| import sys | |
| from io import BytesIO | |
| import base64 | |
| from PIL import Image | |
| import torch | |
| from transformers import CLIPProcessor, CLIPTextModel, CLIPVisionModelWithProjection | |
| device = torch.device('cuda' if torch.cuda.is_available() else 'cpu') | |
| class EndpointHandler(): | |
| def __init__(self, path=""): | |
| self.text_model = CLIPTextModel.from_pretrained("rbanfield/clip-vit-large-patch14") | |
| self.image_model = CLIPVisionModelWithProjection.from_pretrained("rbanfield/clip-vit-large-patch14") | |
| self.processor = CLIPProcessor.from_pretrained("rbanfield/clip-vit-large-patch14") | |
| def __call__(self, data): | |
| inputs = data.pop("inputs", None) | |
| print(inputs, file=sys.stderr) | |
| text_input = inputs["text"] if "text" in inputs else None | |
| image_input = inputs["image"] if "image" in inputs else None | |
| if text_input: | |
| print("in text mode", file=sys.stderr) | |
| print(text_input, file=sys.stderr) | |
| processor = self.processor(text=text_input, return_tensors="pt", padding=True) | |
| with torch.no_grad(): | |
| return self.text_model(**processor).pooler_output.tolist() | |
| elif image_input: | |
| print("in image mode", file=sys.stderr) | |
| print(image_input, file=sys.stderr) | |
| image = Image.open(BytesIO(base64.b64decode(image_input))) | |
| processor = self.processor(images=image, return_tensors="pt") | |
| with torch.no_grad(): | |
| return self.image_model(**processor).image_embeds.tolist() | |
| else: | |
| return None |