{"id":30170,"date":"2025-04-08T16:38:31","date_gmt":"2025-04-08T08:38:31","guid":{"rendered":"https:\/\/www.aisharenet.com\/?p=30170"},"modified":"2025-04-08T16:38:31","modified_gmt":"2025-04-08T08:38:31","slug":"dolphinmianxiangyaai","status":"publish","type":"post","link":"https:\/\/www.kdjingpai.com\/pt\/dolphinmianxiangyaai\/","title":{"rendered":"Dolphin\uff1a\u9762\u5411\u4e9a\u6d32\u8bed\u8a00\u8bc6\u522b\u4e0e\u8bed\u97f3\u8f6c\u6587\u672c\u6a21\u578b"},"content":{"rendered":"<p>Dolphin \u662f\u7531 DataoceanAI \u548c\u6e05\u534e\u5927\u5b66\u5408\u4f5c\u5f00\u53d1\u7684\u4e00\u4e2a\u5f00\u6e90\u6a21\u578b\uff0c\u4e13\u6ce8\u4e8e\u4e9a\u6d32\u8bed\u8a00\u7684\u8bed\u97f3\u8bc6\u522b\u548c\u8bed\u8a00\u8bc6\u522b\u3002\u5b83\u652f\u6301\u4e1c\u4e9a\u3001\u5357\u4e9a\u3001\u4e1c\u5357\u4e9a\u53ca\u4e2d\u4e1c\u5730\u533a\u7684 40 \u79cd\u8bed\u8a00\uff0c\u4ee5\u53ca 22 \u79cd\u4e2d\u56fd\u65b9\u8a00\u3002\u6a21\u578b\u57fa\u4e8e\u8d85\u8fc7 21 \u4e07\u5c0f\u65f6\u7684\u97f3\u9891\u6570\u636e\u8bad\u7ec3\uff0c\u7ed3\u5408\u4e86\u4e13\u6709\u548c\u516c\u5f00\u6570\u636e\u96c6\u3002Dolphin \u80fd\u5c06\u8bed\u97f3\u8f6c\u4e3a\u6587\u672c\uff0c\u8fd8\u80fd\u68c0\u6d4b\u8bed\u97f3\u90e8\u5206\uff08VAD\uff09\u3001\u5206\u5272\u97f3\u9891\u548c\u8bc6\u522b\u8bed\u8a00\uff08LID\uff09\u3002\u5b83\u8bbe\u8ba1\u7b80\u5355\uff0c\u4ee3\u7801\u548c\u90e8\u5206\u6a21\u578b\u5728 GitHub \u4e0a\u514d\u8d39\u5f00\u653e\uff0c\u9002\u5408\u5f00\u53d1\u8005\u4f7f\u7528\u3002<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"aligncenter size-full wp-image-30171\" title=\"Dolphin\uff1a\u9762\u5411\u4e9a\u6d32\u8bed\u8a00\u8bc6\u522b\u4e0e\u8bed\u97f3\u8f6c\u6587\u672c\u6a21\u578b-1\" src=\"https:\/\/www.kdjingpai.com\/wp-content\/uploads\/2025\/04\/0b5a18bfa01e372.jpg\" alt=\"Dolphin\uff1a\u9762\u5411\u4e9a\u6d32\u8bed\u8a00\u8bc6\u522b\u4e0e\u8bed\u97f3\u8f6c\u6587\u672c\u6a21\u578b-1\" width=\"1051\" height=\"241\" srcset=\"https:\/\/www.kdjingpai.com\/wp-content\/uploads\/2025\/04\/0b5a18bfa01e372.jpg 1051w, https:\/\/www.kdjingpai.com\/wp-content\/uploads\/2025\/04\/0b5a18bfa01e372-768x176.jpg 768w, https:\/\/www.kdjingpai.com\/wp-content\/uploads\/2025\/04\/0b5a18bfa01e372-18x4.jpg 18w\" sizes=\"auto, (max-width: 1051px) 100vw, 