{"id":30867,"date":"2025-05-05T19:27:11","date_gmt":"2025-05-05T11:27:11","guid":{"rendered":"https:\/\/www.aisharenet.com\/?p=30867"},"modified":"2025-07-26T09:30:41","modified_gmt":"2025-07-26T01:30:41","slug":"kimi-audio","status":"publish","type":"post","link":"https:\/\/www.kdjingpai.com\/de\/kimi-audio\/","title":{"rendered":"Kimi-Audio\uff1a\u5f00\u6e90\u97f3\u9891\u5904\u7406\u4e0e\u5bf9\u8bdd\u57fa\u7840\u6a21\u578b"},"content":{"rendered":"<p>Kimi-Audio \u662f\u7531 Moonshot AI \u5f00\u53d1\u7684\u4e00\u6b3e\u5f00\u6e90\u97f3\u9891\u57fa\u7840\u6a21\u578b\uff0c\u4e13\u6ce8\u4e8e\u97f3\u9891\u7406\u89e3\u3001\u751f\u6210\u548c\u5bf9\u8bdd\u3002\u5b83\u652f\u6301\u591a\u79cd\u97f3\u9891\u5904\u7406\u4efb\u52a1\uff0c\u4f8b\u5982\u8bed\u97f3\u8bc6\u522b\u3001\u97f3\u9891\u95ee\u7b54\u548c\u8bed\u97f3\u60c5\u611f\u8bc6\u522b\u3002\u6a21\u578b\u7ecf\u8fc7\u8d85\u8fc7 1300 \u4e07\u5c0f\u65f6\u7684\u97f3\u9891\u6570\u636e\u9884\u8bad\u7ec3\uff0c\u7ed3\u5408\u521b\u65b0\u7684\u6df7\u5408\u67b6\u6784\uff0c\u5728\u591a\u4e2a\u97f3\u9891\u57fa\u51c6\u6d4b\u8bd5\u4e2d\u8868\u73b0\u4f18\u5f02\u3002Kimi-Audio \u63d0\u4f9b\u6a21\u578b\u6743\u91cd\u3001\u63a8\u7406\u4ee3\u7801\u548c\u8bc4\u4f30\u5de5\u5177\u5305\uff0c\u65b9\u4fbf\u5f00\u53d1\u8005\u5728\u7814\u7a76\u548c\u5e94\u7528\u4e2d\u96c6\u6210\u3002\u5b83\u9002\u5408\u9700\u8981\u9ad8\u6548\u97f3\u9891\u5904\u7406\u548c\u5bf9\u8bdd\u80fd\u529b\u7684\u573a\u666f\uff0c\u6587\u6863\u6e05\u6670\uff0c\u652f\u6301 Docker \u90e8\u7f72\uff0c\u793e\u533a\u6d3b\u8dc3\uff0c\u6301\u7eed\u66f4\u65b0\u3002<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"aligncenter size-full wp-image-30868\" title=\"Kimi-Audio\uff1a\u5f00\u6e90\u97f3\u9891\u5904\u7406\u4e0e\u5bf9\u8bdd\u57fa\u7840\u6a21\u578b-1\" src=\"https:\/\/www.kdjingpai.com\/wp-content\/uploads\/2025\/05\/f2d9f1271e1d837.png\" alt=\"Kimi-Audio\uff1a\u5f00\u6e90\u97f3\u9891\u5904\u7406\u4e0e\u5bf9\u8bdd\u57fa\u7840\u6a21\u578b-1\" width=\"812\" height=\"611\" srcset=\"https:\/\/www.kdjingpai.com\/wp-content\/uploads\/2025\/05\/f2d9f1271e1d837.png 812w, https:\/\/www.kdjingpai.com\/wp-content\/uploads\/2025\/05\/f2d9f1271e1d837-768x578.png 