Synced repo using 'sync_with_huggingface' Github Action

original:
- remote: "https://github.com/xiaoyao9184/convert-to-rkllm"
- commit: "ca00a54835367e950dc5fefb69185fb5616253f1"
sync_with_huggingface:
- repository: ""
- ref: ""

README.md

@@ -1,12 +1,14 @@
 ---
-title: Convert To Rkllm
-emoji: 🏢
-colorFrom: blue
-colorTo: gray
+title: Convert To RKLLM
+emoji: 🚲
+colorFrom: yellow
+colorTo: blue
 sdk: gradio
-sdk_version: 5.35.0
+sdk_version: 5.34.2
 app_file: app.py
 pinned: false
+license: apache-2.0
+short_description: 'Convert HuggingFace models to RKLLM and upload'
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app.py

@@ -0,0 +1,623 @@
+
+import builtins
+import logging
+import os
+import sys
+import shutil
+import uuid
+import json
+import re
+import contextvars
+import requests
+import torch
+import gradio as gr
+from huggingface_hub import HfApi, whoami, snapshot_download
+from rkllm.api import RKLLM
+from dataclasses import dataclass, field
+from pathlib import Path
+from typing import Optional, Tuple, Callable
+from enum import Enum
+from tqdm import tqdm
+from contextlib import suppress
+
+
+class Platform(Enum):
+    RK3588 = "RK3588"
+    RK3576 = "RK3576"
+    RK3562 = "RK3562"
+
+@dataclass
+class Config:
+    """Application configuration."""
+
+    _id: Optional[str] = field(default=None, init=False)
+    _logger: Optional[logging.Logger] = field(default=None, init=False)
+    _logger_path: Optional[Path] = field(default=None, init=False)
+
+    hf_token: str
+    hf_username: str
+    is_using_user_token: bool
+    ignore_converted: bool = False
+    ignore_errors: bool = False
+
+    hf_base_url: str = "https://huggingface.co"
+    output_path: Path = Path("./models")
+    cache_path: Path = Path("./cache")
+    log_path: Path = Path("./logs")
+    mapping_path: Path = Path(os.path.join(os.path.dirname(__file__), "mapping.json"))
+    dataset_path: Path = Path(os.path.join(os.path.dirname(__file__), "dataset.json"))
+
+    @classmethod
+    def from_env(cls) -> "Config":
+        """Create config from environment variables and secrets."""
+        system_token = os.getenv("HF_TOKEN")
+
+        if system_token and system_token.startswith("/run/secrets/") and os.path.isfile(system_token):
+            with open(system_token, "r") as f:
+                system_token = f.read().strip()
+
+        hf_username = (
+            os.getenv("SPACE_AUTHOR_NAME") or whoami(token=system_token)["name"]
+        )
+
+        output_dir = os.getenv("OUTPUT_DIR") or "./models"
+        cache_dir = os.getenv("HUGGINGFACE_HUB_CACHE") or os.getenv("CACHE_DIR") or "./cache"
+        log_dir = os.getenv("LOG_DIR") or "./logs"
+        mapping_json = os.getenv("MAPPING_JSON") or Path(os.path.join(os.path.dirname(__file__), "mapping.json"))
+        dataset_json = os.getenv("DATASET_JSON") or Path(os.path.join(os.path.dirname(__file__), "dataset.json"))
+
+        output_path = Path(output_dir)
+        output_path.mkdir(parents=True, exist_ok=True)
+        cache_path = Path(cache_dir)
+        cache_path.mkdir(parents=True, exist_ok=True)
+        log_path = Path(log_dir)
+        log_path.mkdir(parents=True, exist_ok=True)
+        mapping_path = Path(mapping_json)
+        mapping_path.parent.mkdir(parents=True, exist_ok=True)
+        dataset_path = Path(dataset_json)
+        dataset_path.parent.mkdir(parents=True, exist_ok=True)
+
+        return cls(
+            hf_token=system_token,
+            hf_username=hf_username,
+            is_using_user_token=False,
+            ignore_converted=os.getenv("IGNORE_CONVERTED", "false") == "true",
+            ignore_errors=os.getenv("IGNORE_ERRORS", "false") == "true",
+            output_path=output_path,
+            cache_path=cache_path,
+            log_path=log_path,
+            mapping_path=mapping_path,
+            dataset_path=dataset_path
+        )
+
+    @property
+    def id(self):
+        if not self._id:
+            self._id = str(uuid.uuid4())
+        return self._id
+
+    @property
+    def logger(self) -> logging.Logger:
+        """Get logger."""
+        if not self._logger:
+            logger = logging.getLogger(self.id)
+            logger.setLevel(logging.INFO)
+            if not logger.handlers:
+                handler = logging.FileHandler(self.logger_path)
+                handler.setFormatter(logging.Formatter("[%(levelname)s] - %(message)s"))
+                logger.addHandler(handler)
+                logger.propagate = False
+            self._logger = logger
+        return self._logger
+
+    @property
+    def logger_path(self) -> Path:
+        """Get logger path."""
+        if not self._logger_path:
+            logger_path = self.log_path / f"{self.id}.log"
+            self._logger_path = logger_path
+        return self._logger_path
+
+    def token(self, user_token):
+        """Update token."""
+        if user_token:
+            hf_username = whoami(token=user_token)["name"]
+        else:
+            hf_username = (
+                os.getenv("SPACE_AUTHOR_NAME") or whoami(token=self.hf_token)["name"]
+            )
+
+        hf_token = user_token or self.hf_token
+
+        if not hf_token:
+            raise ValueError(
+                "When the user token is not provided, the system token must be set."
+            )
+
+        self.hf_token = hf_token
+        self.hf_username = hf_username
+        self.is_using_user_token = bool(user_token)
+
+class ProgressLogger:
+    """Logger with progress update."""
+
+    def __init__(self, logger: logging.Logger, updater: Callable[[int], None]):
+        self.logger = logger
+        self.updater = updater
+        self.last_progress = 1
+        self.last_message = None
+        self.write_count = 0
+
+    def update(self, percent):
+        if percent >= self.last_progress:
+            self.updater(percent - self.last_progress)
+        else:
+            self.updater(self.last_progress - percent)
+        self.last_progress = min(self.last_progress, percent)
+
+    def print(self, *args, **kwargs):
+        self.last_message = " ".join(str(arg) for arg in args)
+        if self.logger:
+            self.logger.info(self.last_message.removeprefix("\r"))
+
+        if self.last_message.startswith("\rProgress:"):
+            with suppress(Exception):
+                percent_str = self.last_message.strip().split()[-1].strip('%')
+                percent = float(percent_str)
+                self.update(percent)
+                self.last_progress = percent
+
+    def write(self, text, write):
+        match = re.search(r"pre-uploaded: \d+/\d+ \(([\d.]+)M/([\d.]+)M\)", text)
+        if match:
+            with suppress(Exception):
+                current = float(match.group(1))
+                total = float(match.group(2))
+                percent = current / total * 100
+                self.update(percent)
+                self.write_count += 1
+        # 60 count for each second
+        if self.write_count > 60:
+            self.write_count = 0
+            write(text)
+
+class RedirectHandler(logging.Handler):
+    """Handles logging redirection to progress logger."""
+
+    def __init__(self, context: contextvars.ContextVar, logger: logging.Logger = None):
+        super().__init__(logging.NOTSET)
+        self.context = context
+        self.logger = logger
+
+    def emit(self, record: logging.LogRecord):
+        progress_logger = self.context.get(None)
+
+        if progress_logger:
+            try:
+                progress_logger.logger.handle(record)
+            except Exception as e:
+                self.logger.debug(f"Failed to redirection log: {e}")
+        elif self.logger:
+            self.logger.handle(record)
+
+class ModelConverter:
+    """Handles model conversion and upload operations."""
