hugging-apps/echo-memory
0
1import torch, math2from einops import rearrange, repeat3from .sd_unet import Timesteps, PushBlock, PopBlock, Attention, GEGLU, ResnetBlock, AttentionBlock, DownSampler, UpSampler4 5 6class TemporalResnetBlock(torch.nn.Module):7 def __init__(self, in_channels, out_channels, temb_channels=None, groups=32, eps=1e-5):8 super().__init__()9 self.norm1 = torch.nn.GroupNorm(num_groups=groups, num_channels=in_channels, eps=eps, affine=True)10 self.conv1 = torch.nn.Conv3d(in_channels, out_channels, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0))11 if temb_channels is not None:12 self.time_emb_proj = torch.nn.Linear(temb_channels, out_channels)13 self.norm2 = torch.nn.GroupNorm(num_groups=groups, num_channels=out_channels, eps=eps, affine=True)14 self.conv2 = torch.nn.Conv3d(out_channels, out_channels, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0))15 self.nonlinearity = torch.nn.SiLU()16 self.conv_shortcut = None17 if in_channels != out_channels:18 self.conv_shortcut = torch.nn.Conv3d(in_channels, out_channels, kernel_size=1, stride=1, padding=0, bias=True)19 20 def forward(self, hidden_states, time_emb, text_emb, res_stack, **kwargs):21 x = rearrange(hidden_states, "f c h w -> 1 c f h w")22 x = self.norm1(x)23 x = self.nonlinearity(x)24 x = self.conv1(x)25 if time_emb is not None:26 emb = self.nonlinearity(time_emb)27 emb = self.time_emb_proj(emb)28 emb = repeat(emb, "b c -> b c f 1 1", f=hidden_states.shape[0])29 x = x + emb30 x = self.norm2(x)31 x = self.nonlinearity(x)32 x = self.conv2(x)33 if self.conv_shortcut is not None:34 hidden_states = self.conv_shortcut(hidden_states)35 x = rearrange(x[0], "c f h w -> f c h w")36 hidden_states = hidden_states + x37 return hidden_states, time_emb, text_emb, res_stack38 39 40def get_timestep_embedding(41 timesteps: torch.Tensor,42 embedding_dim: int,43 flip_sin_to_cos: bool = False,44 downscale_freq_shift: float = 1,45 scale: float = 1,46 max_period: int = 10000,47 computation_device = None,48):49 """50 This matches the implementation in Denoising Diffusion Probabilistic Models: Create sinusoidal timestep embeddings.51 52 :param timesteps: a 1-D Tensor of N indices, one per batch element.53 These may be fractional.54 :param embedding_dim: the dimension of the output. :param max_period: controls the minimum frequency of the55 embeddings. :return: an [N x dim] Tensor of positional embeddings.56 """57 assert len(timesteps.shape) == 1, "Timesteps should be a 1d-array"58 59 half_dim = embedding_dim // 260 exponent = -math.log(max_period) * torch.arange(61 start=0, end=half_dim, dtype=torch.float32, device=timesteps.device if computation_device is None else computation_device62 )63 exponent = exponent / (half_dim - downscale_freq_shift)64 65 emb = torch.exp(exponent).to(timesteps.device)66 emb = timesteps[:, None].float() * emb[None, :]67 68 # scale embeddings69 emb = scale * emb70 71 # concat sine and cosine embeddings72 emb = torch.cat([torch.sin(emb), torch.cos(emb)], dim=-1)73 74 # flip sine and cosine embeddings75 if flip_sin_to_cos:76 emb = torch.cat([emb[:, half_dim:], emb[:, :half_dim]], dim=-1)77 78 # zero pad79 if embedding_dim % 2 == 1:80 emb = torch.nn.functional.pad(emb, (0, 1, 0, 0))81 return emb82 83 84class TemporalTimesteps(torch.nn.Module):85 def __init__(self, num_channels: int, flip_sin_to_cos: bool, downscale_freq_shift: float, computation_device = None):86 super().__init__()87 self.num_channels = num_channels88 self.flip_sin_to_cos = flip_sin_to_cos89 self.downscale_freq_shift = downscale_freq_shift90 self.computation_device = computation_device91 92 def forward(self, timesteps):93 t_emb = get_timestep_embedding(94 timesteps,95 self.num_channels,96 flip_sin_to_cos=self.flip_sin_to_cos,97 downscale_freq_shift=self.downscale_freq_shift,98 computation_device=self.computation_device,99 )100 return t_emb101 102 103class TrainableTemporalTimesteps(torch.nn.Module):104 def __init__(self, num_channels: int, flip_sin_to_cos: bool, downscale_freq_shift: float, num_frames: int):105 super().__init__()106 timesteps = PositionalID()(num_frames)107 embeddings = get_timestep_embedding(timesteps, num_channels, flip_sin_to_cos, downscale_freq_shift)108 self.embeddings = torch.nn.Parameter(embeddings)109 110 def forward(self, timesteps):111 t_emb = self.embeddings[timesteps]112 return t_emb113 114 115class PositionalID(torch.nn.Module):116 def __init__(self, max_id=25, repeat_length=20):117 super().__init__()118 self.max_id = max_id119 self.repeat_length = repeat_length120 121 def frame_id_to_position_id(self, frame_id):122 if frame_id < self.max_id:123 position_id = frame_id124 else:125 position_id = (frame_id - self.max_id) % (self.repeat_length * 2)126 if position_id < self.repeat_length:127 position_id = self.max_id - 2 - position_id128 else:129 position_id = self.max_id - 2 * self.repeat_length + position_id130 return position_id131 132 def forward(self, num_frames, pivot_frame_id=0):133 position_ids = [self.frame_id_to_position_id(abs(i-pivot_frame_id)) for i in range(num_frames)]134 position_ids = torch.IntTensor(position_ids)135 return position_ids136 137 138class TemporalAttentionBlock(torch.nn.Module):139 140 def __init__(self, num_attention_heads, attention_head_dim, in_channels, cross_attention_dim=None, add_positional_conv=None):141 super().__init__()142 143 self.positional_embedding_proj = torch.nn.Sequential(144 torch.nn.Linear(in_channels, in_channels * 4),145 torch.nn.SiLU(),146 torch.nn.Linear(in_channels * 4, in_channels)147 )148 if add_positional_conv is not None:149 self.positional_embedding = TrainableTemporalTimesteps(in_channels, True, 0, add_positional_conv)150 self.positional_conv = torch.nn.Conv3d(in_channels, in_channels, kernel_size=3, padding=1, padding_mode="reflect")151 else:152 self.positional_embedding = TemporalTimesteps(in_channels, True, 0)153 self.positional_conv = None154 155 self.norm_in = torch.nn.LayerNorm(in_channels)156 self.act_fn_in = GEGLU(in_channels, in_channels * 4)157 self.ff_in = torch.nn.Linear(in_channels * 4, in_channels)158 159 self.norm1 = torch.nn.LayerNorm(in_channels)160 self.attn1 = Attention(161 q_dim=in_channels,162 num_heads=num_attention_heads,163 head_dim=attention_head_dim,164 bias_out=True165 )166 167 self.norm2 = torch.nn.LayerNorm(in_channels)168 self.attn2 = Attention(169 q_dim=in_channels,170 kv_dim=cross_attention_dim,171 num_heads=num_attention_heads,172 head_dim=attention_head_dim,173 bias_out=True174 )175 176 self.norm_out = torch.nn.LayerNorm(in_channels)177 self.act_fn_out = GEGLU(in_channels, in_channels * 4)178 self.ff_out = torch.nn.Linear(in_channels * 4, in_channels)179 180 def forward(self, hidden_states, time_emb, text_emb, res_stack, **kwargs):181 182 batch, inner_dim, height, width = hidden_states.shape183 pos_emb = torch.arange(batch)184 pos_emb = self.positional_embedding(pos_emb).to(dtype=hidden_states.dtype, device=hidden_states.device)185 pos_emb = self.positional_embedding_proj(pos_emb)186 187 hidden_states = rearrange(hidden_states, "T C H W -> 1 C T H W") + rearrange(pos_emb, "T C -> 1 C T 1 1")188 if self.positional_conv is not None:189 hidden_states = self.positional_conv(hidden_states)190 hidden_states = rearrange(hidden_states[0], "C T H W -> (H W) T C")191 192 residual = hidden_states193 hidden_states = self.norm_in(hidden_states)194 hidden_states = self.act_fn_in(hidden_states)195 hidden_states = self.ff_in(hidden_states)196 hidden_states = hidden_states + residual197 198 norm_hidden_states = self.norm1(hidden_states)199 attn_output = self.attn1(norm_hidden_states, encoder_hidden_states=None)200 hidden_states = attn_output + hidden_states201 202 norm_hidden_states = self.norm2(hidden_states)203 attn_output = self.attn2(norm_hidden_states, encoder_hidden_states=text_emb.repeat(height * width, 1))204 hidden_states = attn_output + hidden_states205 206 residual = hidden_states207 hidden_states = self.norm_out(hidden_states)208 hidden_states = self.act_fn_out(hidden_states)209 hidden_states = self.ff_out(hidden_states)210 hidden_states = hidden_states + residual211 212 hidden_states = hidden_states.reshape(height, width, batch, inner_dim).permute(2, 3, 0, 1)213 214 return hidden_states, time_emb, text_emb, res_stack215 216 217class PopMixBlock(torch.nn.Module):218 def __init__(self, in_channels=None):219 super().