stable-diffusion-webui/modules/sd_hijack_xlmr.py

35 lines
1.5 KiB
Python
Raw Normal View History

2022-12-31 08:06:35 -07:00
import open_clip.tokenizer
import torch
from modules import sd_hijack_clip, devices
from modules.shared import opts
class FrozenXLMREmbedderWithCustomWords(sd_hijack_clip.FrozenCLIPEmbedderWithCustomWords):
def __init__(self, wrapped, hijack):
super().__init__(wrapped, hijack)
self.id_start = wrapped.config.bos_token_id
self.id_end = wrapped.config.eos_token_id
self.id_pad = wrapped.config.pad_token_id
self.comma_token = self.tokenizer.get_vocab().get(',', None) # alt diffusion doesn't have </w> bits for comma
def encode_with_transformers(self, tokens):
# there's no CLIP Skip here because all hidden layers have size of 1024 and the last one uses a
# trained layer to transform those 1024 into 768 for unet; so you can't choose which transformer
# layer to work with - you have to use the last
attention_mask = (tokens != self.id_pad).to(device=tokens.device, dtype=torch.int64)
features = self.wrapped(input_ids=tokens, attention_mask=attention_mask)
z = features['projection_state']
return z
def encode_embedding_init_text(self, init_text, nvpt):
embedding_layer = self.wrapped.roberta.embeddings
ids = self.wrapped.tokenizer(init_text, max_length=nvpt, return_tensors="pt", add_special_tokens=False)["input_ids"]
embedded = embedding_layer.token_embedding.wrapped(ids.to(devices.device)).squeeze(0)
return embedded