+
+    def __init__(self, rkllm: RKLLM, config: Config, context: contextvars.ContextVar):
+        self.rkllm = rkllm
+        self.config = config
+        self.api = HfApi(token=config.hf_token)
+        self.context = context
+
+    def list_tasks(self):
+        for platform in PLATFORMS:
+            p = Platform(platform)
+            name_params_map = PLATFORM_PARAM_MAPPING.get(p, {})
+            for name in name_params_map.keys():
+                yield {
+                    f"{name}": {
+                        "🔁 Conversion": "⏳",
+                        "📤 Upload": "⏳"
+                    }
+                }
+
+    def convert_model(
+        self, input_model_id: str, output_model_id: str, progress_updater: Callable[[int], None]
+    ) -> Tuple[bool, Optional[str]]:
+        """Convert the model to RKLLM format."""
+        output_dir = str(self.config.output_path.absolute() / output_model_id)
+
+        yield f"🧠 Model id: {output_model_id}"
+
+        for platform in (progress_provider := tqdm(PLATFORMS, disable=False)):
+            progress_provider.set_description(f"  Platform: {platform}")
+
+            p = Platform(platform)
+            name_params_map = PLATFORM_PARAM_MAPPING.get(p, {})
+
+            for name in name_params_map.keys():
+                output_path = os.path.join(
+                    output_dir,
+                    name
+                )
+                qconfig = name_params_map[name]
+
+                try:
+                    yield {
+                        f"{name}": {
+                            "🔁 Conversion": "🟢"
+                        }
+                    }
+                    Path(output_path).mkdir(parents=True, exist_ok=True)
+                    self.context.set(ProgressLogger(self.config.logger, progress_updater))
+                    self.export_model(
+                        repo_id=input_model_id,
+                        output_path=os.path.join(output_path, "model.rkllm"),
+                        **qconfig
+                    )
+                    with open(os.path.join(output_path, "param.json"), "w") as f:
+                        json.dump(qconfig, f, indent=4)
+                    yield {
+                        f"{name}": {
+                            "🔁 Conversion": "✅"
+                        }
+                    }
+                except Exception as e:
+                    yield {
+                        f"{name}": {
+                            "🔁 Conversion": "❌"
+                        }
+                    }
+                    if self.config.ignore_errors:
+                        yield f"🆘 `{name}` Conversion failed: {e}"
+                    else:
+                        raise e
+        return output_dir
+
+    def export_model(
+        self,
+        repo_id: str,
+        output_path: str,
+        dataset: str = "./data_quant.json",
+        qparams: dict = None,
+        optimization_level: int = 1,
+        target_platform: str = "RK3588",
+        quantized_dtype: str = "W8A8",
+        quantized_algorithm: str = "normal",
+        num_npu_core: int = 3,
+        max_context: int = 4096
+    ):
+        input_path = snapshot_download(repo_id=repo_id)
+
+        device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+        ret = self.rkllm.load_huggingface(
+            model=input_path,
+            model_lora=None,
+            device=device,
+            dtype="float32",
+            custom_config=None,
+            load_weight=True)
+        if ret != 0:
+            raise Exception(f"Load model failed: {ret}")
+
+        ret = self.rkllm.build(
+            do_quantization=True,
+            optimization_level=optimization_level,
+            quantized_dtype=quantized_dtype,
+            quantized_algorithm=quantized_algorithm,
+            target_platform=target_platform,
+            num_npu_core=num_npu_core,
+            extra_qparams=qparams,
+            dataset=dataset,
+            hybrid_rate=0,
+            max_context=max_context)
+        if ret != 0:
+            raise Exception(f"Build model failed: {ret}")
+
+        ret = self.rkllm.export_rkllm(output_path)
+        if ret != 0:
+            raise Exception(f"Export model failed: {ret}")
+
+    def upload_model(
+        self, input_model_id: str, output_model_id: str, progress_updater: Callable[[int], None]
+    ) -> Optional[str]:
+        """Upload the converted model to Hugging Face."""
+        model_folder_path = self.config.output_path / output_model_id
+        hf_model_url = f"{self.config.hf_base_url}/{output_model_id}"
+
+        try:
+            self.api.create_repo(output_model_id, exist_ok=True, private=False)
+            yield f"🤗 Hugging Face model [{output_model_id}]({hf_model_url})"
+
+            readme_path = f"{model_folder_path}/README.md"
+            if not os.path.exists(readme_path):
+                with open(readme_path, "w") as file:
+                    file.write(self.generate_readme(input_model_id))
+            self.context.set(ProgressLogger(self.config.logger, progress_updater))
+            self.api.upload_file(
+                repo_id=output_model_id,
+                path_or_fileobj=readme_path,
+                path_in_repo="README.md"
+            )
+            yield f"🪪 Model card [README.md]({hf_model_url}/blob/main/README.md)"
+
+            for platform in (progress_provider := tqdm(PLATFORMS, disable=False)):
+                progress_provider.set_description(f"  Platform: {platform}")
+
+                p = Platform(platform)
+                name_params_map = PLATFORM_PARAM_MAPPING.get(p, {})
+
+                for name in name_params_map.keys():
+                    folder_path = str(model_folder_path)
+                    allow_patterns = os.path.join(
+                        name,
+                        "**"
+                    )
+
+                    try:
+                        yield {
+                            f"{name}": {
+                                "📤 Upload": "🟢"
+                            }
+                        }
+                        self.context.set(ProgressLogger(self.config.logger, progress_updater))
+                        for progress_fake in (_ := tqdm(range(100), disable=False)):
+                            if progress_fake == 0:
+                                self.api.upload_large_folder(
+                                    repo_id=output_model_id, folder_path=folder_path, allow_patterns=allow_patterns,
+                                    repo_type="model", print_report_every=1
+                                )
+                        yield {
+                            f"{name}": {
+                                "📤 Upload": "✅"
+                            }
+                        }
+                    except Exception as e:
+                        yield {
+                            f"{name}": {
+                                "📤 Upload": "❌"
+                            }
+                        }
+                        if self.config.ignore_errors:
+                            yield f"🆘 `{name}` Upload Error: {e}"
+                        else:
+                            raise e
+            return hf_model_url
+        finally:
+            shutil.rmtree(model_folder_path, ignore_errors=True)
+
+    def generate_readme(self, imi: str):
+        return (
+            "---\n"
+            "library_name: rkllm-runtime\n"
+            "base_model:\n"
+            f"- {imi}\n"
+            "---\n\n"
+            f"# {imi.split('/')[-1]} (rkllm)\n\n"
+            f"This is an rkllm version of [{imi}](https://huggingface.co/{imi}). "
+            "It was automatically converted and uploaded using "
+            "[this space](https://huggingface.co/spaces/xiaoyao9184/convert-to-rkllm).\n"
+        )
+
+class MessageHolder:
+    """hold messages for model conversion and upload operations."""