768w, https:\/\/www.kdjingpai.com\/wp-content\/uploads\/2025\/05\/f2d9f1271e1d837-16x12.png 16w\" sizes=\"auto, (max-width: 812px) 100vw, 812px\" \/><\/p>\n<p>&nbsp;<\/p>\n<h2>\u529f\u80fd\u5217\u8868<\/h2>\n<ul>\n<li>\u8bed\u97f3\u8bc6\u522b\uff08ASR\uff09\uff1a\u5c06\u97f3\u9891\u8f6c\u4e3a\u6587\u672c\uff0c\u652f\u6301\u591a\u8bed\u8a00\u8bed\u97f3\u8f6c\u5199\u3002<\/li>\n<li>\u97f3\u9891\u95ee\u7b54\uff08AQA\uff09\uff1a\u6839\u636e\u97f3\u9891\u5185\u5bb9\u56de\u7b54\u7528\u6237\u63d0\u95ee\uff0c\u7406\u89e3\u97f3\u9891\u4e0a\u4e0b\u6587\u3002<\/li>\n<li>\u97f3\u9891\u5b57\u5e55\uff08AAC\uff09\uff1a\u4e3a\u97f3\u9891\u751f\u6210\u7cbe\u51c6\u7684\u5b57\u5e55\u6216\u63cf\u8ff0\u3002<\/li>\n<li>\u8bed\u97f3\u60c5\u611f\u8bc6\u522b\uff08SER\uff09\uff1a\u5206\u6790\u97f3\u9891\u4e2d\u7684\u60c5\u611f\u72b6\u6001\uff0c\u5982\u9ad8\u5174\u6216\u60b2\u4f24\u3002<\/li>\n<li>\u58f0\u97f3\u4e8b\u4ef6\/\u573a\u666f\u5206\u7c7b\uff08SEC\/ASC\uff09\uff1a\u8bc6\u522b\u97f3\u9891\u4e2d\u7684\u7279\u5b9a\u4e8b\u4ef6\u6216\u573a\u666f\uff0c\u5982\u6c7d\u8f66\u58f0\u6216\u5ba4\u5185\u73af\u5883\u3002<\/li>\n<li>\u6587\u672c\u8f6c\u8bed\u97f3\uff08TTS\uff09\uff1a\u5c06\u6587\u672c\u8f6c\u5316\u4e3a\u81ea\u7136\u8bed\u97f3\uff0c\u652f\u6301\u591a\u79cd\u97f3\u8272\u3002<\/li>\n<li>\u8bed\u97f3\u8f6c\u6362\uff08VC\uff09\uff1a\u6539\u53d8\u8bed\u97f3\u7684\u97f3\u8272\u6216\u98ce\u683c\uff0c\u751f\u6210\u4e2a\u6027\u5316\u97f3\u9891\u3002<\/li>\n<li>\u7aef\u5230\u7aef\u8bed\u97f3\u5bf9\u8bdd\uff1a\u652f\u6301\u8fde\u7eed\u7684\u8bed\u97f3\u4ea4\u4e92\uff0c\u6a21\u62df\u81ea\u7136\u5bf9\u8bdd\u3002<\/li>\n<li>\u6d41\u5f0f\u97f3\u9891\u751f\u6210\uff1a\u901a\u8fc7\u5206\u5757\u6d41\u5f0f\u89e3\u7801\u5668\uff0c\u5b9e\u73b0\u4f4e\u5ef6\u8fdf\u97f3\u9891\u751f\u6210\u3002<\/li>\n<li>\u8bc4\u4f30\u5de5\u5177\u5305\uff1a\u63d0\u4f9b\u6807\u51c6\u5316\u7684\u8bc4\u4f30\u5de5\u5177\uff0c\u65b9\u4fbf\u6bd4\u8f83\u4e0d\u540c\u6a21\u578b\u6027\u80fd\u3002<\/li>\n<\/ul>\n<p>&nbsp;<\/p>\n<h2>\u4f7f\u7528\u5e2e\u52a9<\/h2>\n<h3>\u5b89\u88c5\u6d41\u7a0b<\/h3>\n<p>Kimi-Audio \u7684\u90e8\u7f72\u63a8\u8350\u4f7f\u7528 Docker\uff0c\u786e\u4fdd\u73af\u5883\u4e00\u81f4\u6027\u548c\u5b89\u88c5\u7b80\u4fbf\u3002\u4ee5\u4e0b\u662f\u8be6\u7ec6\u7684\u5b89\u88c5\u6b65\u9aa4\uff1a<\/p>\n<ol>\n<li><strong>\u514b\u9686\u4ed3\u5e93<\/strong><br \/>\n\u5728\u7ec8\u7aef\u8fd0\u884c\u4ee5\u4e0b\u547d\u4ee4\uff0c\u83b7\u53d6 Kimi-Audio \u7684\u4ee3\u7801\u548c\u5b50\u6a21\u5757\uff1a<\/p>\n<pre><code>git clone https:\/\/github.com\/MoonshotAI\/Kimi-Audio.git\r\ncd Kimi-Audio\r\ngit submodule update --init --recursive<\/code><\/pre>\n<\/li>\n<\/ol>\n<ol start=\"2\">\n<li><strong>\u5b89\u88c5\u4f9d\u8d56<\/strong><br \/>\n\u5b89\u88c5 Python \u4f9d\u8d56\uff0c\u63a8\u8350\u4f7f\u7528 Python 3.10 \u73af\u5883\uff1a<\/p>\n<pre><code>pip install -r requirements.txt\r\n<\/code><\/pre>\n<p>\u786e\u4fdd\u5b89\u88c5\u00a0<code>torch<\/code>\u00a0\u548c\u00a0<code>soundfile<\/code>\uff0c\u652f\u6301 GPU \u7684\u7528\u6237\u9700\u5b89\u88c5\u5bf9\u5e94 CUDA \u7248\u672c\u7684 PyTorch\u3002<\/li>\n<li><strong>\u6784\u5efa Docker \u955c\u50cf<\/strong><br \/>\n\u5728 Kimi-Audio \u76ee\u5f55\u4e0b\u6784\u5efa Docker \u955c\u50cf\uff1a<\/p>\n<pre><code>docker build -t kimi-audio:v0.1 .\r\n<\/code><\/pre>\n<p>\u6216\u8005\u4f7f\u7528\u5b98\u65b9\u9884\u6784\u5efa\u955c\u50cf\uff1a<\/p>\n<pre><code>docker pull moonshotai\/kimi-audio:v0.1\r\n<\/code><\/pre>\n<\/li>\n<li><strong>\u8fd0\u884c\u5bb9\u5668<\/strong><br \/>\n\u542f\u52a8 Docker \u5bb9\u5668\uff0c\u5e76\u6302\u8f7d\u672c\u5730\u5de5\u4f5c\u76ee\u5f55\uff1a<\/p>\n<pre><code>docker run -it -v $(pwd):\/app kimi-audio:v0.1 bash\r\n<\/code><\/pre>\n<\/li>\n<li><strong>\u4e0b\u8f7d\u6a21\u578b\u6743\u91cd<\/strong><br \/>\nKimi-Audio \u63d0\u4f9b\u4e24\u4e2a\u4e3b\u8981\u6a21\u578b\uff1a<code>Kimi-Audio-7B<\/code>\uff08\u57fa\u7840\u6a21\u578b\uff09\u548c\u00a0<code>Kimi-Audio-7B-Instruct<\/code>\uff08\u6307\u4ee4\u5fae\u8c03\u6a21\u578b\uff09\u3002\u4ece Hugging Face \u4e0b\u8f7d\uff1a<\/p>\n<ul>\n<li><code>moonshotai\/Kimi-Audio-7B-Instruct<\/code>\uff1a\u9002\u5408\u76f4\u63a5\u4f7f\u7528\u3002<\/li>\n<li><code>moonshotai\/Kimi-Audio-7B<\/code>\uff1a\u9002\u5408\u8fdb\u4e00\u6b65\u5fae\u8c03\u3002<br \/>\n\u4f7f\u7528 Hugging Face CLI \u767b\u5f55\u5e76\u4e0b\u8f7d\uff1a<\/li>\n<\/ul>\n<pre><code>huggingface-cli