lower01

Results 7 issues of lower01

it is 2022.8.17 now,but why the file about train.py and dataset.py are still empty?what happened?

### Question 刘哥,如果使用少部分其它数据进行微调,数据集文件夹的格式应该是怎样的呢?我看到coco2014_val_gpt4_qa_30x3.jsonl中的格式是id-image-instruction-output,意思是需要在相同目录下准备一个对应的图片文件夹(图片名称对应于image编号)吗? 由jsonl文件和图片文件夹两部分组成训练数据?

Can the labels in the data only represent the degree of the node? Or can it represent a numerical representation of arbitrary features? For example, I use 0 to represent...

I saw elsewhere that the project code you shared before included this "preProcess" folder, why not now? I think it is very important for data understanding. ![WW6Y0IX5@UHMZUG6WKB3)WJ](https://github.com/benedekrozemberczki/SimGNN/assets/45301618/6aec279c-08d3-4e2f-bad0-6178f31bb1dd)

您好,请问你们这次微调的LLama3-Chinese-8B-Instruct的模型基础是Llama3-8B-base还是Llama3-8B-instruct? 是直接基于Llama3-8B-instruct模型使用中文指令数据集微调得到的,还是先对Llama3-8B-base进行大规模中文持续预训练后,再通过中文指令数据集微调得到的? 然后LLama3-Chinese-8B-Instruct的微调用了多大的中文数据集呢?

![Image](https://github.com/user-attachments/assets/4ce0b3d2-6ed9-4753-8ed5-5d9f83cb70c2) 例如这个中文流式语音识别服务文档,如果我是想测试whisper-large-v3的ASR效果,我该如何设置参数,之前已经在modelscope社区下载了模型文件到本地了并传到服务器上了: ![Image](https://github.com/user-attachments/assets/198ef060-8831-49a5-8cd4-c4700cfebdbb) 请问我该如何设置这个服务端参数?

question

inputs = tf.keras.Input(shape=(model_settings['fingerprint_size']), name='input') ValueError: Cannot convert '490' to a shape. 如何解决这个问题?代码存在错误,升到v2后是否验证过=train.py?