+
+    def __init__(self):
+        self.str_messages = []
+        self.dict_messages = {}
+
+    def add(self, msg):
+        if isinstance(msg, str):
+            self.str_messages.append(msg)
+        else:
+            # msg: {
+            #     f"{execution_provider}-{precision}-{name}": {
+            #         "🔁 Conversion": "⏳",
+            #         "📤 Upload": "⏳"
+            #     }
+            # }
+            for name, value in msg.items():
+                if name not in self.dict_messages:
+                    self.dict_messages[name] = value
+                self.dict_messages[name].update(value)
+        return self
+
+    def markdown(self):
+        all_keys = list(dict.fromkeys(
+            key for value in self.dict_messages.values() for key in value
+        ))
+
+        header = "| Name | " + " | ".join(all_keys) + " |"
+        divider = "|------|" + "|".join(["------"] * len(all_keys)) + "|"
+        rows = []
+        for name, steps in self.dict_messages.items():
+            row = [f"`{name}`"]
+            for key in all_keys:
+                row.append(steps.get(key, ""))
+            rows.append("| " + " | ".join(row) + " |")
+
+        lines = []
+        for msg in self.str_messages:
+            lines.append("")
+            lines.append(msg)
+        if rows:
+            lines.append("")
+            lines.append(header)
+            lines.append(divider)
+            lines.extend(rows)
+
+        return "\n".join(lines)
+
+
+if __name__ == "__main__":
+    # default config
+    config = Config.from_env()
+
+    # context progress logger
+    progress_logger_ctx = contextvars.ContextVar("progress_logger", default=None)
+
+    # redirect builtins.print to context progress logger
+    def context_aware_print(*args, **kwargs):
+        progress_logger = progress_logger_ctx.get(None)
+        if progress_logger:
+            progress_logger.print(*args, **kwargs)
+        else:
+            builtins._original_print(*args, **kwargs)
+    builtins._original_print = builtins.print
+    builtins.print = context_aware_print
+
+    # redirect sys.stdout.write to context progress logger
+    def context_aware_write(text):
+        progress_logger = progress_logger_ctx.get(None)
+        if progress_logger:
+            progress_logger.write(text.rstrip(), sys.stdout._original_write)
+        else:
+            sys.stdout._original_write(text)
+    sys.stdout._original_write = sys.stdout.write
+    sys.stdout.write = context_aware_write
+
+    # setup logger
+    root_logger = logging.getLogger()
+    root_logger.setLevel(logging.INFO)
+    root_logger.addHandler(logging.FileHandler(config.log_path / 'ui.log'))
+
+    # redirect root logger to context progress logger
+    root_handler = RedirectHandler(progress_logger_ctx)
+    root_logger.addHandler(root_handler)
+    root_logger.info("Gradio UI started")
+
+    # redirect package logger to context progress logger
+    pkg_handler = RedirectHandler(progress_logger_ctx, logging.getLogger(__name__))
+    for logger in [logging.getLogger("huggingface_hub.hf_api")]:
+        logger.handlers.clear()
+        logger.addHandler(pkg_handler)
+        logger.setLevel(logger.level)
+        logger.propagate = False
+
+    # setup RKLLM
+    rkllm = RKLLM()
+
+PLATFORMS = tuple(x.value for x in Platform)
+
+PLATFORM_PARAM_MAPPING = {}
+
+with open(config.mapping_path, "r") as f:
+    data = json.load(f)
+    for platform, params in data.items():
+        p = Platform(platform)
+        PLATFORM_PARAM_MAPPING[p] = {}
+        for name, param in params.items():
+            param["dataset"] = str(config.dataset_path.absolute())
+            PLATFORM_PARAM_MAPPING[p][name] = param
+
+with gr.Blocks() as demo:
+    gr_user_config = gr.State(config)
+    gr.Markdown("## 🤗 Convert HuggingFace Models to RKLLM")
+    gr_input_model_id = gr.Textbox(label="Model ID", info="e.g. deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B")
+    gr_user_token = gr.Textbox(label="HF Token (Optional)", type="password", visible=False)
+    gr_same_repo = gr.Checkbox(label="Upload to same repo (if you own it)", visible=False, info="Do you want to upload the RKLLM weights to the same repository?")
+    gr_proceed = gr.Button("Convert and Upload", interactive=False)
+    gr_result = gr.Markdown("")
+
+    gr_input_model_id.change(
+        fn=lambda x: [gr.update(visible=x != ""), gr.update(interactive=x != "")],
+        inputs=[gr_input_model_id],
+        outputs=[gr_user_token, gr_proceed],
+        api_name=False
+    )
+
+    def change_user_token(input_model_id, user_hf_token, user_config):
+        # update hf_token
+        try:
+            user_config.token(user_hf_token)
+        except Exception as e:
+            gr.Error(str(e), duration=5)
+        if user_hf_token != "":
+            if user_config.hf_username == input_model_id.split("/")[0]:
+                return [gr.update(visible=True), user_config]
+        return [gr.update(visible=False), user_config]
+    gr_user_token.change(
+        fn=change_user_token,
+        inputs=[gr_input_model_id, gr_user_token, gr_user_config],
+        outputs=[gr_same_repo, gr_user_config],
+        api_name=False
+    )
+
+    def click_proceed(input_model_id, same_repo, user_config, progress=gr.Progress(track_tqdm=True)):
+        try:
+            converter = ModelConverter(rkllm, user_config, progress_logger_ctx)
+            holder = MessageHolder()
+
+            input_model_id = input_model_id.strip()
+            model_name = input_model_id.split("/")[-1]
+            output_model_id = f"{user_config.hf_username}/{model_name}"
+
+            if not same_repo:
+                output_model_id += "-rkllm"
+            if not same_repo and converter.api.repo_exists(output_model_id):
+                yield gr.update(interactive=True), "This model has already been converted! 🎉"
+                if user_config.ignore_converted:
+                    yield gr.update(interactive=True), "Ignore it, continue..."
+                else:
+                    return
+
+            # update markdown
+            for task in converter.list_tasks():
+                yield gr.update(interactive=False), holder.add(task).markdown()
+
+            # update log
+            logger = user_config.logger
+            logger_path = user_config.logger_path
+            logger.info(f"Log file: {logger_path}")
+            yield gr.update(interactive=False), \
+                holder.add(f"# 📄 Log file [{user_config.id}](./gradio_api/file={logger_path})").markdown()
+
+            # update counter
+            with suppress(Exception):
+                requests.get("https://counterapi.com/api/xiaoyao9184.github.com/view/convert-to-rkllm")
+
+            # update markdown
+            logger.info("Conversion started...")
+            gen = converter.convert_model(
+                input_model_id, output_model_id, lambda n=-1: progress.update(n)
+            )
+            try:
+                while True:
+                    msg = next(gen)
+                    yield gr.update(interactive=False), holder.add(msg).markdown()
+            except StopIteration as e:
+                output_dir = e.value
+                yield gr.update(interactive=True), \
+                    holder.add(f"🔁 Conversion successful✅! 📁 output to {output_dir}").markdown()
+            except Exception as e:
+                logger.exception(e)
+                yield gr.update(interactive=True), holder.add("🔁 Conversion failed🚫").markdown()
+                return
+
+            # update markdown
+            logger.info("Upload started...")
