QinHsiu comments

Results 15 comments of


                                            QinHsiu

Preprocessing for ml1m?

@Book1996 Can you give me your code of dealing the ml-1m dataset?If you can,Thank you so much! My email: [email protected]

Preprocessing for ml1m?

Can you provide me the ml-1m preprocess code? that would help me a lot, thank you. Email: [email protected]

Preprocessing for ml1m?

@WDdeBWT Can you provide me the ml-1m preprocess code? that would help me a lot, thank you. Email: [email protected]

Preprocessing for ml1m?

@conquerSelf Can you provide me the ml-1m preprocess code? that would help me a lot, thank you. Email: [email protected]

[Bug] RuntimeError: stft requires the return_complex parameter be given for real inputs

I have the same problem, can anyone help me fix this problem?

RuntimeError: "reflection_pad1d_out_template" not implemented for 'Short' : when using separate(...) method

I have the sample problem, RuntimeError: "reflection_pad1d_out_template" not implemented for 'Long'

sh: 1: next: not found

I run the following command: num=os.system("cat test.txt | wc -l") , and face the problem too, sh: 1: 224400: not found

Reproducing ML-1M results from the paper.

I got the result when I set the dropout ratio as 0.1, you can have a try.

qwen2-7b-instruct，作了SFT微调之后，在预测过程中，出现错误

> Hi, it suggests traininig stability issues and it normally depends on how you finetune the model to find the way to mitgate the issue. without more information, lowering learning...

qwen2-7b-instruct，作了SFT微调之后，在预测过程中，出现错误

非常感谢你的回复，我会先做尝试，后续如有问题再来请较，非常感谢！ ---原始邮件--- 发件人: "Yang ***@***.***> 发送时间: 2024年8月1日(周四) 晚上6:06 收件人: ***@***.***>; 抄送: ***@***.******@***.***>; 主题: Re: [QwenLM/Qwen2] qwen2-7b-instruct，作了SFT微调之后，在预测过程中，出现错误 (Issue #808) 如果是全量参数训练，3e-4的学习率大了些，可能会导致训练时梯度爆炸。建议考虑e-5或者e-6的学习率。 — Reply to this email directly, view it on GitHub, or...