Emova-ollm/emova-sft-speech-eval

EMOVA-SFT-Speech-Eval是一个用于全模态指令微调及情感口语对话评估的数据集。该数据集通过文本转语音工具将现有的文本和视觉指令数据集转换而来，包含英语和中文的带图像和不带图像的口语对话子集。数据集包含的例子涵盖了不同的情感类别，如中性、快乐、悲伤、愤怒等。

EMOVA-SFT-Speech-Eval is an evaluation dataset for omni-modal instruction tuning and emotional spoken dialogue. The dataset is created by converting existing text and visual instruction datasets into speech using Text-to-Speech (TTS) tools, including subsets of English and Chinese spoken dialogues with and without image inputs. The examples in the dataset cover various emotional categories such as neutral, happy, sad, angry, etc.