__init__()220 self.mix_factor = torch.nn.Parameter(torch.Tensor([0.5]))221 self.need_proj = in_channels is not None222 if self.need_proj:223 self.proj = torch.nn.Linear(in_channels, in_channels)224 225 def forward(self, hidden_states, time_emb, text_emb, res_stack, **kwargs):226 res_hidden_states = res_stack.pop()227 alpha = torch.sigmoid(self.mix_factor)228 hidden_states = alpha * res_hidden_states + (1 - alpha) * hidden_states229 if self.need_proj:230 hidden_states = hidden_states.permute(0, 2, 3, 1)231 hidden_states = self.proj(hidden_states)232 hidden_states = hidden_states.permute(0, 3, 1, 2)233 res_hidden_states = res_stack.pop()234 hidden_states = hidden_states + res_hidden_states235 return hidden_states, time_emb, text_emb, res_stack236 237 238class SVDUNet(torch.nn.Module):239 def __init__(self, add_positional_conv=None):240 super().__init__()241 self.time_proj = Timesteps(320)242 self.time_embedding = torch.nn.Sequential(243 torch.nn.Linear(320, 1280),244 torch.nn.SiLU(),245 torch.nn.Linear(1280, 1280)246 )247 self.add_time_proj = Timesteps(256)248 self.add_time_embedding = torch.nn.Sequential(249 torch.nn.Linear(768, 1280),250 torch.nn.SiLU(),251 torch.nn.Linear(1280, 1280)252 )253 self.conv_in = torch.nn.Conv2d(8, 320, kernel_size=3, padding=1)254 255 self.blocks = torch.nn.ModuleList([256 # CrossAttnDownBlockSpatioTemporal257 ResnetBlock(320, 320, 1280, eps=1e-6), PushBlock(), TemporalResnetBlock(320, 320, 1280, eps=1e-6), PopMixBlock(), PushBlock(),258 AttentionBlock(5, 64, 320, 1, 1024, need_proj_out=False), PushBlock(), TemporalAttentionBlock(5, 64, 320, 1024, add_positional_conv), PopMixBlock(320), PushBlock(),259 ResnetBlock(320, 320, 1280, eps=1e-6), PushBlock(), TemporalResnetBlock(320, 320, 1280, eps=1e-6), PopMixBlock(), PushBlock(),260 AttentionBlock(5, 64, 320, 1, 1024, need_proj_out=False), PushBlock(), TemporalAttentionBlock(5, 64, 320, 1024, add_positional_conv), PopMixBlock(320), PushBlock(),261 DownSampler(320), PushBlock(),262 # CrossAttnDownBlockSpatioTemporal263 ResnetBlock(320, 640, 1280, eps=1e-6), PushBlock(), TemporalResnetBlock(640, 640, 1280, eps=1e-6), PopMixBlock(), PushBlock(),264 AttentionBlock(10, 64, 640, 1, 1024, need_proj_out=False), PushBlock(), TemporalAttentionBlock(10, 64, 640, 1024, add_positional_conv), PopMixBlock(640), PushBlock(),265 ResnetBlock(640, 640, 1280, eps=1e-6), PushBlock(), TemporalResnetBlock(640, 640, 1280, eps=1e-6), PopMixBlock(), PushBlock(),266 AttentionBlock(10, 64, 640, 1, 1024, need_proj_out=False), PushBlock(), TemporalAttentionBlock(10, 64, 640, 1024, add_positional_conv), PopMixBlock(640), PushBlock(),267 DownSampler(640), PushBlock(),268 # CrossAttnDownBlockSpatioTemporal269 ResnetBlock(640, 1280, 1280, eps=1e-6), PushBlock(), TemporalResnetBlock(1280, 1280, 1280, eps=1e-6), PopMixBlock(), PushBlock(),270 AttentionBlock(20, 64, 1280, 1, 1024, need_proj_out=False), PushBlock(), TemporalAttentionBlock(20, 64, 1280, 1024, add_positional_conv), PopMixBlock(1280), PushBlock(),271 ResnetBlock(1280, 1280, 1280, eps=1e-6), PushBlock(), TemporalResnetBlock(1280, 1280, 1280, eps=1e-6), PopMixBlock(), PushBlock(),272 AttentionBlock(20, 64, 1280, 1, 1024, need_proj_out=False), PushBlock(), TemporalAttentionBlock(20, 64, 1280, 1024, add_positional_conv), PopMixBlock(1280), PushBlock(),273 DownSampler(1280), PushBlock(),274 # DownBlockSpatioTemporal275 ResnetBlock(1280, 1280, 1280, eps=1e-5), PushBlock(), TemporalResnetBlock(1280, 1280, 1280, eps=1e-5), PopMixBlock(), PushBlock(),276 ResnetBlock(1280, 1280, 1280, eps=1e-5), PushBlock(), TemporalResnetBlock(1280, 1280, 1280, eps=1e-5), PopMixBlock(), PushBlock(),277 # UNetMidBlockSpatioTemporal278 ResnetBlock(1280, 1280, 1280, eps=1e-5), PushBlock(), TemporalResnetBlock(1280, 1280, 1280, eps=1e-5), PopMixBlock(), PushBlock(),279 AttentionBlock(20, 64, 1280, 1, 1024, need_proj_out=False), PushBlock(), TemporalAttentionBlock(20, 64, 1280, 1024, add_positional_conv), PopMixBlock(1280),280 ResnetBlock(1280, 1280, 1280, eps=1e-5), PushBlock(), TemporalResnetBlock(1280, 1280, 1280, eps=1e-5), PopMixBlock(),281 # UpBlockSpatioTemporal282 PopBlock(), ResnetBlock(2560, 1280, 1280, eps=1e-6), PushBlock(), TemporalResnetBlock(1280, 1280, 1280, eps=1e-5), PopMixBlock(),283 PopBlock(), ResnetBlock(2560, 1280, 1280, eps=1e-6), PushBlock(), TemporalResnetBlock(1280, 1280, 1280, eps=1e-5), PopMixBlock(),284 PopBlock(), ResnetBlock(2560, 1280, 1280, eps=1e-6), PushBlock(), TemporalResnetBlock(1280, 1280, 1280, eps=1e-5), PopMixBlock(),285 UpSampler(1280),286 # CrossAttnUpBlockSpatioTemporal287 PopBlock(), ResnetBlock(2560, 1280, 1280, eps=1e-6), PushBlock(), TemporalResnetBlock(1280, 1280, 1280, eps=1e-6), PopMixBlock(), PushBlock(),288 AttentionBlock(20, 64, 1280, 1, 1024, need_proj_out=False), PushBlock(), TemporalAttentionBlock(20, 64, 1280, 1024, add_positional_conv), PopMixBlock(1280),289 PopBlock(), ResnetBlock(2560, 1280, 1280, eps=1e-6), PushBlock(), TemporalResnetBlock(1280, 1280, 1280, eps=1e-6), PopMixBlock(), PushBlock(),290 AttentionBlock(20, 64, 1280, 1, 1024, need_proj_out=False), PushBlock(), TemporalAttentionBlock(20, 64, 1280, 1024, add_positional_conv), PopMixBlock(1280),291 PopBlock(), ResnetBlock(1920, 1280, 1280, eps=1e-6), PushBlock(), TemporalResnetBlock(1280, 1280, 1280, eps=1e-6), PopMixBlock(), PushBlock(),292 AttentionBlock(20, 64, 1280, 1, 1024, need_proj_out=False), PushBlock(), TemporalAttentionBlock(20, 64, 1280, 1024, add_positional_conv), PopMixBlock(1280),293 UpSampler(1280),294 # CrossAttnUpBlockSpatioTemporal295 PopBlock(), ResnetBlock(1920, 640, 1280, eps=1e-6), PushBlock(), TemporalResnetBlock(640, 640, 1280, eps=1e-6), PopMixBlock(), PushBlock(),296 AttentionBlock(10, 64, 640, 1, 1024, need_proj_out=False), PushBlock(), TemporalAttentionBlock(10, 64, 640, 1024, add_positional_conv), PopMixBlock(640),297 PopBlock(), ResnetBlock(1280, 640, 1280, eps=1e-6), PushBlock(), TemporalResnetBlock(640, 640, 1280, eps=1e-6), PopMixBlock(), PushBlock(),298 AttentionBlock(10, 64, 640, 1, 1024, need_proj_out=False), PushBlock(), TemporalAttentionBlock(10, 64, 640, 1024, add_positional_conv), PopMixBlock(640),299 PopBlock(), ResnetBlock(960, 640, 1280, eps=1e-6), PushBlock(), TemporalResnetBlock(640, 640, 1280, eps=1e-6), PopMixBlock(), PushBlock(),300 AttentionBlock(10, 64, 640, 1, 1024, need_proj_out=False), PushBlock(), TemporalAttentionBlock(10, 64, 640, 1024, add_positional_conv), PopMixBlock(640),301 UpSampler(640),302 # CrossAttnUpBlockSpatioTemporal303 PopBlock(), ResnetBlock(960, 320, 1280, eps=1e-6), PushBlock(), TemporalResnetBlock(320, 320, 1280, eps=1e-6), PopMixBlock(), PushBlock(),304 AttentionBlock(5, 64, 320, 1, 1024, need_proj_out=False), PushBlock(), TemporalAttentionBlock(5, 64, 320, 1024, add_positional_conv), PopMixBlock(320),305 PopBlock(), ResnetBlock(640, 320, 1280, eps=1e-6), PushBlock(), TemporalResnetBlock(320, 320, 1280, eps=1e-6), PopMixBlock(), PushBlock(),306 