Make text finetuning version of Trompt, just like "transformers_text.py" for FTtransformer?

[hohoCode]

Can you also add an example of using text finetuning (better with LORA) version of Trompt, just like "transformers_text.py" for FTtransformer?

Better to support larger models LORA finetuning such as LLMs.

This will be very useful to compare FTtransformer and Trompt after text finetuned embeddings.

Thanks!

[zechengz]

Thanks for creating this issue. We will look into this soon.

[zechengz]

Hi @hohoCode , thanks for the feedback again.
Do you want to make a contribution to enable passing stype_encoder_dict, such as the one in the FTTransformer

pytorch-frame/torch_frame/nn/models/ft_transformer.py

Lines 58 to 59 in 4a6eeed

	col_names_dict: dict[torch_frame.stype, list[str]],
	stype_encoder_dict: dict[torch_frame.stype, StypeEncoder]

to the Trompt (then it can use the text embeddings)?

pytorch-frame/torch_frame/nn/models/trompt.py

Lines 73 to 75 in 4a6eeed

	self.encoders.append(
	StypeWiseFeatureEncoder(
	out_channels=channels,

Please let me know if you want to make a contribution. Thanks.

[weihua916]

@hohoCode Thanks for your suggestion! I agree it is interesting. Please feel modify the example code, experiment yourself, and let us know how Trompt performs!

[weihua916]

Converting this into discussion.

[hohoCode]

Thanks a lot for your response. I am totally new to the library, it will probably take me quite a while to understand the great work! So at this moment may not be able to do things here unfortunately. But I am highly interested and will keep following your exciting work along the journey!

Thanks a lot. Look forward to more great news from you all!