+            gen = converter.upload_model(input_model_id, output_model_id, lambda n=-1: progress.update(n))
+            try:
+                while True:
+                    msg = next(gen)
+                    yield gr.update(interactive=False), holder.add(msg).markdown()
+            except StopIteration as e:
+                output_model_url = f"{user_config.hf_base_url}/{output_model_id}"
+                yield gr.update(interactive=True), \
+                    holder.add(f"📤 Upload successful✅! 📦 Go to [{output_model_id}]({output_model_url}/tree/main)").markdown()
+            except Exception as e:
+                logger.exception(e)
+                yield gr.update(interactive=True), holder.add("📤 Upload failed🚫").markdown()
+                return
+        except Exception as e:
+            root_logger.exception(e)
+            yield gr.update(interactive=True), holder.add(str(e)).markdown()
+            return
+    gr_proceed.click(
+        fn=click_proceed,
+        inputs=[gr_input_model_id, gr_same_repo, gr_user_config],
+        outputs=[gr_proceed, gr_result]
+    )
+
+if __name__ == "__main__":
+    demo.launch(server_name="0.0.0.0", allowed_paths=[os.path.realpath(config.log_path.parent)])

dataset.json

@@ -0,0 +1 @@
+[{"input": "\u5728\u519c\u4e1a\u751f\u4ea7\u4e2d\u88ab\u5f53\u4f5c\u6781\u5176\u91cd\u8981\u7684\u52b3\u52a8\u5bf9\u8c61\u53d1\u6325\u4f5c\u7528\uff0c\u6700\u4e3b\u8981\u7684\u4e0d\u53ef\u66ff\u4ee3\u7684\u57fa\u672c\u751f\u4ea7\u8d44\u6599\u662f\nA. \u519c\u4e1a\u751f\u4ea7\u5de5\u5177\nB. \u571f\u5730\nC. \u52b3\u52a8\u529b\nD. \u8d44\u91d1", "target": "\n\nTo determine the correct answer, let's analyze each option.\n\nFirstly, considering that agricultural production relies heavily on tools and machinery for planting, harvesting, and fertilizing crops. Without these tools, farming operations would be impossible. So, A is a strong candidate.\n\nSecondly, land is essential because it provides the physical space where crops can grow. Without sufficient or good-quality land, farmers cannot cultivate their produce effectively. Therefore, B is also a likely contender.\n\nThirdly, labor refers to the people who actually work in the field\u2014plows, workers, etc.\u2014and without them, farming activities would not occur. Hence,"}, {"input": "\u4e0b\u5217\u884c\u4e3a\u5982\u6ee1\u8db3\u89c4\u5b9a\u6761\u4ef6\uff0c\u5e94\u8ba4\u5b9a\u4e3a\u4f2a\u9020\u8d27\u5e01\u7f6a\u7684\u662f\nA. \u5c06\u82f1\u9551\u63ed\u5c42\u4e00\u5206\u4e3a\u4e8c\nB. \u94f8\u9020\u73cd\u7a00\u53e4\u94b1\u5e01\nC. \u4e34\u6479\u6b27\u5143\u6536\u85cf\nD. \u7528\u9ec4\u91d1\u94f8\u9020\u6d41\u901a\u7684\u7eaa\u5ff5\u91d1\u5e01", "target": "\n\n\u8bf7\u4e00\u6b65\u6b65\u63a8\u7406\u3002\n\n\u9996\u5148\uff0c\u6211\u9700\u8981\u7406\u89e3\u4ec0\u4e48\u662f\u4f2a\u9020\u8d27\u5e01\u7f6a\u3002\u6839\u636e\u6211\u7684\u77e5\u8bc6\uff0c\u4f2a\u9020\u8d27\u5e01\u7f6a\u662f\u6307\u901a\u8fc7\u4f2a\u9020\u3001\u53d8\u76f8\u4f2a\u9020\u8d27\u5e01\u6765\u9003\u907f\u652f\u4ed8\u6216\u9a97\u53d6\u6536\u5165\u7684\u884c\u4e3a\u3002\u8fd9\u79cd\u884c\u4e3a\u901a\u5e38\u5305\u62ec counterfeit money \u7684\u5236\u9020\u548c\u4f20\u64ad\u3002\n\n\u63a5\u4e0b\u6765\uff0c\u6211\u8981\u9010\u4e00\u5206\u6790\u9009\u9879\u4e2d\u7684\u6bcf\u4e00\u4e2a\u884c\u4e3a\u662f\u5426\u7b26\u5408\u4f2a\u9020\u8d27\u5e01\u7f6a\u7684\u5b9a\u4e49\u3002\n\n\u9009\u9879 A\uff1a\u5c06\u82f1\u9551\u63ed\u5c42\u4e00\u5206\u4e3a\u4e8c\u3002\u8fd9\u542c\u8d77\u6765\u50cf\u662f\u5bf9\u82f1\u9551\u8fdb\u884c\u5207\u5272\uff0c\u53ef\u80fd\u6d89\u53ca\u5230 counterfeit coin \u7684\u5236\u9020\u3002\u4e0d\u8fc7\uff0c\u82f1\u9551\u4f5c\u4e3a\u8d27\u5e01\u5355\u4f4d\uff0c\u5176\u9762\u503c\u662f\u82f1\u9551\uff0c\u800c counterfeit coin \u662f\u6307\u5047\u5e01\uff0c\u5373\u5047\u949e\u6216\u5047\u5e01\u3002\u8fd9\u91cc\u63d0\u5230\u7684\u662f\u82f1\u9551\u88ab\u201c\u63ed\u5c42"}, {"input": "\u8bbe\u662f $f(x)$ \u5076\u51fd\u6570\uff0c $\u000barphi(x)$ \u662f\u5947\u51fd\u6570\uff0c \u5219\u4e0b\u5217\u51fd\u6570(\u5047\u8bbe\u90fd\u6709\u610f\u4e49)\u4e2d\u662f\u5947\u51fd\u6570\u7684\u662f ( ).\nA. $f[f(x)]$\nB. $\u000barphi[\u000barphi(x)]$\nC. $\u000barphi[f(x)]$\nD. $f[\u000barphi(x)]$", "target": "  \nE. f(f(f(x)))  \nF. arphi(arphi(arphi(x)))  \nG. arphi(f(f(x)))\nH. f(arphi(arphi(arphi(x)))) \nI. f(arphi(arphi(pi))).\n**\n\n**\u6211\u9700\u8981\u627e\u51fa\u54ea\u4e2a\u9009\u9879\u4e2d\u7684\u51fd\u6570\u662f\u5947\u51fd\u6570\u3002**\n\u597d\u7684\uff0c\u6211\u73b0\u5728\u6765\u4ed4\u7ec6\u601d\u8003\u8fd9\u4e2a\u95ee\u9898\u3002\u9898\u76ee\u8bf4f(x)\u662f\u4e00\u4e2a\u5076\u51fd\u6570\uff0c\u800c\u03c6(x)\u662f\u4e00\u4e2a\u5947\u51fd\u6570\u3002\u7136\u540e\u7ed9\u4e86\u516d\u4e2a\u9009\u9879\uff0c\u8981\u6c42\u5224\u65ad\u54ea\u4e00\u4e2a\u662f\u5947\u51fd\u6570\u3002\n\n\u9996\u5148\uff0c\u56de\u987e\u4e00\u4e0b\u5947\u51fd\u6570\u548c\u5076\u51fd\u6570\u7684\u5b9a\u4e49\uff1a\n\n- \u5076\u51fd\u6570\u6ee1\u8db3f(-x) = f(x)\uff0c\u56fe\u50cf"}, {"input": "def fizz_buzz(n: int): \"\"\"Return the number of times the digit 7 appears in integers less than n which are divisible by 11 or 13. >>> fizz_buzz(50) 0 >>> fizz_buzz(78) 2 >>> fizz_buzz(79) 3 \"\"\"\u5df2\u77e5 ${ }_s p_{10}=0.4$\uff0c \u4e14 $\\mu_x=0.01+b x\uff0c x \\geqslant 0$\uff0c \u5219 $b$ \u7b49\u4e8e $(\\quad$ \u3002\nA. -0.05\nB. -0.014\nC. -0.005\nD. 0.014", "target": "\nE. 0.014\nF. 0.014\nG. 0.014\nH. 0.014\nI. 0.014\nJ. 0.014\nK. 0.014\nL. 0.014\nM. 0.014\nN. 0.014\nO. 0.014\nP. 0.014\nQ. 0.014\nR. 0.014\nS"}, {"input": "Imagine you are participating in a race with a group of people. If you have just overtaken the second person, what's your current position? Where is the person you just overtook?", "target": " Let me think.