1051px\" \/><\/p>\n<p>&nbsp;<\/p>\n<h2>\u529f\u80fd\u5217\u8868<\/h2>\n<ul>\n<li>\u652f\u6301 40 \u79cd\u4e9a\u6d32\u8bed\u8a00\u548c 22 \u79cd\u4e2d\u56fd\u65b9\u8a00\u7684\u8bed\u97f3\u8f6c\u6587\u672c\u3002<\/li>\n<li>\u63d0\u4f9b\u8bed\u97f3\u6d3b\u52a8\u68c0\u6d4b\uff08VAD\uff09\uff0c\u627e\u51fa\u97f3\u9891\u4e2d\u7684\u8bed\u97f3\u7247\u6bb5\u3002<\/li>\n<li>\u652f\u6301\u97f3\u9891\u5206\u5272\uff0c\u5c06\u957f\u97f3\u9891\u5207\u6210\u5c0f\u6bb5\u5904\u7406\u3002<\/li>\n<li>\u5b9e\u73b0\u8bed\u8a00\u8bc6\u522b\uff08LID\uff09\uff0c\u5224\u65ad\u97f3\u9891\u7684\u8bed\u8a00\u6216\u65b9\u8a00\u3002<\/li>\n<li>\u5f00\u6e90\u4ee3\u7801\u548c\u6a21\u578b\uff0c\u5141\u8bb8\u7528\u6237\u4fee\u6539\u548c\u5b9a\u5236\u3002<\/li>\n<li>\u63d0\u4f9b base \u548c small \u4e24\u79cd\u6a21\u578b\uff0c\u9002\u5e94\u4e0d\u540c\u9700\u6c42\u3002<\/li>\n<li>\u4f7f\u7528\u53cc\u5c42\u6807\u8bb0\u7cfb\u7edf\uff0c\u533a\u5206\u8bed\u8a00\u548c\u5730\u533a\uff08\u5982\u00a0<code>&lt;zh&gt;&lt;CN&gt;<\/code>\uff09\u3002<\/li>\n<\/ul>\n<p>&nbsp;<\/p>\n<h2>\u4f7f\u7528\u5e2e\u52a9<\/h2>\n<p>Dolphin \u7684\u5b89\u88c5\u548c\u4f7f\u7528\u6d41\u7a0b\u7b80\u5355\uff0c\u9002\u5408\u6709\u57fa\u7840\u7f16\u7a0b\u80fd\u529b\u7684\u7528\u6237\u3002\u4ee5\u4e0b\u662f\u8be6\u7ec6\u6b65\u9aa4\u3002<\/p>\n<h3>\u5b89\u88c5\u6d41\u7a0b<\/h3>\n<ol>\n<li><strong>\u51c6\u5907\u73af\u5883<\/strong><br \/>\n\u9700\u8981 Python 3.8 \u6216\u4ee5\u4e0a\u7248\u672c\uff0c\u4ee5\u53ca FFmpeg \u6765\u5904\u7406\u97f3\u9891\u3002<\/p>\n<ul>\n<li>\u68c0\u67e5 Python\uff1a\u5728\u7ec8\u7aef\u8f93\u5165\u00a0<code>python --version<\/code>\uff0c\u786e\u8ba4\u7248\u672c\u3002<\/li>\n<li>\u672a\u5b89\u88c5 Python \u53ef\u4ece\u00a0python.org\u00a0\u4e0b\u8f7d\u3002<\/li>\n<li>\u5b89\u88c5 FFmpeg\uff1a\u6839\u636e\u7cfb\u7edf\u8fd0\u884c\u547d\u4ee4\uff1a\n<ul>\n<li>Ubuntu\/Debian\uff1a\n<pre><code>sudo apt update &amp;&amp; sudo apt install ffmpeg\r\n<\/code><\/pre>\n<\/li>\n<li>macOS\uff1a\n<pre><code>brew install ffmpeg\r\n<\/code><\/pre>\n<\/li>\n<li>Windows\uff1a\n<pre><code>choco install ffmpeg\r\n<\/code><\/pre>\n<\/li>\n<\/ul>\n<p>\u672a\u5b89\u88c5\u5305\u7ba1\u7406\u5de5\u5177\u53ef\u4ece\u00a0FFmpeg \u5b98\u7f51\u00a0\u4e0b\u8f7d\u3002<\/li>\n<\/ul>\n<\/li>\n<li><strong>\u5b89\u88c5 Dolphin<\/strong><br \/>\n\u6709\u4e24\u79cd\u65b9\u5f0f\uff1a<\/p>\n<ul>\n<li><strong>\u7528 pip \u5b89\u88c5<\/strong><br \/>\n\u5728\u7ec8\u7aef\u8f93\u5165\uff1a<\/p>\n<pre><code>pip install -U dataoceanai-dolphin\r\n<\/code><\/pre>\n<p>\u8fd9\u4f1a\u5b89\u88c5\u6700\u65b0\u7a33\u5b9a\u7248\u3002<\/li>\n<li><strong>\u4ece\u6e90\u4ee3\u7801\u5b89\u88c5<\/strong><br \/>\n\u8981\u7528\u6700\u65b0\u5f00\u53d1\u7248\uff0c\u53ef\u4ece GitHub \u83b7\u53d6\uff1a<\/p>\n<ol>\n<li>\u514b\u9686\u4ed3\u5e93\uff1a\n<pre><code>git clone https:\/\/github.com\/DataoceanAI\/Dolphin.git\r\n<\/code><\/pre>\n<\/li>\n<li>\u8fdb\u5165\u76ee\u5f55\uff1a\n<pre><code>cd Dolphin\r\n<\/code><\/pre>\n<\/li>\n<li>\u5b89\u88c5\uff1a\n<pre><code>pip install .\r\n<\/code><\/pre>\n<\/li>\n<\/ol>\n<\/li>\n<\/ul>\n<\/li>\n<li><strong>\u4e0b\u8f7d\u6a21\u578b<\/strong><br \/>\nDolphin \u6709 4 \u79cd\u6a21\u578b\uff0c\u76ee\u524d base\uff08140M \u53c2\u6570\uff09\u548c small\uff08372M \u53c2\u6570\uff09\u53ef\u514d\u8d39\u4e0b\u8f7d\u3002<\/p>\n<ul>\n<li>\u4ece\u00a0<a href=\"https:\/\/huggingface.co\/DataoceanAI\">Hugging Face<\/a>\u00a0\u83b7\u53d6\u6a21\u578b\u6587\u4ef6\u3002<\/li>\n<li>\u4fdd\u5b58\u5230\u6307\u5b9a\u8def\u5f84\uff0c\u5982\u00a0<code>\/data\/models\/dolphin\/<\/code>\u3002<\/li>\n<li>base \u6a21\u578b\u901f\u5ea6\u5feb\uff0csmall \u6a21\u578b\u7cbe\u5ea6\u66f4\u9ad8\u3002<\/li>\n<\/ul>\n<\/li>\n<\/ol>\n<h3>\u4f7f\u7528\u65b9\u6cd5<\/h3>\n<p>\u652f\u6301\u547d\u4ee4\u884c\u548c Python \u64cd\u4f5c\u3002<\/p>\n<h4>\u547d\u4ee4\u884c\u64cd\u4f5c<\/h4>\n<ol>\n<li><strong>\u8bed\u97f3\u8f6c\u6587\u672c<\/strong><br \/>\n\u51c6\u5907\u97f3\u9891\u6587\u4ef6\uff08\u5982\u00a0<code>audio.wav<\/code>\uff09\uff0c\u8f93\u5165\uff1a<\/li>\n<\/ol>\n<pre><code>dolphin audio.wav\r\n<\/code><\/pre>\n<p>\u7cfb\u7edf\u81ea\u52a8\u4e0b\u8f7d\u9ed8\u8ba4\u6a21\u578b\u5e76\u8f93\u51fa\u6587\u672c\u3002\u97f3\u9891\u9700\u4e3a WAV \u683c\u5f0f\uff0c\u53ef\u7528 FFmpeg \u8f6c\u6362\uff1a<\/p>\n<pre><code>ffmpeg -i input.mp3 output.wav\r\n<\/code><\/pre>\n<ol start=\"2\">\n<li><strong>\u6307\u5b9a\u6a21\u578b\u548c\u8def\u5f84<\/strong><br \/>\n\u7528 small \u6a21\u578b\uff1a<\/li>\n<\/ol>\n<pre><code>dolphin audio.wav --model small --model_dir \/data\/models\/dolphin\/\r\n<\/code><\/pre>\n<ol