login\r\n<\/code><\/pre>\n<p>\u6a21\u578b\u4f1a\u81ea\u52a8\u4e0b\u8f7d\u5230\u6307\u5b9a\u8def\u5f84\u3002<\/li>\n<\/ol>\n<h3>\u4f7f\u7528\u65b9\u6cd5<\/h3>\n<p>Kimi-Audio \u7684\u6838\u5fc3\u529f\u80fd\u901a\u8fc7 Python API \u8c03\u7528\u5b9e\u73b0\u3002\u4ee5\u4e0b\u662f\u4e3b\u8981\u529f\u80fd\u7684\u8be6\u7ec6\u64cd\u4f5c\u6d41\u7a0b\uff1a<\/p>\n<h4>1. \u8bed\u97f3\u8bc6\u522b\uff08ASR\uff09<\/h4>\n<p>\u5c06\u97f3\u9891\u6587\u4ef6\u8f6c\u4e3a\u6587\u672c\u3002\u793a\u4f8b\u4ee3\u7801\uff1a<\/p>\n<pre><code>import soundfile as sf\r\nfrom kimia_infer.api.kimia import KimiAudio\r\n# \u52a0\u8f7d\u6a21\u578b\r\nmodel_path = \"moonshotai\/Kimi-Audio-7B-Instruct\"\r\nmodel = KimiAudio(model_path=model_path, load_detokenizer=True)\r\n# \u8bbe\u7f6e\u91c7\u6837\u53c2\u6570\r\nsampling_params = {\r\n\"audio_temperature\": 0.8,\r\n\"audio_top_k\": 10,\r\n\"text_temperature\": 0.0,\r\n\"text_top_k\": 5,\r\n\"audio_repetition_penalty\": 1.0,\r\n\"audio_repetition_window_size\": 64,\r\n\"text_repetition_penalty\": 1.0,\r\n\"text_repetition_window_size\": 16,\r\n}\r\n# \u51c6\u5907\u8f93\u5165\r\nasr_audio_path = \"asr_example.wav\"  # \u786e\u4fdd\u6587\u4ef6\u5b58\u5728\r\nmessages_asr = [\r\n{\"role\": \"user\", \"message_type\": \"text\", \"content\": \"\u8bf7\u8f6c\u5f55\u4ee5\u4e0b\u97f3\u9891\uff1a\"},\r\n{\"role\": \"user\", \"message_type\": \"audio\", \"content\": asr_audio_path}\r\n]\r\n# \u751f\u6210\u6587\u672c\u8f93\u51fa\r\n_, text_output = model.generate(messages_asr, **sampling_params, output_type=\"text\")\r\nprint(\"\u8f6c\u5f55\u7ed3\u679c\uff1a\", text_output)\r\n<\/code><\/pre>\n<p><strong>\u64cd\u4f5c\u6b65\u9aa4<\/strong>\uff1a<\/p>\n<ul>\n<li>\u51c6\u5907 WAV \u683c\u5f0f\u7684\u97f3\u9891\u6587\u4ef6\u3002<\/li>\n<li>\u8bbe\u7f6e\u00a0<code>messages_asr<\/code>\u00a0\u6307\u5b9a\u4efb\u52a1\u4e3a\u8f6c\u5f55\u3002<\/li>\n<li>\u8fd0\u884c\u4ee3\u7801\uff0c\u83b7\u53d6\u6587\u672c\u8f93\u51fa\u3002<\/li>\n<\/ul>\n<h4>2. \u97f3\u9891\u95ee\u7b54\uff08AQA\uff09<\/h4>\n<p>\u6839\u636e\u97f3\u9891\u5185\u5bb9\u56de\u7b54\u95ee\u9898\u3002\u793a\u4f8b\uff1a<\/p>\n<pre><code>qa_audio_path = \"qa_example.wav\"\r\nmessages_qa = [\r\n{\"role\": \"user\", \"message_type\": \"text\", \"content\": \"\u97f3\u9891\u4e2d\u8bf4\u4e86\u4ec0\u4e48\uff1f\"},\r\n{\"role\": \"user\", \"message_type\": \"audio\", \"content\": qa_audio_path}\r\n]\r\n_, text_output = model.generate(messages_qa, **sampling_params, output_type=\"text\")\r\nprint(\"\u56de\u7b54\uff1a\", text_output)\r\n<\/code><\/pre>\n<p><strong>\u64cd\u4f5c\u6b65\u9aa4<\/strong>\uff1a<\/p>\n<ul>\n<li>\u4e0a\u4f20\u5305\u542b\u4fe1\u606f\u7684\u97f3\u9891\u6587\u4ef6\u3002<\/li>\n<li>\u5728\u00a0<code>messages_qa<\/code>\u00a0\u4e2d\u8bbe\u7f6e\u5177\u4f53\u95ee\u9898\u3002<\/li>\n<li>\u83b7\u53d6\u6a21\u578b\u7684\u6587\u672c\u56de\u7b54\u3002<\/li>\n<\/ul>\n<h4>3. \u6587\u672c\u8f6c\u8bed\u97f3\uff08TTS\uff09<\/h4>\n<p>\u5c06\u6587\u672c\u8f6c\u5316\u4e3a\u8bed\u97f3\u8f93\u51fa\u3002\u793a\u4f8b\uff1a<\/p>\n<pre><code>messages_tts = [\r\n{\"role\": \"user\", \"message_type\": \"text\", \"content\": \"\u8bf7\u5c06\u4ee5\u4e0b\u6587\u672c\u8f6c\u4e3a\u8bed\u97f3\uff1a\u4f60\u597d\uff0c\u6b22\u8fce\u4f7f\u7528 Kimi-Audio\uff01\"}\r\n]\r\naudio_output, _ = model.generate(messages_tts, **sampling_params, output_type=\"audio\")\r\nsf.write(\"output.wav\", audio_output, samplerate=16000)\r\n<\/code><\/pre>\n<p><strong>\u64cd\u4f5c\u6b65\u9aa4<\/strong>\uff1a<\/p>\n<ul>\n<li>\u8f93\u5165\u9700\u8981\u8f6c\u6362\u7684\u6587\u672c\u3002<\/li>\n<li>\u8bbe\u7f6e\u00a0<code>output_type=\"audio\"<\/code>\u00a0\u83b7\u53d6\u97f3\u9891\u6570\u636e\u3002<\/li>\n<li>\u4f7f\u7528\u00a0<code>soundfile<\/code>\u00a0\u4fdd\u5b58\u4e3a WAV \u6587\u4ef6\u3002<\/li>\n<\/ul>\n<h4>4. \u7aef\u5230\u7aef\u8bed\u97f3\u5bf9\u8bdd<\/h4>\n<p>\u652f\u6301\u8fde\u7eed\u8bed\u97f3\u4ea4\u4e92\u3002\u793a\u4f8b\uff1a<\/p>\n<pre><code>messages_conversation = [\r\n{\"role\": \"user\", \"message_type\": \"audio\", \"content\": \"conversation_example.wav\"},\r\n{\"role\": \"user\", \"message_type\": \"text\", \"content\": \"\u8bf7\u56de\u590d\u4e00\u6bb5\u8bed\u97f3\uff0c\u4ecb\u7ecd\u4f60\u7684\u529f\u80fd\u3002\"}\r\n]\r\naudio_output, text_output = model.generate(messages_conversation, **sampling_params, output_type=\"both\")\r\nsf.write(\"response.wav\", audio_output, samplerate=16000)\r\nprint(\"\u6587\u672c\u56de\u590d\uff1a\", text_output)\r\n<\/code><\/pre>\n<p><strong>\u64cd\u4f5c\u6b65\u9aa4<\/strong>\uff1a<\/p>\n<ul>\n<li>\u63d0\u4f9b\u521d\u59cb\u8bed\u97f3\u8f93\u5165\u548c\u6587\u672c\u6307\u4ee4\u3002<\/li>\n<li>\u8bbe\u7f6e\u00a0<code>output_type=\"both\"<\/code>\u00a0\u83b7\u53d6\u8bed\u97f3\u548c\u6587\u672c\u56de\u590d\u3002<\/li>\n<li>\u4fdd\u5b58\u97f3\u9891\u8f93\u51fa\u5e76\u67e5\u770b\u6587\u672c\u3002<\/li>\n<\/ul>\n<h4>5. \u4f7f\u7528\u8bc4\u4f30\u5de5\u5177\u5305<\/h4>\n<p>Kimi-Audio \u63d0\u4f9b\u00a0<code>Kimi-Audio-Evalkit<\/code>\u00a0\u7528\u4e8e\u6a21\u578b\u6027\u80fd\u8bc4\u4f30\u3002\u5b89\u88c5\uff1a<\/p>\n<pre><code>git clone https:\/\/github.com\/MoonshotAI\/Kimi-Audio-Evalkit.git\r\ncd Kimi-Audio-Evalkit\r\npip install -r requirements.txt\r\n<\/code><\/pre>\n<p>\u8fd0\u884c\u8bc4\u4f30\uff1a<\/p>\n<pre><code>bash run_audio.sh --model Kimi-Audio --dataset all\r\n<\/code><\/pre>\n<p><strong>\u64cd\u4f5c\u6b65\u9aa4<\/strong>\uff1a<\/p>\n<ul>\n<li>\u4e0b\u8f7d\u6570\u636e\u96c6\u5230\u6307\u5b9a\u76ee\u5f55\u3002<\/li>\n<li>\u914d\u7f6e\u00a0<code>config.yaml<\/code>\u00a0\u6307\u5b9a\u6570\u636e\u96c6\u8def\u5f84\u3002<\/li>\n<li>\u8fd0\u884c\u811a\u672c\u751f\u6210\u8bc4\u4f30\u62a5\u544a\u3002<\/li>\n<\/ul>\n<h3>\u6ce8\u610f\u4e8b\u9879<\/h3>\n<ul>\n<li>\u786e\u4fdd\u97f3\u9891\u6587\u4ef6\u683c\u5f0f\u4e3a WAV\uff0c\u91c7\u6837\u7387 16kHz\u3002<\/li>\n<li>GPU \u73af\u5883\u53ef\u52a0\u901f\u63a8\u7406\uff0c\u63a8\u8350\u4f7f\u7528 CUDA 12.4\u3002<\/li>\n<li>\u6a21\u578b\u52a0\u8f7d\u9700\u8981\u8f83\u5927\u5185\u5b58\uff0c\u5efa\u8bae\u81f3\u5c11 16GB \u663e\u5b58\u3002<\/li>\n<li>\u975e Docker \u90e8\u7f72\u9700\u624b\u52a8\u5b89\u88c5\u7cfb\u7edf\u4f9d\u8d56\uff0c\u53c2\u8003 GitHub \u6587\u6863\u3002<\/li>\n<\/ul>\n<p>&nbsp;<\/p>\n<h2>\u5e94\u7528\u573a\u666f<\/h2>\n<ol>\n<li><strong>\u667a\u80fd\u5ba2\u670d<\/strong><br \/>\nKimi-Audio \u53ef\u7528\u4e8e\u6784\u5efa\u8bed\u97f3\u4ea4\u4e92\u7684\u5ba2\u670d\u7cfb\u7edf\u3002\u5b83\u901a\u8fc7\u8bed\u97f3\u8bc6\u522b\u8f6c\u5f55\u7528\u6237\u95ee\u9898\uff0c\u7ed3\u5408\u97f3\u9891\u95ee\u7b54\u63d0\u4f9b\u89e3\u7b54\uff0c\u5e76\u751f\u6210\u81ea\u7136\u8bed\u97f3\u56de\u590d\u3002\u9002\u5408\u7535\u5546\u5e73\u53f0\u6216\u6280\u672f\u652f\u6301\u573a\u666f\uff0c\u63d0\u5347\u7528\u6237\u4f53\u9a8c\u3002<\/li>\n<li><strong>\u6559\u80b2\u8f85\u52a9<\/strong><br \/>\n\u5728\u8bed\u8a00\u5b66\u4e60\u4e2d\uff0cKimi-Audio \u53ef\u8f6c\u5f55\u5b66\u751f\u53d1\u97f3\uff0c\u5206\u6790\u60c5\u611f\u548c\u8bed\u8c03\uff0c\u63d0\u4f9b\u53cd\u9988\u3002\u5b83\u8fd8\u80fd\u5c06\u6559\u5b66\u6587\u672c\u8f6c\u4e3a\u8bed\u97f3\uff0c\u751f\u6210\u542c\u529b\u6750\u6599\uff0c\u9002\u5408\u5728\u7ebf\u6559\u80b2\u5e73\u53f0\u3002<\/li>\n<li><strong>\u5185\u5bb9\u521b\u4f5c<\/strong><br \/>\n\u89c6\u9891\u5236\u4f5c\u8005\u53ef\u4f7f\u7528 Kimi-Audio \u751f\u6210\u5b57\u5e55\u6216\u914d\u97f3\u3002\u5b83\u80fd\u4e3a\u89c6\u9891\u81ea\u52a8\u751f\u6210\u7cbe\u51c6\u5b57\u5e55\uff0c\u6216\u5c06\u811a\u672c\u8f6c\u4e3a\u591a\u79cd\u97f3\u8272\u7684\u8bed\u97f3\uff0c\u7b80\u5316\u540e\u671f\u5236\u4f5c\u3002<\/li>\n<li><strong>\u533b\u7597\u8bb0\u5f55<\/strong><br \/>\n\u533b\u751f\u53ef\u901a\u8fc7\u8bed\u97f3\u8f93\u5165\u75c5\u4f8b\uff0cKimi-Audio \u8f6c\u5f55\u4e3a\u6587\u672c\u5e76\u5206\u7c7b\u60c5\u611f\uff0c\u8f85\u52a9\u8bca\u65ad\u60a3\u8005\u60c5\u7eea\u72b6\u6001\u3002\u9002\u7528\u4e8e\u533b\u9662\u4fe1\u606f\u7cfb\u7edf\u3002<\/li>\n<\/ol>\n<p>&nbsp;<\/p>\n<h2>QA<\/h2>\n<ol>\n<li><strong>Kimi-Audio \u652f\u6301\u54ea\u4e9b\u8bed\u8a00\uff1f<\/strong><br \/>\nKimi-Audio \u652f\u6301\u591a\u8bed\u8a00\u8bed\u97f3\u8bc6\u522b\u548c\u751f\u6210\uff0c\u5c24\u5176\u5728\u82f1\u8bed\u548c\u4e2d\u6587\u4e0a\u8868\u73b0\u4f18\u5f02\u3002\u5176\u4ed6\u8bed\u8a00\u652f\u6301\u9700\u53c2\u8003\u5b98\u65b9\u6587\u6863\u3002<\/li>\n<li><strong>\u5982\u4f55\u4f18\u5316\u63a8\u7406\u901f\u5ea6\uff1f<\/strong><br \/>\n\u4f7f\u7528 GPU \u52a0\u901f\uff0c\u5b89\u88c5\u00a0<code>flash-attn<\/code>\u00a0\u5e93\uff0c\u8bbe\u7f6e\u00a0<code>torch_dtype=torch.bfloat16<\/code>\u3002\u6b64\u5916\uff0c\u8c03\u6574\u00a0<code>audio_top_k<\/code>\u00a0\u548c\u00a0<code>text_top_k<\/code>\u00a0\u53c2\u6570\u53ef\u5e73\u8861\u901f\u5ea6\u548c\u8d28\u91cf\u3002<\/li>\n<li><strong>\u6a21\u578b\u662f\u5426\u652f\u6301\u5b9e\u65f6\u5bf9\u8bdd\uff1f<\/strong><br \/>\n\u662f\u7684\uff0cKimi-Audio \u7684\u6d41\u5f0f\u89e3\u7801\u5668\u652f\u6301\u4f4e\u5ef6\u8fdf\u97f3\u9891\u751f\u6210\uff0c\u9002\u5408\u5b9e\u65f6\u8bed\u97f3\u4ea4\u4e92\u3002<\/li>\n<li><strong>\u5982\u4f55\u6dfb\u52a0\u81ea\u5b9a\u4e49\u6570\u636e\u96c6\u8fdb\u884c\u8bc4\u4f30\uff1f<\/strong><br \/>\n\u5728\u00a0<code>Kimi-Audio-Evalkit<\/code>\u00a0\u4e2d\u521b\u5efa JSONL \u6587\u4ef6\uff0c\u5305\u542b\u00a0<code>index<\/code>\u3001<code>audio_path<\/code>\u00a0\u548c\u00a0<code>question<\/code>\u00a0\u5b57\u6bb5\u3002\u4fee\u6539\u00a0<code>config.yaml<\/code>\u00a0\u6307\u5b9a\u6570\u636e\u96c6\u8def\u5f84\uff0c\u8fd0\u884c\u8bc4\u4f30\u811a\u672c\u5373\u53ef\u3002<\/li>\n<\/ol>\n","protected":false},"excerpt":{"rendered":"<p>Kimi-Audio \u662f\u7531 Moonshot AI \u5f00\u53d1\u7684\u4e00\u6b3e\u5f00\u6e90\u97f3\u9891\u57fa\u7840\u6a21\u578b\uff0c\u4e13\u6ce8\u4e8e\u97f3\u9891\u7406\u89e3\u3001\u751f\u6210\u548c\u5bf9\u8bdd\u3002\u5b83\u652f\u6301\u591a\u79cd\u97f3\u9891\u5904\u7406\u4efb\u52a1\uff0c\u4f8b\u5982\u8bed\u97f3\u8bc6\u522b\u3001\u97f3\u9891\u95ee\u7b54\u548c\u8bed\u97f3\u60c5\u611f\u8bc6\u522b\u3002\u6a21\u578b\u7ecf\u8fc7\u8d85\u8fc7 1300 \u4e07\u5c0f\u65f6\u7684\u97f3\u9891\u6570\u636e\u9884\u8bad\u7ec3\uff0c\u7ed3\u5408\u521b\u65b0\u7684\u6df7\u5408\u67b6\u6784\uff0c\u5728&#8230;<\/p>\n","protected":false},"author":1,"featured_media":62316,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[20,392,400],"tags":[230,215,216],"class_list":["post-30867","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-tool","category-models","category-speech-model","tag-aikaiyuanxiangmu","tag-aiwenbenzhuanyuyin","tag-aiyuyinzhuanwenben"],"_links":{"self":[{"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/posts\/30867","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/comments?post=30867"}],"version-history":[{"count":0,"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/posts\/30867\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/media\/62316"}],"wp:attachment":[{"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/media?parent=30867"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/categories?post=30867"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/tags?post=30867"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}