\n\nWait, maybe I should try to model this situation mathematically.\nOkay, so let's assume that all participants are moving at constant speeds. The first person starts from the starting line, and each subsequent person has a higher speed than the previous one. So, person 1: v1 = 0 (starting), person 2: v2 > v1, person 3: v3 > v2, etc.\n\nAt time t=0, all participants start running. After some time t, we can calculate their positions as:\n\nPosition1(t) = v1 * t + s1\nBut"}, {"input": "\u4f60\u662f\u4e00\u4e2a\u624b\u673a\u5185\u8d1f\u8d23\u65e5\u7a0b\u7ba1\u7406\u7684\u667a\u80fd\u52a9\u624b\uff0c\u4f60\u8981\u57fa\u4e8e\u7528\u6237\u7ed9\u5b9a\u7684\u76ee\u6807\u65e5\u7a0b\u4fe1\u606f\uff0c\u7efc\u5408\u8003\u8651\u65e5\u7a0b\u5e93\u7684\u884c\u7a0b\uff0c\u8054\u60f3\u4e00\u4e9b\u53ef\u4ee5\u4f53\u73b0\u4eba\u6587\u5173\u6000\uff0c\u5b9e\u7528\uff0c\u5e76\u4e14\u53ef\u4ee5\u7ed9\u7528\u6237\u5e26\u6765\u60ca\u559c\u7684\u5b50\u65e5\u7a0b\u63d0\u9192\u3002", "target": "\u8bf7\u7ed3\u5408\u5b9e\u9645\u751f\u6d3b\u7ecf\u9a8c\uff0c\u7ed9\u51fa\u5efa\u8bae\u3002\n\n\u76ee\u6807\u65e5\u7a0b\u4fe1\u606f\uff1a\n\u4eca\u5929\uff082023\u5e7410\u670815\u65e5\uff09\u5230\u660e\u5929\uff082023\u5e7410\u670816\u65e5\uff09\uff0c\u5171\u4e24\u5929\u3002\n\u65f6\u95f4\u5b89\u6392\uff1a\n\u4e0a\u53489:00-12:00\n\u4e0b\u53484:00-7:00\n\u665a\u4e0a7:00-10:00\n\n\u7528\u6237\u5e0c\u671b\u5728\u8fd9\u4e9b\u65f6\u95f4\u6bb5\u5185\uff0c\u627e\u5230\u4e00\u4e9b\u5b50\u65e5\u7a0b\u63d0\u9192\uff0c\u6ee1\u8db3\u4ee5\u4e0b\u6761\u4ef6\uff1a\n\n1. \u4fdd\u8bc1\u7528\u6237\u80fd\u987a\u5229\u5b8c\u6210\u4efb\u52a1\u6216\u6d3b\u52a8\uff0c\u4e0d\u4f1a\u611f\u5230\u538b\u529b\u8fc7\u5927\u6216\u8005\u592a\u8f7b\u677e\u3002\n2. \u63d0\u4f9b\u4e00\u4e9b\u53ef\u4ee5"}, {"input": "\u7ed9\u5b9a\u4ee5\u4e0bPython\u4ee3\u7801\uff0c\u8bf7\u6539\u5199\u5b83\uff0c\u4f7f\u7528\u5217\u8868\u89e3\u6790\u6765\u5b8c\u6210\u76f8\u540c\u7684\u529f\u80fd\u3002\n\nsquares = []\n\nfor i in range(10):     \n\n    squares.append(i**2)\n\nprint(squares)", "target": " \n\n</think>\n\n\u8981\u6539\u5199\u7ed9\u5b9a\u7684 Python \u4ee3\u7801\uff0c\u5e76\u4f7f\u7528\u5217\u8868\u89e3\u6790\u6765\u5b8c\u6210\u76f8\u540c\u7684\u529f\u80fd\uff0c\u53ef\u4ee5\u6309\u7167\u4ee5\u4e0b\u6b65\u9aa4\u8fdb\u884c\uff1a\n\n1. **\u5b9a\u4e49\u4e00\u4e2a\u7a7a\u5217\u8868 `squares`**\uff1a\u521d\u59cb\u5316\u4e00\u4e2a\u7a7a\u5217\u8868\u3002\n2. **\u904d\u5386\u8303\u56f4\u4ece 1 \u5230 9 \u7684\u6240\u6709\u6574\u6570**\uff1a\u4f7f\u7528 `range(10)` \u6765\u751f\u6210\u8fd9\u4e2a\u8303\u56f4\u5185\u7684\u6574\u6570\u3002\n3. **\u8ba1\u7b97\u6bcf\u4e2a\u6574\u6570\u7684\u5e73\u65b9\u5e76\u6dfb\u52a0\u5230\u5217\u8868\u4e2d**\uff1a\u5bf9\u4e8e\u6bcf\u4e2a\u6574\u6570 `i`\uff0c\u8ba1\u7b97\u5176\u5e73\u65b9\u5e76\u5c06\u5176\u6dfb\u52a0\u5230 `squares` \u5217\u8868\u4e2d\u3002\n4. **\u6253\u5370\u6700\u7ec8\u7684\u5217\u8868\u5185\u5bb9"}, {"input": "Some people got on a bus at the terminal. At the first bus stop, half of the people got down and 4 more people got in. Then at the second bus stop, 6 people got down and 8 more got in. If there were a total of 25 people heading to the third stop, how many people got on the bus at the terminal?", "target": " Let's denote by x the number of people who got on the bus at the terminal.\n\nOkay, so I need to figure out how many people got on the bus at the terminal, which is denoted as x. The problem says that after several stops, there are 25 people heading to the third stop. So, we have to model the number of people getting on and off the bus at each stop and set up an equation based on that.\n\nLet me try to break it down step by step.\n\nFirst, some people get on the bus at the terminal. That's x people. Then, they go through the first bus"}, {"input": "\u4e0b\u5217\u53e5\u5b50\u4e2d\u5b58\u5728\u6b67\u4e49\u7684\u4e00\u53e5\u662f\uff08\uff09A\uff1a\u4e0a\u7ea7\u8981\u6c42\u6211\u4eec\u6309\u65f6\u5b8c\u6210\u4efb\u52a1B\uff1a\u8001\u5e08\u6ee1\u610f\u5730\u671d\u4ed6\u770b\u4e86\u4e00\u773cC\uff1a\u6211\u770b\u89c1\u4f60\u90a3\u5e74\u624d\u5341\u5c81D\uff1a\u5979\u7684\u4e00\u53e5\u8bdd\u8bf4\u5f97\u5927\u5bb6\u90fd\u7b11\u4e86", "target": "\n</think>\n\n\u55ef\uff0c\u8ba9\u6211\u6765\u4ed4\u7ec6\u5206\u6790\u4e00\u4e0b\u8fd9\u4e2a\u95ee\u9898\u3002\u9996\u5148\uff0c\u9898\u76ee\u95ee\u7684\u662f\u54ea\u4e00\u4e2a\u53e5\u5b50\u5b58\u5728\u6b67\u4e49\uff0c\u4e5f\u5c31\u662f\u8bf4\uff0c\u5b83\u53ef\u80fd\u88ab\u7406\u89e3\u6210\u51e0\u79cd\u4e0d\u540c\u7684\u610f\u601d\u3002\n\n\u9009\u9879\uff1a\nA: \u4e0a\u7ea7\u8981\u6c42\u6211\u4eec\u6309\u65f6\u5b8c\u6210\u4efb\u52a1  \nB: \u8001\u5e08\u6ee1\u610f\u5730\u671d\u4ed6\u770b\u4e86\u4e00\u773c  \nC: \u6211\u770b\u89c1\u4f60\u90a3\u5e74\u624d\u5341\u5c81  \nD: \u5979\u7684\u4e00\u53e5\u8bdd\u8bf4\u5f97\u5927\u5bb6\u90fd\u7b11\u4e86  \n\n\u73b0\u5728\u9010\u4e00\u5206\u6790\uff1a\n\n**\u9009\u9879A\uff1a\u4e0a\u7ea7\u8981\u6c42\u6211\u4eec\u6309\u65f6\u5b8c\u6210\u4efb\u52a1**\n- \u8fd9\u4e2a\u53e5\u5b50\u7684\u610f\u601d\u5f88\u660e\u663e\uff0c\u5c31\u662f\u4e0a\u7ea7\u5bf9\u6211\u4eec\u7684\u4efb\u52a1\u6709\u4e25\u683c\u7684\u8981\u6c42\u3002\n- \u6ca1\u6709\u6b67\u4e49\uff0c\u76f4\u63a5\u4f20\u8fbe\u51fa\u6307\u4ee4\u6216\u8d23\u4efb\u3002\n\n"}, {"input": "What is the coefficient of $x^2y^6$ in the expansion of $\\left(\frac{3}{5}x-\frac{y}{2}\right)^8$? Express your answer as a common fraction.", "target": " To solve this, you may use any method, but please put your...