(
self,
image_size,
in_channels,
model_channels,
out_channels,
num_res_blocks,
attention_resolutions,
dropout=0,
channel_mult=(1, 2, 4, 8),
conv_resample=True,
dims=2,
num_classes=None,
use_checkpoint=False,
use_fp16=False,
num_heads=-1,
num_head_channels=-1,
num_heads_upsample=-1,
use_scale_shift_norm=False,
resblock_updown=False,
use_new_attention_order=False,
use_spatial_transformer=False, # custom transformer support
transformer_depth=1, # custom transformer support
context_dim=None, # custom transformer support
n_embed=None, # custom support for prediction of discrete ids into codebook of first stage vq model
legacy=True,
disable_self_attentions=None,
num_attention_blocks=None,
disable_middle_self_attn=False,
use_linear_in_transformer=False,
)
| 442 | """ |
| 443 | |
| 444 | def __init__( |
| 445 | self, |
| 446 | image_size, |
| 447 | in_channels, |
| 448 | model_channels, |
| 449 | out_channels, |
| 450 | num_res_blocks, |
| 451 | attention_resolutions, |
| 452 | dropout=0, |
| 453 | channel_mult=(1, 2, 4, 8), |
| 454 | conv_resample=True, |
| 455 | dims=2, |
| 456 | num_classes=None, |
| 457 | use_checkpoint=False, |
| 458 | use_fp16=False, |
| 459 | num_heads=-1, |
| 460 | num_head_channels=-1, |
| 461 | num_heads_upsample=-1, |
| 462 | use_scale_shift_norm=False, |
| 463 | resblock_updown=False, |
| 464 | use_new_attention_order=False, |
| 465 | use_spatial_transformer=False, # custom transformer support |
| 466 | transformer_depth=1, # custom transformer support |
| 467 | context_dim=None, # custom transformer support |
| 468 | n_embed=None, # custom support for prediction of discrete ids into codebook of first stage vq model |
| 469 | legacy=True, |
| 470 | disable_self_attentions=None, |
| 471 | num_attention_blocks=None, |
| 472 | disable_middle_self_attn=False, |
| 473 | use_linear_in_transformer=False, |
| 474 | ): |
| 475 | super().__init__() |
| 476 | if use_spatial_transformer: |
| 477 | assert context_dim is not None, 'Fool!! You forgot to include the dimension of your cross-attention conditioning...' |
| 478 | |
| 479 | if context_dim is not None: |
| 480 | assert use_spatial_transformer, 'Fool!! You forgot to use the spatial transformer for your cross-attention conditioning...' |
| 481 | from omegaconf.listconfig import ListConfig |
| 482 | if type(context_dim) == ListConfig: |
| 483 | context_dim = list(context_dim) |
| 484 | |
| 485 | if num_heads_upsample == -1: |
| 486 | num_heads_upsample = num_heads |
| 487 | |
| 488 | if num_heads == -1: |
| 489 | assert num_head_channels != -1, 'Either num_heads or num_head_channels has to be set' |
| 490 | |
| 491 | if num_head_channels == -1: |
| 492 | assert num_heads != -1, 'Either num_heads or num_head_channels has to be set' |
| 493 | |
| 494 | self.image_size = image_size |
| 495 | self.in_channels = in_channels |
| 496 | self.model_channels = model_channels |
| 497 | self.out_channels = out_channels |
| 498 | if isinstance(num_res_blocks, int): |
| 499 | self.num_res_blocks = len(channel_mult) * [num_res_blocks] |
| 500 | else: |
| 501 | if len(num_res_blocks) != len(channel_mult): |
nothing calls this directly
no test coverage detected