptr, state_enum):
+        """
+        推理回调函数
+        
+        Args:
+            result_ptr: 结果指针
+            userdata_ptr: 用户数据指针
+            state_enum: 状态枚举
+        """
+        state = LLMCallState(state_enum)
+        
+        if state == LLMCallState.RKLLM_RUN_NORMAL:
+            result = result_ptr.contents
+            print(f"result: {result}")
+            # 获取最后隐藏层输出作为嵌入
+            if result.last_hidden_layer.hidden_states and result.last_hidden_layer.embd_size > 0:
+                embd_size = result.last_hidden_layer.embd_size
+                num_tokens = result.last_hidden_layer.num_tokens
+                
+                print(f"获取到嵌入向量：维度={embd_size}, 令牌数={num_tokens}")
+                
+                # 将 C 数组转换为 numpy 数组
+                # 这里我们取最后一个 token 的隐藏状态作为句子嵌入
+                if num_tokens > 0:
+                    # 获取最后一个 token 的嵌入（shape: [embd_size]）
+                    last_token_embedding = np.array([
+                        result.last_hidden_layer.hidden_states[(num_tokens-1) * embd_size + i] 
+                        for i in range(embd_size)
+                    ])
+                    
+                    self.current_result = {
+                        'embedding': last_token_embedding,
+                        'embd_size': embd_size,
+                        'num_tokens': num_tokens
+                    }
+                    
+                    print(f"嵌入向量范数: {np.linalg.norm(last_token_embedding):.4f}")
+                    print(f"嵌入向量前10维: {last_token_embedding[:10]}")
+                
+        elif state == LLMCallState.RKLLM_RUN_ERROR:
+            print("推理过程发生错误")
+            
+    def init_model(self):
+        """初始化模型"""
+        try:
+            print(f"初始化 RKLLM 运行时，库路径: {self.library_path}")
+            self.runtime = RKLLMRuntime(self.library_path)
+            
+            print("创建默认参数...")
+            params = self.runtime.create_default_param()
+            
+            # 配置参数
+            params.model_path = self.model_path.encode('utf-8')
+            params.max_context_len = 1024  # 设置上下文长度
+            params.max_new_tokens = 1      # 嵌入任务不需要生成新token
+            params.temperature = 1.0       # 嵌入任务温度设置
+            params.top_k = 1              # 嵌入任务不需要采样
+            params.top_p = 1.0            # 嵌入任务不需要采样
+            
+            # 扩展参数配置
+            params.extend_param.base_domain_id = 1  # 建议为 >1B 模型设置为1
+            params.extend_param.embed_flash = 0     # 是否使用flash存储Embedding
+            params.extend_param.enabled_cpus_num = 4  # 启用的CPU核心数
+            params.extend_param.enabled_cpus_mask = 0x0F  # CPU核心掩码
+            
+            print(f"初始化模型: {self.model_path}")
+            self.runtime.init(params, self.callback_function)
+            self.runtime.set_chat_template("","","")
+            print("模型初始化成功！")
+            
+        except Exception as e:
+            print(f"模型初始化失败: {e}")
+            raise
+            
+    def get_detailed_instruct(self, task_description: str, query: str) -> str:
+        """
+        构建指令提示词（参考 README 中的用法）
+        
+        Args:
+            task_description: 任务描述
+            query: 查询文本
+            
+        Returns:
+            格式化的指令提示词
+        """
+        return f'Instruct: {task_description}\nQuery: {query}'
+    
+    def encode_text(self, text: str, task_description: str = None) -> np.ndarray:
+        """
+        编码文本为嵌入向量
+        
+        Args:
+            text: 要编码的文本
+            task_description: 任务描述，如果提供则使用指令提示
+            
+        Returns:
+            嵌入向量（numpy数组）
+        """
+        try:
+            # 如果提供了任务描述，则使用指令提示
+            if task_description:
+                input_text = self.get_detailed_instruct(task_description, text)
+            else:
+                input_text = text
+                
+            print(f"编码文本: {input_text[:100]}{'...' if len(input_text) > 100 else ''}")
+            