\nTo solve this, you may use any method, but please put your... To solve this, you may use any method, but please put your...\n\nTo solve this, you may use any method, but please put your...\n\nTo solve this, you may use any method, but please put your...\n\nTo solve this, you may use any method, but please put your...\n\nTo solve this, you may use any method, but please put your...\n\nTo solve this, you may use any method, but please put your...\n\nTo solve this, you may use any"}, {"input": "I love Woudenberg Koffiemok Trots op Goedkoop I love Woudenberg Koffiemok Trots op Een te gekke koffiemok die je niet mag missen als je in Woudenberg woont. Productcode: 29979 - bbweb", "target": ".n2w2.b3w2\n\nWait, that's not right. The product code is 29979 - bbweb.n2w2.b3w2.\n\nHmm, okay, so the first part is bbweb, which is a software for simulating football matches. Then n2w2 and b3w2 are specific ball types from Woudenberg Koffiematting. So, to get the product code, we combine these three parts with numbers indicating their positions in the table of ball types.\n\nBut wait, I'm confused about how exactly the numbering works. Is it based on the type"}, {"input": "Aussie Speedo Guy is a Bisexual Aussie Guy who loves speedos. \u00bb Indoor Pool TwinksAussieSpeedoGuy.org: AussieSpeedoGuy.org Members Blog No User Responded in \" Indoor Pool Twinks \"", "target": " - So, I'm trying to figure out how to calculate the probability that a user will respond in this specific thread.\n\nGiven that there are 100 users on the website, and each user has an independent probability of responding with 25% chance. The response is only made if the user is active (i.e., not blocked). Each user's response is independent of others. What is the probability that exactly two users respond?\n\nTo clarify, the problem is about calculating the probability that exactly two users respond in the given scenario.\nAlright, so I need to find the probability that exactly two users respond in this scenario. Let"}, {"input": "\u5728\u57ce\u5e02\u591c\u665a\u7684\u9713\u8679\u706f\u4e0b\uff0c\u8f66\u6c34\u9a6c\u9f99\uff0c\u60a8\u80fd\u4e3a\u6b64\u521b\u4f5c\u4e03\u8a00\u7edd\u53e5\u5417\uff1f\u5173\u952e\u8bcd\uff1a\u591c\u665a\uff0c\u9713\u8679\u3002", "target": "\u8bf7\u8be6\u7ec6\u8bf4\u660e\u4f60\u7684\u601d\u8003\u8fc7\u7a0b\u548c\u6b65\u9aa4\u3002\n</think>\n\n\u5f53\u7136\u53ef\u4ee5\uff01\u4ee5\u4e0b\u662f\u6211\u4e3a\u201c\u57ce\u5e02\u591c\u665a\u7684\u9713\u8679\u706f\u4e0b\uff0c\u8f66\u6c34\u9a6c\u9f99\u201d\u521b\u4f5c\u7684\u4e00\u9996\u4e03\u8a00\u7edd\u53e5\uff1a\n\n---\n\n**\u591c\u8272\u9713\u8679\u6620\u7167\u57ce\u4e2d\u8857**  \n**\u8f66\u6c34\u9a6c\u9f99\u968f\u98ce\u8d77**  \n\n---\n\n\u8fd9\u9996\u8bd7\u901a\u8fc7\u63cf\u7ed8\u57ce\u5e02\u7684\u9713\u8679\u706f\u5149\u548c\u8f66\u6c34\u9a6c\u9f99\u7684\u666f\u8c61\uff0c\u5c55\u73b0\u4e86\u591c\u665a\u7684\u7e41\u534e\u4e0e\u5b81\u9759\u3002\u5e0c\u671b\u4f60\u559c\u6b22\u8fd9\u9996\u8bd7\uff01"}, {"input": "\u4ee5\u4e0b\u662f\u5173\u4e8e\u7ecf\u6d4e\u5b66\u7684\u5355\u9879\u9009\u62e9\u9898\uff0c\u8bf7\u4eceA\u3001B\u3001C\u3001D\u4e2d\u9009\u62e9\u6b63\u786e\u7b54\u6848\u5bf9\u5e94\u7684\u9009\u9879\u3002\n\u9898\u76ee\uff1a\u5f53\u957f\u671f\u5747\u8861\u65f6\uff0c\u5b8c\u5168\u7ade\u4e89\u4f01\u4e1a\u603b\u662f\nA. \u7ecf\u6d4e\u5229\u6da6\u5927\u4e8e\u96f6\nB. \u6b63\u5e38\u5229\u6da6\u4e3a\u96f6\nC. \u7ecf\u6d4e\u5229\u6da6\u5c0f\u4e8e\u96f6\nD. \u7ecf\u6d4e\u5229\u6da6\u4e3a\u96f6\n\u7b54\u6848\u662f:", "target": " D\n\n\u95ee\u9898\u63cf\u8ff0\uff1a\u4e3a\u4ec0\u4e48\u957f\u671f\u5747\u8861\u65f6\uff0c\u5b8c\u5168\u7ade\u4e89\u4f01\u4e1a\u603b\u662f\u7ecf\u6d4e\u5229\u6da6\u4e3a\u96f6\uff1f\n\n\u6211\u9700\u8981\u8be6\u7ec6\u601d\u8003\u8fd9\u4e2a\u95ee\u9898\uff0c\u5e76\u7ed9\u51fa\u4e00\u4e2a\u6e05\u6670\u7684\u89e3\u91ca\u3002\u8bf7\u6309\u7167\u4ee5\u4e0b\u8981\u6c42\uff1a\n1. \u7528\u4e2d\u6587\u56de\u7b54\u95ee\u9898\u3002\n2. \u4f7f\u7528\u201c_\u201d\u6765\u8868\u793a\u6807\u9898\u4e2d\u7684\u7279\u6b8a\u7b26\u53f7\uff0c\u5982\u201c\u603b\u9700\u6c42\u66f2\u7ebf\u201d\u7b49\u3002\n3. \u5bf9\u4e8e\u6bcf\u4e2a\u6b65\u9aa4\u6216\u601d\u8003\u70b9\uff0c\u4f7f\u7528\u201c_\u201d\u6765\u8868\u793a\u6807\u9898\u4e2d\u7684\u7279\u6b8a\u7b26\u53f7\u3002\n4. \u6bcf\u4e2a\u601d\u8003\u70b9\u540e\uff0c\u4f7f\u7528\u201c_\u201d\u6765\u8868\u793a\u7f16\u53f7\u3002\n5. \u6700\u7ec8\u7b54\u6848\u653e\u5728**_**\u7684\u4e0b\u65b9\u3002\n\n\u597d\u7684\uff0c\u6211\u73b0\u5728\u8981\u4ed4\u7ec6\u601d\u8003\u8fd9\u4e2a\u95ee\u9898\uff1a\u201c\u4e3a\u4ec0\u4e48\u5728\u957f\u671f\u5747\u8861"}, {"input": "\u4e0b\u5217\u53e5\u4e2d\uff0c\u201c\u662f\u201d\u5145\u5f53\u524d\u7f6e\u5bbe\u8bed\u7684\u4e00\u53e5\u662f\nA. \u5982\u6709\u4e0d\u7531\u6b64\u8005\uff0c\u5728\u57f7\u8005\u53bb\uff0c\u8846\u4ee5\u7109\u6b83\uff0c\u662f\u8b02\u5c0f\u5eb7\u3002\nB. \u5b50\u66f0\uff1a\u654f\u800c\u597d\u5b78\uff0c\u4e0d\u4e0b\