HF hosted interface API for a finetuned model with LoRA

Hi All,

I need a help. When I fine-tuned a flan-t5 model with lora, I got adapter_config.json and adaptor_model.bin instead of config.json and pytorch_model.bin. I used trainer pushing to hub. And if I am understanding correctly, it seems that the hf interface API doesn’t real the adapter_config because it’s asking for config.json.

How can I make this work? The model works fine generating outputs in notebook. Any idea to fix it?

Got an answer. According to HF, “Inference API does not yet support adapter-transformers models for this pipeline type.”