search

tencent-ailab / IP-Adapter

The image prompt adapter is designed to enable a pretrained text-to-image diffusion model to generate images with image prompt.
Apache License 2.0
5.32k stars 337 forks source link

train faceid with/without text? #413

Open rybinsky opened 3 months ago

rybinsky commented 3 months ago

Hi, a saw next code in MyDataset in file tutorial_train_faceid.py: self.data = json.load(open(json_file)) # list of dict: [{"image_file": "1.png", "id_embed_file": "faceid.bin"}] There is no field 'text', but in __get_item__ method of this Dataset a see: item = self.data[idx] text = item["text"] So, should a have a text caption for each image in dataset or no? What should be written there?

alexblattner commented 1 month ago

they messed up, yes. just have "text" in json