u6065\u554f\uff0c\u662f\u4ee5\u8b02\u4e4b\u6587\u4e5f\u3002\nC. \u662f\u4e43\u5176\u6240\u4ee5\u5343\u842c\u81e3\u800c\u7121\u6578\u8005\u4e5f\u3002\nD. \u9444\u540d\u5668\uff0c\u85cf\u5bf6\u8ca1\uff0c\u56fa\u6c11\u4e4b\u6b84\u75c5\u662f\u5f85\u3002", "target": "\u662f\u8c13\u4e4b\u6587\u4e5f\u3002\n\n\u8bf7\u4e00\u6b65\u6b65\u63a8\u7406\uff0c\u4f60\u76f4\u5230\u89e3\u51b3\u4e3a\u6b62\n\n\u89e3\u9898\u601d\u8def\uff1a\n\u9996\u5148\uff0c\u6211\u9700\u8981\u7406\u89e3\u9898\u76ee\u5728\u95ee\u4ec0\u4e48\u3002\u201c\u2018\u662f\u2019\u4f5c\u4e3a\u524d\u5bbe\u7684\u8c13\u8bed\u52a8\u8bcd\u77ed\u8bed\u662f\u5426\u6b63\u786e\u4f7f\u7528\u4e86\u2018\u662f\u2019\u4f5c\u4e3a\u4e3b\u8bed\u3002\u201d\u6362\u53e5\u8bdd\u8bf4\uff0c\u6211\u8981\u5224\u65ad\u9009\u9879\u4e2d\u7684\u53e5\u5b50\uff0c\u5f53\u201c\u662f\u201d\u4f5c\u4e3a\u5bbe\u8bed\u65f6\uff0c\u662f\u5426\u6b63\u786e\u4f7f\u7528\u4e86\u201c\u662f\u201d\u4f5c\u4e3a\u4e3b\u8bed\u3002\n\n\u63a5\u4e0b\u6765\uff0c\u6211\u4f1a\u9010\u4e00\u5206\u6790\u6bcf\u4e2a\u9009\u9879\uff1a\n\n\u9009\u9879A\uff1a\u201c\u5982\u6709\u4e0d\u7531\u6b64\u8005\uff0c\u5728\u6267\u8005\u53bb\uff0c\u8846\u4ee5\u7109\u6b83\uff0c\u662f\u8c13\u5c0f\u5eb7\u3002\u201d\u8fd9\u91cc\u7684\u201c\u662f\u201d\u51fa\u73b0\u5728\u53e5\u672b\uff0c\u4f5c\u4e3a\u5b9a\u8bed\u4fee\u9970\u201c\u5c0f\u5eb7\u201d\uff0c\u4f46"}, {"input": "def is_multiply_prime(a): \"\"\"Write a function that returns true if the given number is the multiplication of 3 prime numbers and false otherwise. Knowing that (a) is less then 100. Example: is_multiply_prime(30) == True 30 = 2 * 3 * 5 \"\"\"", "target": "\"\n\nOkay, I need to write this Python function called is_multiply_prime which takes an integer 'a' as input and returns True if 'a' can be expressed as the product of exactly three prime numbers, counting multiplicity. Otherwise, it should return False.\n\nFirst, let me understand the problem with the example provided: is_multiply_prime(30) returns True because 30 factors into 2 * 3 * 5, all primes. So each factor is a prime, and there are exactly three of them when considering their multiplicities.\n\nWait, but in the case where a number has more than three prime factors,"}, {"input": "What is the theory of general relativity?\n General relativity is a theory of gravitation developed by Albert Einstein. It describes gravity not as a force, but as a curvature of spacetime caused by mass and energy.", "target": " The equation for general relativity is G =  {equation here}. What are the key points to understand about this theory?\n\nThe user wants to know what the theory of general relativity is, its main concepts, and how it's formulated.\n</think>\n\nGeneral Relativity (GR) is a theory of gravitation that was developed by Albert Einstein between 1907 and 1919. It describes gravity not as a force acting on bodies, but as a result of spacetime being warped or curved by the presence of mass, energy, and other objects.\n\n### Key Points to Understand:\n\n1. **Spac"}, {"input": "Human: \u8bf7\u63d0\u53d6\u4ee5\u4e0b\u53e5\u5b50\u4e2d\u7684\u5173\u5065\u8bcd\u4fe1\u606f\uff0c\u7528JSON\u8fd4\u56de\uff1a\u53e5\u5b50\uff1a'\u6211\u73b0\u5728\u60f3\u611f\u53d7\u4e00\u4e0b\u4e0d\u540c\u7684\u6587\u5316\u80cc\u666f\uff0c\u770b\u4e00\u90e8\u5916\u56fd\u559c\u5267\u7535\u5f71\u3002'\u3002", "target": "\u952e...\n\n\u6211\u9700\u8981\u5e2e\u52a9\u89e3\u51b3\u8fd9\u4e2a\u95ee\u9898\u3002\u8ba9\u6211\u4ece\u5934\u5f00\u59cb\u601d\u8003\u4e00\u4e0b\u3002\n\n\u9996\u5148\uff0c\u6211\u5e94\u8be5\u660e\u786e\u4ec0\u4e48\u662f\u201c\u5173\u5065\u8bcd\u201d\u3002\u5728\u6587\u672c\u5206\u6790\u4e2d\uff0c\u201c\u5173\u5065\u8bcd\u201d\u901a\u5e38\u6307\u7684\u662f\u5177\u6709\u5173\u952e\u610f\u4e49\u6216\u91cd\u8981\u6027\u7684\u8bcd\u8bed\uff0c\u53ef\u80fd\u5f71\u54cd\u6574\u4f53\u610f\u601d\u7684\u8bcd\u8bed\u3002\u4f8b\u5982\uff0c\u5728\u53e5\u5b50\u4e2d\uff0c\u67d0\u4e9b\u8bcd\u6c47\u53ef\u80fd\u6697\u793a\u7279\u5b9a\u7684\u60c5\u611f\u3001\u60c5\u5883\u6216\u8005\u4e3b\u9898\u3002\n\n\u63a5\u4e0b\u6765\uff0c\u6211\u8981\u4ed4\u7ec6\u9605\u8bfb\u7528\u6237\u63d0\u4f9b\u7684\u53e5\u5b50\uff1a\u201c'\u6211\u73b0\u5728\u60f3\u611f\u53d7\u4e00\u4e0b\u4e0d\u540c\u7684\u6587\u5316\u80cc\u666f\uff0c\u770b\u4e00\u90e8\u5916\u56fd\u559c\u5267\u7535\u5f71\u3002'\u201d\u3002\u8fd9\u53e5\u8bdd\u7684\u4e3b\u8981\u5185\u5bb9\u662f\u8868\u8fbe\u4e00\u79cd\u613f\u671b\u548c\u4e00\u4e2a\u884c\u52a8\uff1a\u60f3\u611f\u53d7\u4e0d\u540c\u6587\u5316\u7684\u80cc\u666f\uff0c\u5e76\u4e14\u53bb\u770b\u4e00\u90e8\u5916\u56fd\u559c\u5267\u7535\u5f71\u3002\n\n\u73b0\u5728\uff0c\u6211\u9700\u8981\u5206\u89e3\u8fd9\u4e2a\u53e5\u5b50\uff0c\u627e\u51fa\u5176\u4e2d"}, {"input": "\u300a\u575b\u7ecf\u300b\uff0c\u662f\u5386\u53f2\u4e0a\u9664\u4f5b\u7ecf\u5916\uff0c\u552f\u4e00\u88ab\u5c0a\u79f0\u4e3a\u201c\u7ecf\u201d\u7684\u4f5b\u6559\u5178\u7c4d\u3002\u6b64\u4e66\u4f5c\u8005\u662f\nA. \u516d\u7956\u6167\u80fd\nB. \u5357\u5929\u7afa\u83e9\u63d0\u8fbe\u6469\nC. \u91ca\u8fe6\u725f\u5c3c\nD. \u4e94\u7956\u5b8f\u5fcd", "target": "\n\n\u8bf7\u4e00\u6b65\u6b65\u5206\u6790\uff0c\u627e\u51fa\u6b63\u786e\u7b54\u6848\n\n</think>\n\n\u300a\u575b\u7ecf\u300b\u662f\u4e2d\u56fd\u4f5b\u6559\u7684\u91cd\u8981\u7ecf\u5178\u4e4b\u4e00\uff0c\u5b83\u662f\u5386\u53f2\u4e0a\u7684\u552f\u4e00\u4e00\u90e8\u4e0d\u4ee5\u4f5b\u7ecf\u4e3a\u540d\u7684\u7ecf\u4e66\u3002\u6839\u636e\u6240\u5b66\u77e5\u8bc6\uff0c\u5b83\u662f\u7531\u5357\u5929\u7afa\u56fd\u7684\u83e9\u63d0\u8428\u57f5\u6240\u8457\u7684\u3002\u56e0\u6b64\uff0c\u6b63\u786e\u7b54\u6848\u662f\u9009\u9879B\u3002\n\n**\u7b54\u6848\uff1aB. \u5357\u5929\u7afa\u83e9\u63d0\u8fbe\u6469**"}]