+            # 准备输入
+            rk_input = RKLLMInput()
+            rk_input.input_type = RKLLMInputType.RKLLM_INPUT_PROMPT
+            c_prompt = input_text.encode('utf-8')
+            rk_input._union_data.prompt_input = c_prompt
+            
+            # 准备推理参数
+            infer_params = RKLLMInferParam()
+            infer_params.mode = RKLLMInferMode.RKLLM_INFER_GET_LAST_HIDDEN_LAYER  # 获取隐藏层输出
+            infer_params.keep_history = 0  # 不保留历史
+            
+            # 清空之前的结果
+            self.current_result = None
+            self.runtime.clear_kv_cache(False)
+            
+            # 执行推理
+            start_time = time.time()
+            self.runtime.run(rk_input, infer_params)
+            end_time = time.time()
+            
+            print(f"推理耗时: {end_time - start_time:.3f}秒")
+            
+            if self.current_result and 'embedding' in self.current_result:
+                # 对嵌入向量进行L2标准化
+                embedding = self.current_result['embedding']
+                normalized_embedding = embedding / np.linalg.norm(embedding)
+                return normalized_embedding
+            else:
+                raise RuntimeError("未能获取到有效的嵌入向量")
+                
+        except Exception as e:
+            print(f"编码文本时发生错误: {e}")
+            raise
+    
+    def compute_similarity(self, emb1: np.ndarray, emb2: np.ndarray) -> float:
+        """
+        计算两个嵌入向量的余弦相似度
+        
+        Args:
+            emb1: 第一个嵌入向量
+            emb2: 第二个嵌入向量
+            
+        Returns:
+            余弦相似度值
+        """
+        return np.dot(emb1, emb2)
+    
+    def test_embedding_similarity(self):
+        """测试嵌入相似度计算"""
+        print("\n" + "="*50)
+        print("测试嵌入相似度计算")
+        print("="*50)
+        
+        # 测试文本
+        task_description = "Given a web search query, retrieve relevant passages that answer the query"
+        
+        queries = [
+            "What is the capital of China?",
+            "Explain gravity"
+        ]
+        
+        documents = [
+            "The capital of China is Beijing.",
+            "Gravity is a force that attracts two bodies towards each other. It gives weight to physical objects and is responsible for the movement of planets around the sun."
+        ]
+        
+        # 编码查询（使用指令）
+        print("\n编码查询文本:")
+        query_embeddings = []
+        for i, query in enumerate(queries):
+            print(f"\n查询 {i+1}: {query}")
+            emb = self.encode_text(query, task_description)
+            query_embeddings.append(emb)
+        
+        # 编码文档（不使用指令）
+        print("\n编码文档文本:")
+        doc_embeddings = []
+        for i, doc in enumerate(documents):
+            print(f"\n文档 {i+1}: {doc}")
+            emb = self.encode_text(doc)
+            doc_embeddings.append(emb)
+        
+        # 计算相似度矩阵
+        print("\n计算相似度矩阵:")
+        print("查询 vs 文档相似度:")
+        print("-" * 30)
+        
+        similarities = []
+        for i, q_emb in enumerate(query_embeddings):
+            row_similarities = []
+            for j, d_emb in enumerate(doc_embeddings):
+                sim = self.compute_similarity(q_emb, d_emb)
+                row_similarities.append(sim)
+                print(f"查询{i+1} vs 文档{j+1}: {sim:.4f}")
+            similarities.append(row_similarities)
+            print()
+        
+        return similarities
+    
+    def test_multilingual_embedding(self):
+        """测试多语言嵌入能力"""
+        print("\n" + "="*50)
+        print("测试多语言嵌入能力")
+        print("="*50)
+        
+        # 多语言测试文本（相同含义的不同语言）
+        texts = {
+            "英语": "Hello, how are you?",
+            "中文": "你好，你好吗？",
+            "法语": "Bonjour, comment allez-vous?",
+            "西班牙语": "Hola, ¿cómo estás?",
+            "日语": "こんにちは、元気ですか？"
+        }
+        
+        embeddings = {}
+        print("\n编码多语言文本:")
+        for lang, text in texts.items():
