Nope, you can have any format. But you will have to provide the data loader function as an input parameter. Only then the `train` function will know how to handle the data.
Using your modal training code I got error on trainer stating AttributeError: 'TrainingArguments' object has no attribute 'dataset_num_proc' Can you help me .
@@AIBites Thank you so much for asking and for your willingness to provide resources that cater to my learning preferences. I deeply appreciate it. I believe hands-on videos would be exceptionally beneficial for me, as they offer a more direct and engaging way to grasp the applications of theoretical concepts in real-life scenarios. The practical aspect of learning through such videos can significantly enhance understanding and retention. Once again, thank you for your thoughtful video!!!!!!
I'm getting error RuntimeError: No GPU found. A GPU is needed for quantization.
Yes, could you confirm if the machine you are running on has a GPU. Are you running locally or on the cloud? please clarify
fp16 also has a bigger mantissa than bfloat which benefits normalized or bounded activation functions (e.g. sigmoid)
How did you push your own custom dataset on huggingface?
Checkout the bunch of commands available in the HF command line tools. It's quite easy actually
do we need to prepare data set in specific format as Instruction and Response one ? means is it mandatory to follow the same way ?
Nope, you can have any format. But you will have to provide the data loader function as an input parameter. Only then the `train` function will know how to handle the data.
Using your modal training code I got error on trainer stating
AttributeError: 'TrainingArguments' object has no attribute 'dataset_num_proc'
Can you help me .
sorry about the late reply, but did you manage to fix this or is it still a problem?
Nice video. When can I find the model downloaded in colab? Can you help me?
I am not sure if I shared the model. The training code is however available on the github page. Hope you found that.
Amazing video! Very helpful!
Glad it was helpful!
Your video is really helpful!!! thank you !!!!
You're welcome 🙂 so would you like more of hands on videos or theoretical paper videos?
@@AIBites Thank you so much for asking and for your willingness to provide resources that cater to my learning preferences. I deeply appreciate it. I believe hands-on videos would be exceptionally beneficial for me, as they offer a more direct and engaging way to grasp the applications of theoretical concepts in real-life scenarios. The practical aspect of learning through such videos can significantly enhance understanding and retention. Once again, thank you for your thoughtful video!!!!!!
Very detailed and informative vedio!
thanks. Glad it was useful.
This is really helpful!
thank you :)
Thank you
You are welcome 🙂
Good
thank you! 🙂