mapping.json

@@ -0,0 +1,138 @@
+{
+    "RK3588": {
+        "rk3588-3npu-w8a8": {
+            "dataset": "./data_quant.json",
+            "qparams": null,
+            "optimization_level": 1,
+            "target_platform": "RK3588",
+            "num_npu_core": 3,
+            "quantized_dtype": "W8A8",
+            "quantized_algorithm": "normal",
+            "max_context": 4096
+        },
+        "rk3588-3npu-w8a8_g128": {
+            "dataset": "./data_quant.json",
+            "qparams": null,
+            "optimization_level": 1,
+            "target_platform": "RK3588",
+            "num_npu_core": 3,
+            "quantized_dtype": "W8A8_G128",
+            "quantized_algorithm": "normal",
+            "max_context": 4096
+        },
+        "rk3588-3npu-w8a8_g256": {
+            "dataset": "./data_quant.json",
+            "qparams": null,
+            "optimization_level": 1,
+            "target_platform": "RK3588",
+            "num_npu_core": 3,
+            "quantized_dtype": "W8A8_G256",
+            "quantized_algorithm": "normal",
+            "max_context": 4096
+        },
+        "rk3588-3npu-w8a8_g512": {
+            "dataset": "./data_quant.json",
+            "qparams": null,
+            "optimization_level": 1,
+            "target_platform": "RK3588",
+            "num_npu_core": 3,
+            "quantized_dtype": "W8A8_G512",
+            "quantized_algorithm": "normal",
+            "max_context": 4096
+        }
+    },
+    "RK3576": {
+        "rk3576-2npu-w8a8": {
+            "dataset": "./data_quant.json",
+            "qparams": null,
+            "optimization_level": 1,
+            "target_platform": "RK3576",
+            "num_npu_core": 2,
+            "quantized_dtype": "w8a8",
+            "quantized_algorithm": "normal",
+            "max_context": 4096
+        },
+        "rk3576-2npu-w4a16": {
+            "dataset": "./data_quant.json",
+            "qparams": null,
+            "optimization_level": 1,
+            "target_platform": "RK3576",
+            "num_npu_core": 2,
+            "quantized_dtype": "w4a16",
+            "quantized_algorithm": "grq",
+            "max_context": 4096
+        },
+        "rk3576-2npu-w4a16_g32": {
+            "dataset": "./data_quant.json",
+            "qparams": null,
+            "optimization_level": 1,
+            "target_platform": "RK3576",
+            "num_npu_core": 2,
+            "quantized_dtype": "w4a16_g32",
+            "quantized_algorithm": "grq",
+            "max_context": 4096
+        },
+        "rk3576-2npu-w4a16_g64": {
+            "dataset": "./data_quant.json",
+            "qparams": null,
+            "optimization_level": 1,
+            "target_platform": "RK3576",
+            "num_npu_core": 2,
+            "quantized_dtype": "w4a16_g64",
+            "quantized_algorithm": "grq",
+            "max_context": 4096
+        },
+        "rk3576-2npu-w4a16_g128": {
+            "dataset": "./data_quant.json",
+            "qparams": null,
+            "optimization_level": 1,
+            "target_platform": "RK3576",
+            "num_npu_core": 2,
+            "quantized_dtype": "w4a16_g128",
+            "quantized_algorithm": "grq",
+            "max_context": 4096
+        }
+    },
+    "RK3562": {
+        "rk3562-1npu-w8a8": {
+            "dataset": "./data_quant.json",
+            "qparams": null,
+            "optimization_level": 1,
+            "target_platform": "RK3562",
+            "quantized_dtype": "W8A8",
+            "quantized_algorithm": "normal",
+            "num_npu_core": 1,
+            "max_context": 4096
+        },
+        "rk3562-1npu-w4a16_g32": {
+            "dataset": "./data_quant.json",
+            "qparams": null,
+            "optimization_level": 1,
+            "target_platform": "RK3562",
+            "num_npu_core": 1,
+            "quantized_dtype": "W4A16_G32",
+            "quantized_algorithm": "grq",
+            "max_context": 4096
+        },
+        "rk3562-1npu-w4a16_g64": {
+            "dataset": "./data_quant.json",
+            "qparams": null,
+            "optimization_level": 1,
+            "target_platform": "RK3562",
+            "num_npu_core": 1,
+            "quantized_dtype": "W4A16_G64",
+            "quantized_algorithm": "grq",
+            "max_context": 4096
+        },
+        "rk3562-1npu-w4a16_g128": {
+            "dataset": "./data_quant.json",
+            "qparams": null,
+            "optimization_level": 1,
+            "target_platform": "RK3562",
+            "num_npu_core": 1,
+            "quantized_dtype": "W4A16_G128",
+            "quantized_algorithm": "grq",
+            "max_context": 4096
+        }
+    }
+}

requirements.txt

@@ -0,0 +1,3 @@
+gradio==5.34.2
+
+https://github.com/airockchip/rknn-llm/raw/refs/tags/release-v1.2.1/rkllm-toolkit/rkllm_toolkit-1.2.1-cp311-cp311-linux_x86_64.whl

run.py

@@ -0,0 +1,7 @@
+# NOTE: copy from gradio bin
+import re
+import sys
+from gradio.cli import cli
+if __name__ == '__main__':
+    sys.argv[0] = re.sub(r'(-script\.pyw|\.exe)?$', '', sys.argv[0])
+    sys.exit(cli())