start=\"3\">\n<li><strong>\u6307\u5b9a\u8bed\u8a00\u548c\u5730\u533a<\/strong><br \/>\n\u7528\u53cc\u5c42\u6807\u8bb0\u8bc6\u522b\u4e2d\u6587\u666e\u901a\u8bdd\uff1a<\/li>\n<\/ol>\n<pre><code>dolphin audio.wav --model small --model_dir \/data\/models\/dolphin\/ --lang_sym \"zh\" --region_sym \"CN\"\r\n<\/code><\/pre>\n<ul>\n<li><code>lang_sym<\/code>\u00a0\u662f\u8bed\u8a00\u4ee3\u7801\uff0c\u5982 &#8220;zh&#8221;\uff08\u4e2d\u6587\uff09\u3002<\/li>\n<li><code>region_sym<\/code>\u00a0\u662f\u5730\u533a\u4ee3\u7801\uff0c\u5982 &#8220;CN&#8221;\uff08\u4e2d\u56fd\u5927\u9646\uff09\u3002<br \/>\n\u5b8c\u6574\u8bed\u8a00\u5217\u8868\u89c1\u00a0<a href=\"https:\/\/github.com\/DataoceanAI\/Dolphin\/blob\/main\/languages.md\">languages.md<\/a>\u3002<\/li>\n<\/ul>\n<ol start=\"4\">\n<li><strong>\u586b\u5145\u77ed\u97f3\u9891<\/strong><br \/>\n\u97f3\u9891\u4e0d\u8db3 30 \u79d2\u65f6\uff0c\u53ef\u7528\u00a0<code>--padding_speech true<\/code>\u00a0\u586b\u5145\uff1a<\/li>\n<\/ol>\n<pre><code>dolphin audio.wav --model small --model_dir \/data\/models\/dolphin\/ --lang_sym \"zh\" --region_sym \"CN\" --padding_speech true\r\n<\/code><\/pre>\n<h4>Python \u4ee3\u7801\u64cd\u4f5c<\/h4>\n<ol>\n<li><strong>\u52a0\u8f7d\u97f3\u9891\u548c\u6a21\u578b<\/strong><br \/>\n\u5728 Python \u4e2d\u8fd0\u884c\uff1a<\/li>\n<\/ol>\n<pre><code>import dolphin\r\nwaveform = dolphin.load_audio(\"audio.wav\")  # \u52a0\u8f7d\u97f3\u9891\r\nmodel = dolphin.load_model(\"small\", \"\/data\/models\/dolphin\/\", \"cuda\")  # \u52a0\u8f7d\u6a21\u578b\r\n<\/code><\/pre>\n<ul>\n<li><code>\"cuda\"<\/code>\u00a0\u7528 GPU\uff0c\u65e0 GPU \u6539\u4e3a\u00a0<code>\"cpu\"<\/code>\u3002<\/li>\n<\/ul>\n<ol start=\"2\">\n<li><strong>\u6267\u884c\u8bc6\u522b<\/strong><br \/>\n\u5904\u7406\u97f3\u9891\u5e76\u8f93\u51fa\uff1a<\/p>\n<pre><code>result = model(waveform)  # \u8f6c\u6587\u672c\r\nprint(result.text)  # \u663e\u793a\u7ed3\u679c\r\n<\/code><\/pre>\n<\/li>\n<li><strong>\u6307\u5b9a\u8bed\u8a00\u548c\u5730\u533a<\/strong><br \/>\n\u6dfb\u52a0\u53c2\u6570\uff1a<\/p>\n<pre><code>result = model(waveform, lang_sym=\"zh\", region_sym=\"CN\")\r\nprint(result.text)\r\n<\/code><\/pre>\n<\/li>\n<\/ol>\n<h3>\u7279\u8272\u529f\u80fd\u64cd\u4f5c<\/h3>\n<ul>\n<li><strong>\u8bed\u97f3\u6d3b\u52a8\u68c0\u6d4b\uff08VAD\uff09<\/strong><br \/>\n\u81ea\u52a8\u68c0\u6d4b\u8bed\u97f3\u7247\u6bb5\u5e76\u6807\u6ce8\u65f6\u95f4\uff0c\u5982\uff1a<\/p>\n<pre><code>0.0-2.5s: \u4f60\u597d\r\n3.0-4.5s: \u4eca\u5929\u5929\u6c14\u5f88\u597d\r\n<\/code><\/pre>\n<\/li>\n<li><strong>\u8bed\u8a00\u8bc6\u522b\uff08LID\uff09<\/strong><br \/>\n\u5224\u65ad\u97f3\u9891\u8bed\u8a00\uff0c\u4f8b\u5982\uff1a<\/p>\n<pre><code>dolphin audio.wav --model small --model_dir \/data\/models\/dolphin\/\r\n<\/code><\/pre>\n<p>\u8f93\u51fa\u5982\u00a0<code>&lt;zh&gt;<\/code>\uff08\u4e2d\u6587\uff09\u6216\u00a0<code>&lt;ja&gt;<\/code>\uff08\u65e5\u8bed\uff09\u3002<\/li>\n<li><strong>\u53cc\u5c42\u8bed\u8a00\u6807\u8bb0<\/strong><br \/>\n\u7528\u4e24\u7ea7\u6807\u8bb0\u533a\u5206\u8bed\u8a00\u548c\u5730\u533a\uff0c\u5982\u00a0<code>&lt;zh&gt;&lt;CN&gt;<\/code>\uff08\u4e2d\u6587\u666e\u901a\u8bdd\uff09\u3001<code>&lt;zh&gt;&lt;TW&gt;<\/code>\uff08\u53f0\u6e7e\u666e\u901a\u8bdd\uff09\uff0c\u63d0\u5347\u4e9a\u6d32\u8bed\u8a00\u5904\u7406\u80fd\u529b\u3002<\/li>\n<li><strong>\u6a21\u578b\u67b6\u6784<\/strong><br \/>\n\u91c7\u7528 CTC-Attention \u67b6\u6784\uff0c\u7f16\u7801\u5668\u7528 E-Branchformer\uff0c\u89e3\u7801\u5668\u7528 Transformer\uff0c\u4e13\u4e3a\u4e9a\u6d32\u8bed\u8a00\u4f18\u5316\u3002<\/li>\n<\/ul>\n<p>&nbsp;<\/p>\n<h2>\u5e94\u7528\u573a\u666f<\/h2>\n<ol>\n<li><strong>\u4f1a\u8bae\u8bb0\u5f55<\/strong><br \/>\n\u5c06\u4e9a\u6d32\u591a\u8bed\u8a00\u4f1a\u8bae\u5f55\u97f3\u8f6c\u4e3a\u6587\u672c\uff0c\u9002\u5408\u56fd\u9645\u6216\u5730\u65b9\u4f1a\u8bae\u3002<\/li>\n<li><strong>\u65b9\u8a00\u7814\u7a76<\/strong><br \/>\n\u5206\u6790 22 \u79cd\u4e2d\u56fd\u65b9\u8a00\u7684\u8bed\u97f3\u7279\u5f81\uff0c\u751f\u6210\u7814\u7a76\u6570\u636e\u3002<\/li>\n<li><strong>\u667a\u80fd\u8bbe\u5907\u5f00\u53d1<\/strong><br \/>\n\u96c6\u6210\u5230\u667a\u80fd\u8bbe\u5907\uff0c\u5b9e\u73b0\u4e9a\u6d32\u8bed\u8a00\u7684\u8bed\u97f3\u63a7\u5236\u3002<\/li>\n<\/ol>\n<p>&nbsp;<\/p>\n<h2>QA<\/h2>\n<ol>\n<li><strong>\u652f\u6301\u54ea\u4e9b\u8bed\u8a00\uff1f<\/strong><br \/>\n\u652f\u6301 40 \u79cd\u4e9a\u6d32\u8bed\u8a00\u548c 22 \u79cd\u4e2d\u56fd\u65b9\u8a00\uff0c\u8be6\u89c1\u00a0<a href=\"https:\/\/github.com\/DataoceanAI\/Dolphin\/blob\/main\/languages.md\">languages.md<\/a>\u3002<\/li>\n<li><strong>\u9700\u8981 GPU \u5417\uff1f<\/strong><br \/>\n\u4e0d\u9700\u8981\u3002CPU \u53ef\u8fd0\u884c\uff0cGPU\uff08\u652f\u6301 CUDA\uff09\u66f4\u5feb\u3002<\/li>\n<li><strong>base \u548c small \u6a21\u578b\u6709\u4f55\u4e0d\u540c\uff1f<\/strong><br \/>\nbase \u6a21\u578b\u5c0f\uff08140M \u53c2\u6570\uff09\uff0c\u9519\u8bef\u7387 33.3%\uff1bsmall \u6a21\u578b\u5927\uff08372M \u53c2\u6570\uff09\uff0c\u9519\u8bef\u7387 25.2%\u3002<\/li>\n<\/ol>\n","protected":false},"excerpt":{"rendered":"<p>Dolphin \u662f\u7531 DataoceanAI \u548c\u6e05\u534e\u5927\u5b66\u5408\u4f5c\u5f00\u53d1\u7684\u4e00\u4e2a\u5f00\u6e90\u6a21\u578b\uff0c\u4e13\u6ce8\u4e8e\u4e9a\u6d32\u8bed\u8a00\u7684\u8bed\u97f3\u8bc6\u522b\u548c\u8bed\u8a00\u8bc6\u522b\u3002\u5b83\u652f\u6301\u4e1c\u4e9a\u3001\u5357\u4e9a\u3001\u4e1c\u5357\u4e9a\u53ca\u4e2d\u4e1c\u5730\u533a\u7684 40 \u79cd\u8bed\u8a00\uff0c\u4ee5\u53ca 22 \u79cd\u4e2d\u56fd\u65b9\u8a00\u3002\u6a21\u578b\u57fa\u4e8e\u8d85\u8fc7 21 \u4e07\u5c0f\u65f6\u7684\u97f3\u9891\u6570\u636e\u8bad\u7ec3\uff0c\u7ed3\u5408\u4e86&#8230;<\/p>\n","protected":false},"author":1,"featured_media":62213,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[20],"tags":[230,216],"class_list":["post-30170","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-tool","tag-aikaiyuanxiangmu","tag-aiyuyinzhuanwenben"],"_links":{"self":[{"href":"https:\/\/www.kdjingpai.com\/pt\/wp-json\/wp\/v2\/posts\/30170","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.kdjingpai.com\/pt\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.kdjingpai.com\/pt\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.kdjingpai.com\/pt\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.kdjingpai.com\/pt\/wp-json\/wp\/v2\/comments?post=30170"}],"version-history":[{"count":0,"href":"https:\/\/www.kdjingpai.com\/pt\/wp-json\/wp\/v2\/posts\/30170\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.kdjingpai.com\/pt\/wp-json\/wp\/v2\/media\/62213"}],"wp:attachment":[{"href":"https:\/\/www.kdjingpai.com\/pt\/wp-json\/wp\/v2\/media?parent=30170"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.kdjingpai.com\/pt\/wp-json\/wp\/v2\/categories?post=30170"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.kdjingpai.com\/pt\/wp-json\/wp\/v2\/tags?post=30170"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}