+            print(f"\n{lang}: {text}")
+            emb = self.encode_text(text)
+            embeddings[lang] = emb
+        
+        # 计算跨语言相似度
+        print("\n跨语言相似度:")
+        print("-" * 30)
+        
+        languages = list(texts.keys())
+        for i, lang1 in enumerate(languages):
+            for j, lang2 in enumerate(languages):
+                if i <= j:
+                    sim = self.compute_similarity(embeddings[lang1], embeddings[lang2])
+                    print(f"{lang1} vs {lang2}: {sim:.4f}")
+    
+    def test_code_embedding(self):
+        """测试代码嵌入能力"""
+        print("\n" + "="*50)
+        print("测试代码嵌入能力")
+        print("="*50)
+        
+        # 代码示例
+        codes = {
+            "Python函数": """
+def fibonacci(n):
+    if n <= 1:
+        return n
+    return fibonacci(n-1) + fibonacci(n-2)
+""",
+            "JavaScript函数": """
+function fibonacci(n) {
+    if (n <= 1) return n;
+    return fibonacci(n-1) + fibonacci(n-2);
+}
+""",
+            "C++函数": """
+int fibonacci(int n) {
+    if (n <= 1) return n;
+    return fibonacci(n-1) + fibonacci(n-2);
+}
+""",
+            "数组排序": """
+def bubble_sort(arr):
+    n = len(arr)
+    for i in range(n):
+        for j in range(0, n-i-1):
+            if arr[j] > arr[j+1]:
+                arr[j], arr[j+1] = arr[j+1], arr[j]
+"""
+        }
+        
+        embeddings = {}
+        print("\n编码代码文本:")
+        for name, code in codes.items():
+            print(f"\n{name}:")
+            print(code[:100] + "..." if len(code) > 100 else code)
+            emb = self.encode_text(code)
+            embeddings[name] = emb
+        
+        # 计算代码相似度
+        print("\n代码相似度:")
+        print("-" * 30)
+        
+        code_names = list(codes.keys())
+        for i, name1 in enumerate(code_names):
+            for j, name2 in enumerate(code_names):
+                if i <= j:
+                    sim = self.compute_similarity(embeddings[name1], embeddings[name2])
+                    print(f"{name1} vs {name2}: {sim:.4f}")
+    
+    def cleanup(self):
+        """清理资源"""
+        if self.runtime:
+            try:
+                self.runtime.destroy()
+                print("模型资源已清理")
+            except Exception as e:
+                print(f"清理资源时发生错误: {e}")
+
+def main():
+    """主函数"""
+    import argparse
+    
+    # 解析命令行参数
+    parser = argparse.ArgumentParser(description='Qwen3-Embedding-0.6B 推理测试')
+    parser.add_argument('model_path', help='模型文件路径(.rkllm格式)')
+    parser.add_argument('--library_path', default="./librkllmrt.so", help='RKLLM库文件路径(默认为./librkllmrt.so)')
+    args = parser.parse_args()
+    
+    # 检查文件是否存在
+    if not os.path.exists(args.model_path):
+        print(f"错误: 模型文件不存在: {args.model_path}")
+        print("请确保:")
+        print("1. 已下载 Qwen3-Embedding-0.6B 模型")
+        print("2. 已使用 rkllm-convert.py 将模型转换为 .rkllm 格式")
+        return
+    
+    if not os.path.exists(args.library_path):
+        print(f"错误: RKLLM 库文件不存在: {args.library_path}")
+        print("请确保 librkllmrt.so 在当前目录或 LD_LIBRARY_PATH 中")
+        return
+    
+    print("Qwen3-Embedding-0.6B 推理测试")
+    print("=" * 50)
+    
+    # 创建测试器
+    tester = Qwen3EmbeddingTester(args.model_path, args.library_path)
+    
+    try:
+        # 初始化模型
+        tester.init_model()
+        
+        # 运行测试
+        print("\n开始运行嵌入测试...")
+        
+        # 测试基础嵌入相似度
+        tester.test_embedding_similarity()
+        
+        # 测试多语言嵌入
+        tester.test_multilingual_embedding()
+        
+        # 测试代码嵌入
+        tester.test_code_embedding()
+        
+        print("\n" + "="*50)
+        print("所有测试完成！")
+        print("="*50)
+        
+    except KeyboardInterrupt:
+        print("\n测试被用户中断")
+    except Exception as e:
+        print(f"\n测试过程中发生错误: {e}")
+        import traceback
+        traceback.print_exc()
+    finally:
+        # 清理资源
+        tester.cleanup()
+
+
+if __name__ == "__main__":
+    main()