Views
No views yet
TOK → use <s0><s1> in your prompt1from diffusers import AutoPipelineForText2Image
2import torch
3from huggingface_hub import hf_hub_download
4from safetensors.torch import load_file
5
6pipeline = AutoPipelineForText2Image.from_pretrained('stabilityai/stable-diffusion-xl-base-1.0', torch_dtype=torch.float16).to('cuda')
7pipeline.load_lora_weights('multimodalart/polipure-repeats-3-steps-1200-rank64-with-prior', weight_name='pytorch_lora_weights.safetensors')
8embedding_path = hf_hub_download(repo_id='multimodalart/polipure-repeats-3-steps-1200-rank64-with-prior', filename="embeddings.safetensors", repo_type="model")
9state_dict = load_file(embedding_path)
10pipeline.load_textual_inversion(state_dict["clip_l"], token=["<s0>", "<s1>"], text_encoder=pipe.text_encoder, tokenizer=pipe.tokenizer)
11pipeline.load_textual_inversion(state_dict["clip_g"], token=["<s0>", "<s1>"], text_encoder=pipe.text_encoder_2, tokenizer=pipe.tokenizer_2)
12
13image = pipeline('<s0><s1>').images[0]