Open dahu1 opened 11 months ago
我在gigaspeech 上看到k2_gigaspeech模型的效果是10.40 / 10.51 ,如图所示:
而且我看k2 gigaspeech的测试结果 也是这个,如图所示:
但是为什么在Leaderboard上的结果对不上呢?如图所示:
Because this leaderboard applies far more advanced text normalizations than an academic baseline recipe.
我在gigaspeech 上看到k2_gigaspeech模型的效果是10.40 / 10.51 ,如图所示:
而且我看k2 gigaspeech的测试结果 也是这个,如图所示:
但是为什么在Leaderboard上的结果对不上呢?如图所示: