Hour
Hello,
Thank you for this version and for sharing ! It works very well :)
Would it be complicated to add hour recognition ?
Thx in advance for the answer,
Sanok
Hello,
Thank you for the good words :)
Actually it wouldn't be too complicated with the correct dataset which unfortunately I don't have.
Maybe you could try using one of the new LLM to generate enough examples and then train a camembert model on it.
Training part is actually not very complicated with the tools available in transformers library.
Thanks,
Jean-Baptiste
Thank you very much for your answer ! I will try to train one then :)
Sanok
Just a small question as I'm not very expert on the matter : is it better to train your model or to start with a less trained camembert model ?
Thanks in advance
This is actually a good question, I am not sure if there is really a best practice. But for sure if you add a new class, I don't think you have a choice than starting from an "untrained" camembert model.
And even if not I would say that it's probably best to start from scratch because the training I did might have move the weights in a direction that are not profitable for your class. But that's really just my intuition, probably best to try both approaches and see what perform best.
Please let me know what worked best for you, I would be curious to know.
Jean-Baptiste
Thank you again for your answer. I'll try and let you know.
Asanosuke