Unable to load model from HF

[levente-murgas]

Hi!

First of all, amazing work! I'm trying to load the model with the pretrained weights from HF, but I'm receiving an error, while doing so.
My first attempt:
model = AutoModelForSeq2SeqLM.from_pretrained("flbbb/locost-gsg-pretrained")
And my second attempt:
pipe = pipeline("text2text-generation", model="flbbb/locost-gsg-pretrained")
Both resulting in the error:
ValueError: Unrecognized model in flbbb/locost-gsg-pretrained. Should have a `model_type` key in its config.json, or contain one of the following strings in its name: albert, align, altclip, audio-spectrogram-transformer, ...

Is there a chance that I can get some guidance on how to properly load your model?

Thanks!

[flbbb]

Hi and thanks for your message!
We did not fully integrated our model with the HF ecosystem. Which means you need to use our own class and call .pretrained on this class, instead of using the AutoModel or the pipeline.

Install following the README then:

from models import LOCOSTForConditionalGeneration
model = LOCOSTForConditionalGeneration.from_pretrained("flbbb/locost-gsg-pretrained")