fb700/chatglm-fitness-RLHF · Discussions

Resources

View closed (2)

Adding `safetensors` variant of this model

#25 opened 8 months ago by

大佬有没有兴趣再搞一搞llama3-8b

#24 opened about 1 year ago by

Hello

#23 opened over 1 year ago by

model = model.half().quantize(4).cuda() 运行显示错误"AttributeError: 'Linear' object has no attribute 'bias'"

#22 opened over 1 year ago by

请问训练所用的数据集能否公开？

#21 opened over 1 year ago by

请问作者，rlhf的actor loss是否下降和正常收敛呢？能不能给一些经验的超参数设置？请教了

#20 opened over 1 year ago by

rlhf

#19 opened over 1 year ago by

Cannot copy out of meta tensor; no data!，报错代码地方为model = model.half().quantize(4).cuda() ，猜测是量化相关问题或者作者的模型上传的时候有遗漏的文件

#18 opened almost 2 years ago by

关于两种加载模型文件方式的区别

#17 opened almost 2 years ago by

如何在多gpu上加载

#16 opened almost 2 years ago by

关于模型权重

#15 opened almost 2 years ago by

Lora和RLHF训练的代码开源了吗

#14 opened almost 2 years ago by

最大token

#12 opened almost 2 years ago by

还可以再次进行自我认知的lora的训练吗

#11 opened almost 2 years ago by

api 调用您的模型出现错误

#10 opened almost 2 years ago by

牛的无限长的原理和chatglm2是一个道理吗？

#9 opened almost 2 years ago by

测试了一下，很好用，比chatglm2还好用。期待更多作品。

#8 opened almost 2 years ago by

训练代码

#6 opened almost 2 years ago by

启动加载很慢，需要130秒

#5 opened almost 2 years ago by

的确很好用！！

#4 opened almost 2 years ago by

会不会基于ChatGLM2-6B进行迭代？

#3 opened almost 2 years ago by

牛逼

#2 opened almost 2 years ago by