| 563 | return [49406] + bpe_tokens + [49407] * (77 - len(bpe_tokens) - 1) |
| 564 | |
| 565 | class StableDiffusion: |
| 566 | def __init__(self): |
| 567 | self.alphas_cumprod = Tensor.empty(1000) |
| 568 | self.model = namedtuple("DiffusionModel", ["diffusion_model"])(diffusion_model = UNetModel()) |
| 569 | self.first_stage_model = AutoencoderKL() |
| 570 | self.cond_stage_model = namedtuple("CondStageModel", ["transformer"])(transformer = namedtuple("Transformer", ["text_model"])(text_model = CLIPTextTransformer())) |
| 571 | |
| 572 | # TODO: make __call__ run the model |
| 573 | |
| 574 | # ** ldm.models.autoencoder.AutoencoderKL (done!) |
| 575 | # 3x512x512 <--> 4x64x64 (16384) |