search

SteveKGYang / MentalLLaMA

This repository introduces MentaLLaMA, the first open-source instruction following large language model for interpretable mental health analysis.
MIT License
192 stars 18 forks source link

Getting an error on mentalBart model #4

Open PRANJALRANA11 opened 8 months ago

PRANJALRANA11 commented 8 months ago

image

TianlinZhang668 commented 8 months ago

I have added [merges.txt], please try again

PRANJALRANA11 commented 8 months ago

Thanks for that I have one more querry I have then used the model using pipeline to generate text but it is generating some random one

SteveKGYang commented 8 months ago

Firstly, make sure your code and padding sides are correct. As BART model is small in size, we expect its capability to be weaker. Please use the same prompt templates we provided when making inferences.

PRANJALRANA11 commented 8 months ago

Ok

On Wed, 27 Dec 2023, 2:49 pm SteveKGYang, @.***> wrote:

Firstly, make sure your code and padding sides are correct. As BART model is small in size, we expect it to be somehow overfitting. Please use the same prompt templates we provided when making inferences.

— Reply to this email directly, view it on GitHub https://github.com/SteveKGYang/MentalLLaMA/issues/4#issuecomment-1870109229, or unsubscribe https://github.com/notifications/unsubscribe-auth/A62HX4NG2S36QCC63J4SWU3YLPR2VAVCNFSM6AAAAABBD4T2GOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQNZQGEYDSMRSHE . You are receiving this because you authored the thread.Message ID: @.***>

PRANJALRANA11 commented 8 months ago

But I have checked out that templates too that are provided as examples on hugging faces and there also to output seems quiet off

On Wed, 27 Dec 2023, 2:51 pm pranjal rana, @.***> wrote:

Ok

On Wed, 27 Dec 2023, 2:49 pm SteveKGYang, @.***> wrote:

Firstly, make sure your code and padding sides are correct. As BART model is small in size, we expect it to be somehow overfitting. Please use the same prompt templates we provided when making inferences.

— Reply to this email directly, view it on GitHub https://github.com/SteveKGYang/MentalLLaMA/issues/4#issuecomment-1870109229, or unsubscribe https://github.com/notifications/unsubscribe-auth/A62HX4NG2S36QCC63J4SWU3YLPR2VAVCNFSM6AAAAABBD4T2GOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQNZQGEYDSMRSHE . You are receiving this because you authored the thread.Message ID: @.***>

TianlinZhang668 commented 8 months ago

Could you try to use your own machine to run the experiments rather than using huggingface website?