AttentionBlock(5, 64, 320, 1, 1024, need_proj_out=False), PushBlock(), TemporalAttentionBlock(5, 64, 320, 1024, add_positional_conv), PopMixBlock(320),307 PopBlock(), ResnetBlock(640, 320, 1280, eps=1e-6), PushBlock(), TemporalResnetBlock(320, 320, 1280, eps=1e-6), PopMixBlock(), PushBlock(),308 AttentionBlock(5, 64, 320, 1, 1024, need_proj_out=False), PushBlock(), TemporalAttentionBlock(5, 64, 320, 1024, add_positional_conv), PopMixBlock(320),309 ])310 311 self.conv_norm_out = torch.nn.GroupNorm(32, 320, eps=1e-05, affine=True)312 self.conv_act = torch.nn.SiLU()313 self.conv_out = torch.nn.Conv2d(320, 4, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))314 315 316 def build_mask(self, data, is_bound):317 T, C, H, W = data.shape318 t = repeat(torch.arange(T), "T -> T H W", T=T, H=H, W=W)319 h = repeat(torch.arange(H), "H -> T H W", T=T, H=H, W=W)320 w = repeat(torch.arange(W), "W -> T H W", T=T, H=H, W=W)321 border_width = (T + H + W) // 6322 pad = torch.ones_like(t) * border_width323 mask = torch.stack([324 pad if is_bound[0] else t + 1,325 pad if is_bound[1] else T - t,326 pad if is_bound[2] else h + 1,327 pad if is_bound[3] else H - h,328 pad if is_bound[4] else w + 1,329 pad if is_bound[5] else W - w330 ]).min(dim=0).values331 mask = mask.clip(1, border_width)332 mask = (mask / border_width).to(dtype=data.dtype, device=data.device)333 mask = rearrange(mask, "T H W -> T 1 H W")334 return mask335 336 337 def tiled_forward(338 self, sample, timestep, encoder_hidden_states, add_time_id,339 batch_time=25, batch_height=128, batch_width=128,340 stride_time=5, stride_height=64, stride_width=64,341 progress_bar=lambda x:x342 ):343 data_device = sample.device344 computation_device = self.conv_in.weight.device345 torch_dtype = sample.dtype346 T, C, H, W = sample.shape347 348 weight = torch.zeros((T, 1, H, W), dtype=torch_dtype, device=data_device)349 values = torch.zeros((T, 4, H, W), dtype=torch_dtype, device=data_device)350 351 # Split tasks352 tasks = []353 for t in range(0, T, stride_time):354 for h in range(0, H, stride_height):355 for w in range(0, W, stride_width):356 if (t-stride_time >= 0 and t-stride_time+batch_time >= T)\357 or (h-stride_height >= 0 and h-stride_height+batch_height >= H)\358 or (w-stride_width >= 0 and w-stride_width+batch_width >= W):359 continue360 tasks.append((t, t+batch_time, h, h+batch_height, w, w+batch_width))361 362 # Run363 for tl, tr, hl, hr, wl, wr in progress_bar(tasks):364 sample_batch = sample[tl:tr, :, hl:hr, wl:wr].to(computation_device)365 sample_batch = self.forward(sample_batch, timestep, encoder_hidden_states, add_time_id).to(data_device)366 mask = self.build_mask(sample_batch, is_bound=(tl==0, tr>=T, hl==0, hr>=H, wl==0, wr>=W))367 values[tl:tr, :, hl:hr, wl:wr] += sample_batch * mask368 weight[tl:tr, :, hl:hr, wl:wr] += mask369 values /= weight370 return values371 372 373 def forward(self, sample, timestep, encoder_hidden_states, add_time_id, use_gradient_checkpointing=False, **kwargs):374 # 1. time375 timestep = torch.tensor((timestep,)).to(sample.device)376 t_emb = self.time_proj(timestep).to(sample.dtype)377 t_emb = self.time_embedding(t_emb)378 379 add_embeds = self.add_time_proj(add_time_id.flatten()).to(sample.dtype)380 add_embeds = add_embeds.reshape((-1, 768))381 add_embeds = self.add_time_embedding(add_embeds)382 383 time_emb = t_emb + add_embeds384 385 # 2. pre-process386 height, width = sample.shape[2], sample.shape[3]387 hidden_states = self.conv_in(sample)388 text_emb = encoder_hidden_states389 res_stack = [hidden_states]390 391 # 3. blocks392 def create_custom_forward(module):393 def custom_forward(*inputs):394 return module(*inputs)395 return custom_forward396 for i, block in enumerate(self.blocks):397 if self.training and use_gradient_checkpointing and not (isinstance(block, PushBlock) or isinstance(block, PopBlock) or isinstance(block, PopMixBlock)):398 hidden_states, time_emb, text_emb, res_stack = torch.utils.checkpoint.checkpoint(399 create_custom_forward(block),400 hidden_states, time_emb, text_emb, res_stack,401 use_reentrant=False,402 )403 else:404 hidden_states, time_emb, text_emb, res_stack = block(hidden_states, time_emb, text_emb, res_stack)405 406 # 4. output407 hidden_states = self.conv_norm_out(hidden_states)408 hidden_states = self.conv_act(hidden_states)409 hidden_states = self.conv_out(hidden_states)410 411 return hidden_states412 413 @staticmethod414 def state_dict_converter():415 return SVDUNetStateDictConverter()416 417 418 419class SVDUNetStateDictConverter:420 def __init__(self):421 pass422 423 def get_block_name(self, names):424 if names[0] in ["down_blocks", "mid_block", "up_blocks"]:425 if names[4] in ["norm", "proj_in"]:426 return ".".join(names[:4] + ["transformer_blocks"])427 elif names[4] in ["time_pos_embed"]:428 return ".".join(names[:4] + ["temporal_transformer_blocks"])429 elif names[4] in ["proj_out"]:430 return ".".join(names[:4] + ["time_mixer"])431 else:432 return ".".join(names[:5])433 return ""434 435 def from_diffusers(self, state_dict):436 rename_dict = {437 "time_embedding.linear_1": "time_embedding.0",438 "time_embedding.linear_2": "time_embedding.2",439 "add_embedding.linear_1": "add_time_embedding.0",440 "add_embedding.linear_2": "add_time_embedding.2",441 "conv_in": "conv_in",442 "conv_norm_out": "conv_norm_out",443 "conv_out": "conv_out",444 }445 blocks_rename_dict = [446 "down_blocks.0.resnets.0.spatial_res_block", None, "down_blocks.0.resnets.0.temporal_res_block", "down_blocks.0.resnets.0.time_mixer", None,447 "down_blocks.0.attentions.0.transformer_blocks", None, "down_blocks.0.attentions.0.temporal_transformer_blocks", "down_blocks.0.attentions.0.time_mixer", None,448 "down_blocks.0.resnets.1.spatial_res_block", None, "down_blocks.0.resnets.1.temporal_res_block", "down_blocks.0.resnets.1.time_mixer", None,449 "down_blocks.0.attentions.1.transformer_blocks", None, "down_blocks.0.attentions.1.temporal_transformer_blocks", "down_blocks.0.attentions.1.time_mixer", None,450 "down_blocks.0.downsamplers.0.conv", None,451 "down_blocks.1.resnets.0.spatial_res_block", None, "down_blocks.1.resnets.0.temporal_res_block", "down_blocks.1.resnets.0.time_mixer", None,452 "down_blocks.1.attentions.0.transformer_blocks", None, "down_blocks.1.attentions.0.temporal_transformer_blocks", "down_blocks.1.attentions.0.time_mixer", None,453 "down_blocks.1.resnets.1.spatial_res_block", None, "down_blocks.1.resnets.1.temporal_res_block", "down_blocks.1.resnets.1.time_mixer", None,454 "down_blocks.1.attentions.1.transformer_blocks", None, "down_blocks.1.attentions.1.temporal_transformer_blocks", "down_blocks.1.attentions.1.time_mixer", None,455 "down_blocks.1.downsamplers.0.conv", None,456 "down_blocks.2.resnets.0.spatial_res_block", None, "down_blocks.2.resnets.0.temporal_res_block", "down_blocks.2.resnets.0.time_mixer", None,457 "down_blocks.2.attentions.0.transformer_blocks", None, "down_blocks.2.attentions.0.temporal_transformer_blocks", "down_blocks.2.attentions.0.time_mixer", None,458 "down_blocks.2.resnets.1.spatial_res_block", None, "down_blocks.2.resnets.1.temporal_res_block", "down_blocks.2.resnets.1.time_mixer", None,459 "down_blocks.2.attentions.1.transformer_blocks", None, "down_blocks.2.attentions.1.temporal_transformer_blocks", "down_blocks.2.attentions.1.time_mixer", None,460 "down_blocks.2.downsamplers.0.conv", None,461 "down_blocks.3.resnets.0.spatial_res_block", None, "down_blocks.3.resnets.0.temporal_res_block", "down_blocks.3.resnets.0.time_mixer", None,462 "down_blocks.3.resnets.1.spatial_res_block", None, "down_blocks.3.resnets.1.temporal_res_block", "down_blocks.3.resnets.1.time_mixer", None,463 "mid_block.mid_block.resnets.0.spatial_res_block", None, "mid_block.mid_block.resnets.0.temporal_res_block", "mid_block.mid_block.resnets.0.time_mixer", None,464 "mid_block.mid_block.attentions.0.transformer_blocks", None, "mid_block.mid_block.attentions.0.temporal_transformer_blocks", "mid_block.mid_block.attentions.0.time_mixer",465 "mid_block.mid_block.resnets.1.spatial_res_block", None, "mid_block.mid_block.resnets.1.temporal_res_block", "mid_block.mid_block.resnets.1.time_mixer",466 None, "up_blocks.0.resnets.0.spatial_res_block", None, "up_blocks.0.resnets.0.temporal_res_block", "up_blocks.0.resnets.0.time_mixer",467 None, "up_blocks.0.resnets.1.spatial_res_block", None, "up_blocks.0.resnets.1.temporal_res_block", "up_blocks.0.resnets.1.time_mixer",468 None, "up_blocks.0.resnets.2.spatial_res_block", None, "up_blocks.0.resnets.2.temporal_res_block", "up_blocks.0.resnets.2.time_mixer",469 "up_blocks.0.upsamplers.0.conv",470 None, "up_blocks.1.resnets.0.spatial_res_block", None, "up_blocks.1.resnets.0.temporal_res_block", "up_blocks.1.resnets.0.time_mixer", None,471 "up_blocks.1.attentions.0.transformer_blocks", None, "up_blocks.1.attentions.0.temporal_transformer_blocks", "up_blocks.1.attentions.0.time_mixer",472 None, "up_blocks.1.resnets.1.spatial_res_block", None, "up_blocks.1.resnets.1.temporal_res_block", "up_blocks.1.resnets.1.time_mixer", None,473 "up_blocks.1.attentions.1.transformer_blocks", None, "up_blocks.1.attentions.1.temporal_transformer_blocks", "up_blocks.1.attentions.1.time_mixer",474 None, "up_blocks.1.resnets.2.spatial_res_block", None, "up_blocks.1.resnets.2.temporal_res_block", "up_blocks.1.resnets.2.time_mixer", None,475 "up_blocks.1.attentions.2.transformer_blocks", None, "up_blocks.1.attentions.2.temporal_transformer_blocks", "up_blocks.1.attentions.2.time_mixer",476 "up_blocks.1.upsamplers.0.conv",477 None, "up_blocks.2.resnets.0.spatial_res_block", None, "up_blocks.2.resnets.0.temporal_res_block", "up_blocks.2.resnets.0.time_mixer", None,478 "up_blocks.2.attentions.0.transformer_blocks", None, "up_blocks.2.attentions.0.temporal_transformer_blocks", "up_blocks.2.attentions.0.time_mixer",479 None, "up_blocks.2.resnets.1.spatial_res_block", None, "up_blocks.2.resnets.1.temporal_res_block", "up_blocks.2.resnets.1.time_mixer", None,480 "up_blocks.2.attentions.1.transformer_blocks", None, "up_blocks.2.attentions.1.temporal_transformer_blocks", "up_blocks.2.attentions.1.time_mixer",481 None, "up_blocks.2.resnets.2.spatial_res_block", None, "up_blocks.2.resnets.2.temporal_res_block", "up_blocks.2.resnets.2.time_mixer", None,482 "up_blocks.2.attentions.2.transformer_blocks", None, "up_blocks.2.attentions.2.temporal_transformer_blocks", "up_blocks.2.attentions.2.time_mixer",483 "up_blocks.2.upsamplers.0.conv",484 None, "up_blocks.3.resnets.0.spatial_res_block", None, "up_blocks.3.resnets.0.temporal_res_block", "up_blocks.3.resnets.0.time_mixer", None,485 "up_blocks.3.attentions.0.transformer_blocks", None, "up_blocks.3.attentions.0.temporal_transformer_blocks", "up_blocks.3.attentions.0.time_mixer",486 None, "up_blocks.3.resnets.1.spatial_res_block", None, "up_blocks.3.resnets.1.temporal_res_block", "up_blocks.3.resnets.1.time_mixer", None,487 "up_blocks.3.attentions.1.transformer_blocks", None, "up_blocks.3.attentions.1.temporal_transformer_blocks", "up_blocks.3.attentions.1.time_mixer",488 None, "up_blocks.3.resnets.2.spatial_res_block", None, "up_blocks.3.resnets.2.temporal_res_block", "up_blocks.3.resnets.2.time_mixer", None,489 "up_blocks.3.attentions.2.transformer_blocks", None, "up_blocks.3.attentions.2.temporal_transformer_blocks", "up_blocks.3.attentions.2.time_mixer",490 ]491 blocks_rename_dict = {i:j for j,i in enumerate(blocks_rename_dict) if i is not None}492 state_dict_ = {}493 for name, param in sorted(state_dict.items()):494 names = name.split(".")495 if names[0] == "mid_block":496 names = ["mid_block"] + names497 if names[-1] in ["weight", "bias"]:498 name_prefix = ".".join(names[:-1])499 if name_prefix in rename_dict:500 state_dict_[rename_dict[name_prefix] + "." + names[-1]] = param501 else:502 block_name = self.get_block_name(names)503 if "resnets" in block_name and block_name in blocks_rename_dict:504 rename = ".".join(["blocks", str(blocks_rename_dict[block_name])] + names[5:])505 state_dict_[rename] = param506 elif ("downsamplers" in block_name or "upsamplers" in block_name) and block_name in blocks_rename_dict:507 rename = ".".join(["blocks", str(blocks_rename_dict[block_name])] + names[-2:])508 state_dict_[rename] = param509 elif "attentions" in block_name and block_name in blocks_rename_dict:510 attention_id = names[5]511 if "transformer_blocks" in names:512 suffix_dict = {513 "attn1.to_out.0": "attn1.to_out",514 "attn2.to_out.0": "attn2.to_out",515 "ff.net.0.proj": "act_fn.proj",516 "ff.net.2": "ff",517 }518 suffix = ".".join(names[6:-1])519 suffix = suffix_dict.get(suffix, suffix)520 rename = ".".join(["blocks", str(blocks_rename_dict[block_name]), "transformer_blocks", attention_id, suffix, names[-1]])521 elif "temporal_transformer_blocks" in names:522 suffix_dict = {523 "attn1.to_out.0": "attn1.to_out",524 "attn2.to_out.0": "attn2.to_out",525 "ff_in.net.0.proj": "act_fn_in.proj",526 "ff_in.net.2": "ff_in",527 "ff.net.0.proj": "act_fn_out.proj",528 "ff.net.2": "ff_out",529 "norm3": "norm_out",530 }531 suffix = ".".join(names[6:-1])532 suffix = suffix_dict.get(suffix, suffix)533 rename = ".".join(["blocks", str(blocks_rename_dict[block_name]), suffix, names[-1]])534 elif "time_mixer" in block_name:535 rename = ".".join(["blocks", str(blocks_rename_dict[block_name]), "proj", names[-1]])536 else:537 suffix_dict = {538 "linear_1": "positional_embedding_proj.0",539 "linear_2": "positional_embedding_proj.2",540 }541 suffix = names[-2]542 suffix = suffix_dict.get(suffix, suffix)543 rename = ".".join(["blocks", str(blocks_rename_dict[block_name]), suffix, names[-1]])544 state_dict_[rename] = param545 else:546 print(name)547 else:548 block_name = self.get_block_name(names)549 if len(block_name)>0 and block_name in blocks_rename_dict:550 rename = ".".join(["blocks", str(blocks_rename_dict[block_name]), names[-1]])551 state_dict_[rename] = param552 return state_dict_553 554 555 def from_civitai(self, state_dict, add_positional_conv=None):556 rename_dict = {557 "model.diffusion_model.input_blocks.0.0.bias": "conv_in.bias",558 "model.diffusion_model.input_blocks.0.0.weight": "conv_in.weight",559 "model.diffusion_model.input_blocks.1.0.emb_layers.1.bias": "blocks.0.time_emb_proj.bias",560 "model.diffusion_model.input_blocks.1.0.emb_layers.1.weight": "blocks.0.time_emb_proj.weight",561 "model.diffusion_model.input_blocks.1.0.in_layers.0.bias": "blocks.0.norm1.bias",562 "model.diffusion_model.input_blocks.1.0.in_layers.0.weight": "blocks.0.norm1.weight",563 "model.diffusion_model.input_blocks.1.0.in_layers.2.bias": "blocks.0.conv1.bias",564 "model.diffusion_model.input_blocks.1.0.in_layers.2.weight": "blocks.0.conv1.weight",565 "model.diffusion_model.input_blocks.1.0.out_layers.0.bias": "blocks.0.norm2.bias",566 "model.diffusion_model.input_blocks.1.0.out_layers.0.weight": "blocks.0.norm2.weight",567 "model.diffusion_model.input_blocks.1.0.out_layers.3.bias": "blocks.0.conv2.bias",568 "model.diffusion_model.input_blocks.1.0.out_layers.3.weight": "blocks.0.conv2.weight",569 "model.diffusion_model.input_blocks.1.0.time_mixer.mix_factor": "blocks.3.mix_factor",570 "model.diffusion_model.input_blocks.1.0.time_stack.emb_layers.1.bias": "blocks.2.time_emb_proj.bias",571 "model.diffusion_model.input_blocks.1.0.time_stack.emb_layers.1.weight": "blocks.2.time_emb_proj.weight",572 "model.diffusion_model.input_blocks.1.0.time_stack.in_layers.0.bias": "blocks.2.norm1.bias",573 "model.diffusion_model.input_blocks.1.0.time_stack.in_layers.0.weight": "blocks.2.norm1.weight",574 "model.diffusion_model.input_blocks.1.0.time_stack.in_layers.2.bias": "blocks.2.conv1.bias",575 "model.diffusion_model.input_blocks.1.0.time_stack.in_layers.2.weight": "blocks.2.conv1.weight",576 "model.diffusion_model.input_blocks.1.0.time_stack.out_layers.0.bias": "blocks.2.norm2.bias",577 "model.diffusion_model.input_blocks.1.0.time_stack.out_layers.0.weight": "blocks.2.norm2.weight",578 "model.diffusion_model.input_blocks.1.0.time_stack.out_layers.3.bias": "blocks.2.conv2.bias",579 "model.diffusion_model.input_blocks.1.0.time_stack.out_layers.3.weight": "blocks.2.conv2.weight",580 "model.diffusion_model.input_blocks.1.1.norm.bias": "blocks.5.norm.bias",581 "model.diffusion_model.input_blocks.1.1.norm.weight": "blocks.5.norm.weight",582 "model.diffusion_model.input_blocks.1.1.proj_in.bias": "blocks.5.proj_in.bias",583 "model.diffusion_model.input_blocks.1.1.proj_in.weight": "blocks.5.proj_in.weight",584 "model.diffusion_model.input_blocks.1.1.proj_out.bias": "blocks.8.proj.bias",585 "model.diffusion_model.input_blocks.1.1.proj_out.weight": "blocks.8.proj.weight",586 "model.diffusion_model.input_blocks.1.1.time_mixer.mix_factor": "blocks.8.mix_factor",587 "model.diffusion_model.input_blocks.1.1.time_pos_embed.0.bias": "blocks.7.positional_embedding_proj.0.bias",588 "model.diffusion_model.input_blocks.1.1.time_pos_embed.0.weight": "blocks.7.positional_embedding_proj.0.weight",589 "model.diffusion_model.input_blocks.1.1.time_pos_embed.2.bias": "blocks.7.positional_embedding_proj.2.bias",590 "model.diffusion_model.input_blocks.1.1.time_pos_embed.2.weight": "blocks.7.positional_embedding_proj.2.weight",591 "model.diffusion_model.input_blocks.1.1.time_stack.0.attn1.to_k.weight": "blocks.7.attn1.to_k.weight",592 "model.diffusion_model.input_blocks.1.1.time_stack.0.attn1.to_out.0.bias": "blocks.7.attn1.to_out.bias",593 "model.diffusion_model.input_blocks.1.1.time_stack.0.attn1.to_out.0.weight": "blocks.7.attn1.to_out.weight",594 "model.diffusion_model.input_blocks.1.1.time_stack.0.attn1.to_q.weight": "blocks.7.attn1.to_q.weight",595 "model.diffusion_model.input_blocks.1.1.time_stack.0.attn1.to_v.weight": "blocks.7.attn1.to_v.weight",596 "model.diffusion_model.input_blocks.1.1.time_stack.0.attn2.to_k.weight": "blocks.7.attn2.to_k.weight",597 "model.diffusion_model.input_blocks.1.1.time_stack.0.attn2.to_out.0.bias": "blocks.7.attn2.to_out.bias",598 "model.diffusion_model.input_blocks.1.1.time_stack.0.attn2.to_out.0.weight": "blocks.7.attn2.to_out.weight",599 "model.diffusion_model.input_blocks.1.1.time_stack.0.attn2.to_q.weight": "blocks.7.attn2.to_q.weight",600 "model.diffusion_model.input_blocks.1.1.time_stack.0.attn2.to_v.weight": "blocks.7.attn2.to_v.weight",601 "model.diffusion_model.input_blocks.1.1.time_stack.0.ff.net.0.proj.bias": "blocks.7.act_fn_out.proj.bias",602 "model.diffusion_model.input_blocks.1.1.time_stack.0.ff.net.0.proj.weight": "blocks.7.act_fn_out.proj.weight",603 "model.diffusion_model.input_blocks.1.1.time_stack.0.ff.net.2.bias": "blocks.7.ff_out.bias",604 "model.diffusion_model.input_blocks.1.1.time_stack.0.ff.net.2.weight": "blocks.7.ff_out.weight",605 "model.diffusion_model.input_blocks.1.1.time_stack.0.ff_in.net.0.proj.bias": "blocks.7.act_fn_in.proj.bias",606 "model.diffusion_model.input_blocks.1.1.time_stack.0.ff_in.net.0.proj.weight": "blocks.7.act_fn_in.proj.weight",607 "model.diffusion_model.input_blocks.1.1.time_stack.0.ff_in.net.2.bias": "blocks.7.ff_in.bias",608 "model.diffusion_model.input_blocks.1.1.time_stack.0.ff_in.net.2.weight": "blocks.7.ff_in.weight",609 "model.diffusion_model.input_blocks.1.1.time_stack.0.norm1.bias": "blocks.7.norm1.bias",610 "model.diffusion_model.input_blocks.1.1.time_stack.0.norm1.weight": "blocks.7.norm1.weight",611 "model.diffusion_model.input_blocks.1.1.time_stack.0.norm2.bias": "blocks.7.norm2.bias",612 "model.diffusion_model.input_blocks.1.1.time_stack.0.norm2.weight": "blocks.7.norm2.weight",613 "model.diffusion_model.input_blocks.1.1.time_stack.0.norm3.bias": "blocks.7.norm_out.bias",614 "model.diffusion_model.input_blocks.1.1.time_stack.0.norm3.weight": "blocks.7.norm_out.weight",615 "model.diffusion_model.input_blocks.1.1.time_stack.0.norm_in.bias": "blocks.7.norm_in.bias",616 "model.diffusion_model.input_blocks.1.1.time_stack.0.norm_in.weight": "blocks.7.norm_in.weight",617 "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.attn1.to_k.weight": "blocks.5.transformer_blocks.0.attn1.to_k.weight",618 "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.attn1.to_out.0.bias": "blocks.5.transformer_blocks.0.attn1.to_out.bias",619 "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.attn1.to_out.0.weight": "blocks.5.transformer_blocks.0.attn1.to_out.weight",620 "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.attn1.to_q.weight": "blocks.5.transformer_blocks.0.attn1.to_q.weight",621 "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.attn1.to_v.weight": "blocks.5.transformer_blocks.0.attn1.to_v.weight",622 "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.attn2.to_k.weight": "blocks.5.transformer_blocks.0.attn2.to_k.weight",623 "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.attn2.to_out.0.bias": "blocks.5.transformer_blocks.0.attn2.to_out.bias",624 "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.attn2.to_out.0.weight": "blocks.5.transformer_blocks.0.attn2.to_out.weight",625 "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.attn2.to_q.weight": "blocks.5.transformer_blocks.0.attn2.to_q.weight",626 "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.attn2.to_v.weight": "blocks.5.transformer_blocks.0.attn2.to_v.weight",627 "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.ff.net.0.proj.bias": "blocks.5.transformer_blocks.0.act_fn.proj.bias",628 "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.ff.net.0.proj.weight": "blocks.5.transformer_blocks.0.act_fn.proj.weight",629 "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.ff.net.2.bias": "blocks.5.transformer_blocks.0.ff.bias",630 "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.ff.net.2.weight": "blocks.5.transformer_blocks.0.ff.weight",631 "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.norm1.bias": "blocks.5.transformer_blocks.0.norm1.bias",632 "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.norm1.weight": "blocks.5.transformer_blocks.0.norm1.weight",633 "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.norm2.bias": "blocks.5.transformer_blocks.0.norm2.bias",634 "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.norm2.weight": "blocks.5.transformer_blocks.0.norm2.weight",635 "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.norm3.bias": "blocks.5.transformer_blocks.0.norm3.bias",636 "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.norm3.weight": "blocks.5.transformer_blocks.0.norm3.weight",637 "model.diffusion_model.input_blocks.10.0.emb_layers.1.bias": "blocks.66.time_emb_proj.bias",638 "model.diffusion_model.input_blocks.10.0.emb_layers.1.weight": "blocks.66.time_emb_proj.weight",639 "model.diffusion_model.input_blocks.10.0.in_layers.0.bias": "blocks.66.norm1.bias",640 "model.diffusion_model.input_blocks.10.0.in_layers.0.weight": "blocks.66.norm1.weight",641 "model.diffusion_model.input_blocks.10.0.in_layers.2.bias": "blocks.66.conv1.bias",642 "model.diffusion_model.input_blocks.10.0.in_layers.2.weight": "blocks.66.conv1.weight",643 "model.diffusion_model.input_blocks.10.0.out_layers.0.bias": "blocks.66.norm2.bias",644 "model.diffusion_model.input_blocks.10.0.out_layers.0.weight": "blocks.66.norm2.weight",645 "model.diffusion_model.input_blocks.10.0.out_layers.3.bias": "blocks.66.conv2.bias",646 "model.diffusion_model.input_blocks.10.0.out_layers.3.weight": "blocks.66.conv2.weight",647 "model.diffusion_model.input_blocks.10.0.time_mixer.mix_factor": "blocks.69.mix_factor",648 "model.diffusion_model.input_blocks.10.0.time_stack.emb_layers.1.bias": "blocks.68.time_emb_proj.bias",649 "model.diffusion_model.input_blocks.10.0.time_stack.emb_layers.1.weight": "blocks.68.time_emb_proj.weight",650 "model.diffusion_model.input_blocks.10.0.time_stack.in_layers.0.bias": "blocks.68.norm1.bias",651 "model.diffusion_model.input_blocks.10.0.time_stack.in_layers.0.weight": "blocks.68.norm1.weight",652 "model.diffusion_model.input_blocks.10.0.time_stack.in_layers.2.bias": "blocks.68.conv1.bias",653 "model.diffusion_model.input_blocks.10.0.time_stack.in_layers.2.weight": "blocks.68.conv1.weight",654 "model.diffusion_model.input_blocks.10.0.time_stack.out_layers.0.bias": "blocks.68.norm2.bias",655 "model.diffusion_model.input_blocks.10.0.time_stack.out_layers.0.weight": "blocks.68.norm2.weight",656 "model.diffusion_model.input_blocks.10.0.time_stack.out_layers.3.bias": "blocks.68.conv2.bias",657 "model.diffusion_model.input_blocks.10.0.time_stack.out_layers.3.weight": "blocks.68.conv2.weight",658 "model.diffusion_model.input_blocks.11.0.emb_layers.1.bias": "blocks.71.time_emb_proj.bias",659 "model.diffusion_model.input_blocks.11.0.emb_layers.1.weight": "blocks.71.time_emb_proj.weight",660 "model.diffusion_model.input_blocks.11.0.in_layers.0.bias": "blocks.71.norm1.bias",661 "model.diffusion_model.input_blocks.11.0.in_layers.0.weight": "blocks.71.norm1.weight",662 "model.diffusion_model.input_blocks.11.0.in_layers.2.bias": "blocks.71.conv1.bias",663 "model.diffusion_model.input_blocks.11.0.in_layers.2.weight": "blocks.71.conv1.weight",664 "model.diffusion_model.input_blocks.11.0.out_layers.0.bias": "blocks.71.norm2.bias",665 "model.diffusion_model.input_blocks.11.0.out_layers.0.weight": "blocks.71.norm2.weight",666 "model.diffusion_model.input_blocks.11.0.out_layers.3.bias": "blocks.71.conv2.bias",667 "model.diffusion_model.input_blocks.11.0.out_layers.3.weight": "blocks.71.conv2.weight",668 "model.diffusion_model.input_blocks.11.0.time_mixer.mix_factor": "blocks.74.mix_factor",669 "model.diffusion_model.input_blocks.11.0.time_stack.emb_layers.1.bias": "blocks.73.time_emb_proj.bias",670 "model.diffusion_model.input_blocks.11.0.time_stack.emb_layers.1.weight": "blocks.73.time_emb_proj.weight",671 "model.diffusion_model.input_blocks.11.0.time_stack.in_layers.0.bias": "blocks.73.norm1.bias",672 "model.diffusion_model.input_blocks.11.0.time_stack.in_layers.0.weight": "blocks.73.norm1.weight",673 "model.diffusion_model.input_blocks.11.0.time_stack.in_layers.2.bias": "blocks.73.conv1.bias",674 "model.diffusion_model.input_blocks.11.0.time_stack.in_layers.2.weight": "blocks.73.conv1.weight",675 "model.diffusion_model.input_blocks.11.0.time_stack.out_layers.0.bias": "blocks.73.norm2.bias",676 "model.diffusion_model.input_blocks.11.0.time_stack.out_layers.0.weight": "blocks.73.norm2.weight",677 "model.diffusion_model.input_blocks.11.0.time_stack.out_layers.3.bias": "blocks.73.conv2.bias",678 "model.diffusion_model.input_blocks.11.0.time_stack.out_layers.3.weight": "blocks.73.conv2.weight",679 "model.diffusion_model.input_blocks.2.0.emb_layers.1.bias": "blocks.10.time_emb_proj.bias",680 "model.diffusion_model.input_blocks.2.0.emb_layers.1.weight": "blocks.10.time_emb_proj.weight",681 "model.diffusion_model.input_blocks.2.0.in_layers.0.bias": "blocks.10.norm1.bias",682 "model.diffusion_model.input_blocks.2.0.in_layers.0.weight": "blocks.10.norm1.weight",683 "model.diffusion_model.input_blocks.2.0.in_layers.2.bias": "blocks.10.conv1.bias",684 "model.diffusion_model.input_blocks.2.0.in_layers.2.weight": "blocks.10.conv1.weight",685 "model.diffusion_model.input_blocks.2.0.out_layers.0.bias": "blocks.10.norm2.bias",686 "model.diffusion_model.input_blocks.2.0.out_layers.0.weight": "blocks.10.norm2.weight",687 "model.diffusion_model.input_blocks.2.0.out_layers.3.bias": "blocks.10.conv2.bias",688 "model.diffusion_model.input_blocks.2.0.out_layers.3.weight": "blocks.10.conv2.weight",689 "model.diffusion_model.input_blocks.2.0.time_mixer.mix_factor": "blocks.13.mix_factor",690 "model.diffusion_model.input_blocks.2.0.time_stack.emb_layers.1.bias": "blocks.12.time_emb_proj.bias",691 "model.diffusion_model.input_blocks.2.0.time_stack.emb_layers.1.weight": "blocks.12.time_emb_proj.weight",692 "model.diffusion_model.input_blocks.2.0.time_stack.in_layers.0.bias": "blocks.12.norm1.bias",693 "model.diffusion_model.input_blocks.2.0.time_stack.in_layers.0.weight": "blocks.12.norm1.weight",694 "model.diffusion_model.input_blocks.2.0.time_stack.in_layers.2.bias": "blocks.12.conv1.bias",695 "model.diffusion_model.input_blocks.2.0.time_stack.in_layers.2.weight": "blocks.12.conv1.weight",696 "model.diffusion_model.input_blocks.2.0.time_stack.out_layers.0.bias": "blocks.12.norm2.bias",697 "model.diffusion_model.input_blocks.2.0.time_stack.out_layers.0.weight": "blocks.12.norm2.weight",698 "model.diffusion_model.input_blocks.2.0.time_stack.out_layers.3.bias": "blocks.12.conv2.bias",699 "model.diffusion_model.input_blocks.2.0.time_stack.out_layers.3.weight": "blocks.12.conv2.weight",700 "model.diffusion_model.input_blocks.2.1.norm.bias": "blocks.15.norm.bias",701 "model.diffusion_model.input_blocks.2.1.norm.weight": "blocks.15.norm.weight",702 "model.diffusion_model.input_blocks.2.1.proj_in.bias": "blocks.15.proj_in.bias",703 "model.diffusion_model.input_blocks.2.1.proj_in.weight": "blocks.15.proj_in.weight",704 "model.diffusion_model.input_blocks.2.1.proj_out.bias": "blocks.18.proj.bias",705 "model.diffusion_model.input_blocks.2.1.proj_out.weight": "blocks.18.proj.weight",706 "model.diffusion_model.input_blocks.2.1.time_mixer.mix_factor": "blocks.18.mix_factor",707 "model.diffusion_model.input_blocks.2.1.time_pos_embed.0.bias": "blocks.17.positional_embedding_proj.0.bias",708 "model.diffusion_model.input_blocks.2.1.time_pos_embed.0.weight": "blocks.17.positional_embedding_proj.0.weight",709 "model.diffusion_model.input_blocks.2.1.time_pos_embed.2.bias": "blocks.17.positional_embedding_proj.2.bias",710 "model.diffusion_model.input_blocks.2.1.time_pos_embed.2.weight": "blocks.17.positional_embedding_proj.2.weight",711 "model.diffusion_model.input_blocks.2.1.time_stack.0.attn1.to_k.weight": "blocks.17.attn1.to_k.weight",712 "model.diffusion_model.input_blocks.2.1.time_stack.0.attn1.to_out.0.bias": "blocks.17.attn1.to_out.bias",713 "model.diffusion_model.input_blocks.2.1.time_stack.0.attn1.to_out.0.weight": "blocks.17.attn1.to_out.weight",714 "model.diffusion_model.input_blocks.2.1.time_stack.0.attn1.to_q.weight": "blocks.17.attn1.to_q.weight",715 "model.diffusion_model.input_blocks.2.1.time_stack.0.attn1.to_v.weight": "blocks.17.attn1.to_v.weight",716 "model.diffusion_model.input_blocks.2.1.time_stack.0.attn2.to_k.weight": "blocks.17.attn2.to_k.weight",717 "model.diffusion_model.input_blocks.2.1.time_stack.0.attn2.to_out.0.bias": "blocks.17.attn2.to_out.bias",718 "model.diffusion_model.input_blocks.2.1.time_stack.0.attn2.to_out.0.weight": "blocks.17.attn2.to_out.weight",719 "model.diffusion_model.input_blocks.2.1.time_stack.0.attn2.to_q.weight": "blocks.17.attn2.to_q.weight",720 "model.diffusion_model.input_blocks.2.1.time_stack.0.attn2.to_v.weight": "blocks.17.attn2.to_v.weight",721 "model.diffusion_model.input_blocks.2.1.time_stack.0.ff.net.0.proj.bias": "blocks.17.act_fn_out.proj.bias",722 "model.diffusion_model.input_blocks.2.1.time_stack.0.ff.net.0.proj.weight": "blocks.17.act_fn_out.proj.weight",723 "model.diffusion_model.input_blocks.2.1.time_stack.0.ff.net.2.bias": "blocks.17.ff_out.bias",724 "model.diffusion_model.input_blocks.2.1.time_stack.0.ff.net.2.weight": "blocks.17.ff_out.weight",725 "model.diffusion_model.input_blocks.2.1.time_stack.0.ff_in.net.0.proj.bias": "blocks.17.act_fn_in.proj.bias",726 "model.diffusion_model.input_blocks.2.1.time_stack.0.ff_in.net.0.proj.weight": "blocks.17.act_fn_in.proj.weight",727 "model.diffusion_model.input_blocks.2.1.time_stack.0.ff_in.net.2.bias": "blocks.17.ff_in.bias",728 "model.diffusion_model.input_blocks.2.1.time_stack.0.ff_in.net.2.weight": "blocks.17.ff_in.weight",729 "model.diffusion_model.input_blocks.2.1.time_stack.0.norm1.bias": "blocks.17.norm1.bias",730 "model.diffusion_model.input_blocks.2.1.time_stack.0.norm1.weight": "blocks.17.norm1.weight",731 "model.diffusion_model.input_blocks.2.1.time_stack.0.norm2.bias": "blocks.17.norm2.bias",732 "model.diffusion_model.input_blocks.2.1.time_stack.0.norm2.weight": "blocks.17.norm2.weight",733 "model.diffusion_model.input_blocks.2.1.time_stack.0.norm3.bias": "blocks.17.norm_out.bias",734 "model.diffusion_model.input_blocks.2.1.time_stack.0.norm3.weight": "blocks.17.norm_out.weight",735 "model.diffusion_model.input_blocks.2.1.time_stack.0.norm_in.bias": "blocks.17.norm_in.bias",736 "model.diffusion_model.input_blocks.2.1.time_stack.0.norm_in.weight": "blocks.17.norm_in.weight",737 "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.attn1.to_k.weight": "blocks.15.transformer_blocks.0.attn1.to_k.weight",738 "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.attn1.to_out.0.bias": "blocks.15.transformer_blocks.0.attn1.to_out.bias",739 "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.attn1.to_out.0.weight": "blocks.15.transformer_blocks.0.attn1.to_out.weight",740 "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.attn1.to_q.weight": "blocks.15.transformer_blocks.0.attn1.to_q.weight",741 "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.attn1.to_v.weight": "blocks.15.transformer_blocks.0.attn1.to_v.weight",742 "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.attn2.to_k.weight": "blocks.15.transformer_blocks.0.attn2.to_k.weight",743 "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.attn2.to_out.0.bias": "blocks.15.transformer_blocks.0.attn2.to_out.bias",744 "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.attn2.to_out.0.weight": "blocks.15.transformer_blocks.0.attn2.to_out.weight",745 "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.attn2.to_q.weight": "blocks.15.transformer_blocks.0.attn2.to_q.weight",746 "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.attn2.to_v.weight": "blocks.15.transformer_blocks.0.attn2.to_v.weight",747 "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.ff.net.0.proj.bias": "blocks.15.transformer_blocks.0.act_fn.proj.bias",748 "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.ff.net.0.proj.weight": "blocks.15.transformer_blocks.0.act_fn.proj.weight",749 "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.ff.net.2.bias": "blocks.15.transformer_blocks.0.ff.bias",750 "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.ff.net.2.weight": "blocks.15.transformer_blocks.0.ff.weight",751 "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.norm1.bias": "blocks.15.transformer_blocks.0.norm1.bias",752 "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.norm1.weight": "blocks.15.transformer_blocks.0.norm1.weight",753 "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.norm2.bias": "blocks.15.transformer_blocks.0.norm2.bias",754 "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.norm2.weight": "blocks.15.transformer_blocks.0.norm2.weight",755 "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.norm3.bias": "blocks.15.transformer_blocks.0.norm3.bias",756 "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.norm3.weight": "blocks.15.transformer_blocks.0.norm3.weight",757 "model.diffusion_model.input_blocks.3.0.op.bias": "blocks.20.conv.bias",758 "model.diffusion_model.input_blocks.3.0.op.weight": "blocks.20.conv.weight",759 "model.diffusion_model.input_blocks.4.0.emb_layers.1.bias": "blocks.22.time_emb_proj.bias",760 "model.diffusion_model.input_blocks.4.0.emb_layers.1.weight": "blocks.22.time_emb_proj.weight",761 "model.diffusion_model.input_blocks.4.0.in_layers.0.bias": "blocks.22.norm1.bias",762 "model.diffusion_model.input_blocks.4.0.in_layers.0.weight": "blocks.22.norm1.weight",763 "model.diffusion_model.input_blocks.4.0.in_layers.2.bias": "blocks.22.conv1.bias",764 "model.diffusion_model.input_blocks.4.0.in_layers.2.weight": "blocks.22.conv1.weight",765 "model.diffusion_model.input_blocks.4.0.out_layers.0.bias": "blocks.22.norm2.bias",766 "model.diffusion_model.input_blocks.4.0.out_layers.0.weight": "blocks.22.norm2.weight",767 "model.diffusion_model.input_blocks.4.0.out_layers.3.bias": "blocks.22.conv2.bias",768 "model.diffusion_model.input_blocks.4.0.out_layers.3.weight": "blocks.22.conv2.weight",769 "model.diffusion_model.input_blocks.4.0.skip_connection.bias": "blocks.22.conv_shortcut.bias",770 "model.diffusion_model.input_blocks.4.0.skip_connection.weight": "blocks.22.conv_shortcut.weight",771 "model.diffusion_model.input_blocks.4.0.time_mixer.mix_factor": "blocks.25.mix_factor",772 "model.diffusion_model.input_blocks.4.0.time_stack.emb_layers.1.bias": "blocks.24.time_emb_proj.bias",773 "model.diffusion_model.input_blocks.4.0.time_stack.emb_layers.1.weight": "blocks.24.time_emb_proj.weight",774 "model.diffusion_model.input_blocks.4.0.time_stack.in_layers.0.bias": "blocks.24.norm1.bias",775 "model.diffusion_model.input_blocks.4.0.time_stack.in_layers.0.weight": "blocks.24.norm1.weight",776 "model.diffusion_model.input_blocks.4.0.time_stack.in_layers.2.bias": "blocks.24.conv1.bias",777 "model.diffusion_model.input_blocks.4.0.time_stack.in_layers.2.weight": "blocks.24.conv1.weight",778 "model.diffusion_model.input_blocks.4.0.time_stack.out_layers.0.bias": "blocks.24.norm2.bias",779 "model.diffusion_model.input_blocks.4.0.time_stack.out_layers.0.weight": "blocks.24.norm2.weight",780 "model.diffusion_model.input_blocks.4.0.time_stack.out_layers.3.bias": "blocks.24.conv2.bias",781 "model.diffusion_model.input_blocks.4.0.time_stack.out_layers.3.weight": "blocks.24.conv2.weight",782 "model.diffusion_model.input_blocks.4.1.norm.bias": "blocks.27.norm.bias",783 "model.diffusion_model.input_blocks.4.1.norm.weight": "blocks.27.norm.weight",784 "model.diffusion_model.input_blocks.4.1.proj_in.bias": "blocks.27.proj_in.bias",785 "model.diffusion_model.input_blocks.4.1.proj_in.weight": "blocks.27.proj_in.weight",786 "model.diffusion_model.input_blocks.4.1.proj_out.bias": "blocks.30.proj.bias",787 "model.diffusion_model.input_blocks.4.1.proj_out.weight": "blocks.30.proj.weight",788 "model.diffusion_model.input_blocks.4.1.time_mixer.mix_factor": "blocks.30.mix_factor",789 "model.diffusion_model.input_blocks.4.1.time_pos_embed.0.bias": "blocks.29.positional_embedding_proj.0.bias",790 "model.diffusion_model.input_blocks.4.1.time_pos_embed.0.weight": "blocks.29.positional_embedding_proj.0.weight",791 "model.diffusion_model.input_blocks.4.1.time_pos_embed.2.bias": "blocks.29.positional_embedding_proj.2.bias",792 "model.diffusion_model.input_blocks.4.1.time_pos_embed.2.weight": "blocks.29.positional_embedding_proj.2.weight",793 "model.diffusion_model.input_blocks.4.1.time_stack.0.attn1.to_k.weight": "blocks.29.attn1.to_k.weight",794 "model.diffusion_model.input_blocks.4.1.time_stack.0.attn1.to_out.0.bias": "blocks.29.attn1.to_out.bias",795 "model.diffusion_model.input_blocks.4.1.time_stack.0.attn1.to_out.0.weight": "blocks.29.attn1.to_out.weight",796 "model.diffusion_model.input_blocks.4.1.time_stack.0.attn1.to_q.weight": "blocks.29.attn1.to_q.weight",797 "model.diffusion_model.input_blocks.4.1.time_stack.0.attn1.to_v.weight": "blocks.29.attn1.to_v.weight",798 "model.diffusion_model.input_blocks.4.1.time_stack.0.attn2.to_k.weight": "blocks.29.attn2.to_k.weight",799 "model.diffusion_model.input_blocks.4.1.time_stack.0.attn2.to_out.0.bias": "blocks.29.attn2.to_out.bias",800 "model.diffusion_model.input_blocks.4.1.time_stack.0.attn2.to_out.0.weight": "blocks.29.attn2.to_out.weight",801 "model.diffusion_model.input_blocks.4.1.time_stack.0.attn2.to_q.weight": "blocks.29.attn2.to_q.weight",802 "model.diffusion_model.input_blocks.4.1.time_stack.0.attn2.to_v.weight": "blocks.29.attn2.to_v.weight",803 "model.diffusion_model.input_blocks.4.1.time_stack.0.ff.net.0.proj.bias": "blocks.29.act_fn_out.proj.bias",804 "model.diffusion_model.input_blocks.4.1.time_stack.0.ff.net.0.proj.weight": "blocks.29.act_fn_out.proj.weight",805 "model.diffusion_model.input_blocks.4.1.time_stack.0.ff.net.2.bias": "blocks.29.ff_out.bias",806 "model.diffusion_model.input_blocks.4.1.time_stack.0.ff.net.2.weight": "blocks.29.ff_out.weight",807 "model.diffusion_model.input_blocks.4.1.time_stack.0.ff_in.net.0.proj.bias": "blocks.29.act_fn_in.proj.bias",808 "model.diffusion_model.input_blocks.4.1.time_stack.0.ff_in.net.0.proj.weight": "blocks.29.act_fn_in.proj.weight",809 "model.diffusion_model.input_blocks.4.1.time_stack.0.ff_in.net.2.bias": "blocks.29.ff_in.bias",810 "model.diffusion_model.input_blocks.4.1.time_stack.0.ff_in.net.2.weight": "blocks.29.ff_in.weight",811 "model.diffusion_model.input_blocks.4.1.time_stack.0.norm1.bias": "blocks.29.norm1.bias",812 "model.diffusion_model.input_blocks.4.1.time_stack.0.norm1.weight": "blocks.29.norm1.weight",813 "model.diffusion_model.input_blocks.4.1.time_stack.0.norm2.bias": "blocks.29.norm2.bias",814 "model.diffusion_model.input_blocks.4.1.time_stack.0.norm2.weight": "blocks.29.norm2.weight",815 "model.diffusion_model.input_blocks.4.1.time_stack.0.norm3.bias": "blocks.29.norm_out.bias",816 "model.diffusion_model.input_blocks.4.1.time_stack.0.norm3.weight": "blocks.29.norm_out.weight",817 "model.diffusion_model.input_blocks.4.1.time_stack.0.norm_in.bias": "blocks.29.norm_in.bias",818 "model.diffusion_model.input_blocks.4.1.time_stack.0.norm_in.weight": "blocks.29.norm_in.weight",819 "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn1.to_k.weight": "blocks.27.transformer_blocks.0.attn1.to_k.weight",820 "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn1.to_out.0.bias": "blocks.27.transformer_blocks.0.attn1.to_out.bias",821 "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn1.to_out.0.weight": "blocks.27.transformer_blocks.0.attn1.to_out.weight",822 "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn1.to_q.weight": "blocks.27.transformer_blocks.0.attn1.to_q.weight",823 "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn1.to_v.weight": "blocks.27.transformer_blocks.0.attn1.to_v.weight",824 "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn2.to_k.weight": "blocks.27.transformer_blocks.0.attn2.to_k.weight",825 "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn2.to_out.0.bias": "blocks.27.transformer_blocks.0.attn2.to_out.bias",826 "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn2.to_out.0.weight": "blocks.27.transformer_blocks.0.attn2.to_out.weight",827 "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn2.to_q.weight": "blocks.27.transformer_blocks.0.attn2.to_q.weight",828 "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn2.to_v.weight": "blocks.27.transformer_blocks.0.attn2.to_v.weight",829 "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.ff.net.0.proj.bias": "blocks.27.transformer_blocks.0.act_fn.proj.bias",830 "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.ff.net.0.proj.weight": "blocks.27.transformer_blocks.0.act_fn.proj.weight",831 "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.ff.net.2.bias": "blocks.27.transformer_blocks.0.ff.bias",832 "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.ff.net.2.weight": "blocks.27.transformer_blocks.0.ff.weight",833 "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.norm1.bias": "blocks.27.transformer_blocks.0.norm1.bias",834 "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.norm1.weight": "blocks.27.transformer_blocks.0.norm1.weight",835 "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.norm2.bias": "blocks.27.transformer_blocks.0.norm2.bias",836 "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.norm2.weight": "blocks.27.transformer_blocks.0.norm2.weight",837 "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.norm3.bias": "blocks.27.transformer_blocks.0.norm3.bias",838 "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.norm3.weight": "blocks.27.transformer_blocks.0.norm3.weight",839 "model.diffusion_model.input_blocks.5.0.emb_layers.1.bias": "blocks.32.time_emb_proj.bias",840 "model.diffusion_model.input_blocks.5.0.emb_layers.1.weight": "blocks.32.time_emb_proj.weight",841 "model.diffusion_model.input_blocks.5.0.in_layers.0.bias": "blocks.32.norm1.bias",842 "model.diffusion_model.input_blocks.5.0.in_layers.0.weight": "blocks.32.norm1.weight",843 "model.diffusion_model.input_blocks.5.0.in_layers.2.bias": "blocks.32.conv1.bias",844 "model.diffusion_model.input_blocks.5.0.in_layers.2.weight": "blocks.32.conv1.weight",845 "model.diffusion_model.input_blocks.5.0.out_layers.0.bias": "blocks.32.norm2.bias",846 "model.diffusion_model.input_blocks.5.0.out_layers.0.weight": "blocks.32.norm2.weight",847 "model.diffusion_model.input_blocks.5.0.out_layers.3.bias": "blocks.32.conv2.bias",848 "model.diffusion_model.input_blocks.5.0.out_layers.3.weight": "blocks.32.conv2.weight",849 "model.diffusion_model.input_blocks.5.0.time_mixer.mix_factor": "blocks.35.mix_factor",850 "model.diffusion_model.input_blocks.5.0.time_stack.emb_layers.1.bias": "blocks.34.time_emb_proj.bias",851 "model.diffusion_model.input_blocks.5.0.time_stack.emb_layers.1.weight": "blocks.34.time_emb_proj.weight",852 "model.diffusion_model.input_blocks.5.0.time_stack.in_layers.0.bias": "blocks.34.norm1.bias",853 "model.diffusion_model.input_blocks.5.0.time_stack.in_layers.0.weight": "blocks.34.norm1.weight",854 "model.diffusion_model.input_blocks.5.0.time_stack.in_layers.2.bias": "blocks.34.conv1.bias",855 "model.diffusion_model.input_blocks.5.0.time_stack.in_layers.2.weight": "blocks.34.conv1.weight",856 "model.diffusion_model.input_blocks.5.0.time_stack.out_layers.0.bias": "blocks.34.norm2.bias",857 "model.diffusion_model.input_blocks.5.0.time_stack.out_layers.0.weight": "blocks.34.norm2.weight",858 "model.diffusion_model.input_blocks.5.0.time_stack.out_layers.3.bias": "blocks.34.conv2.bias",859 "model.diffusion_model.input_blocks.5.0.time_stack.out_layers.3.weight": "blocks.34.conv2.weight",860 "model.diffusion_model.input_blocks.5.1.norm.bias": "blocks.37.norm.bias",861 "model.diffusion_model.input_blocks.5.1.norm.weight": "blocks.37.norm.weight",862 "model.diffusion_model.input_blocks.5.1.proj_in.bias": "blocks.37.proj_in.bias",863 "model.diffusion_model.input_blocks.5.1.proj_in.weight": "blocks.37.proj_in.weight",864 "model.diffusion_model.input_blocks.5.1.proj_out.bias": "blocks.40.proj.bias",865 "model.diffusion_model.input_blocks.5.1.proj_out.weight": "blocks.40.proj.weight",866 "model.diffusion_model.input_blocks.5.1.time_mixer.mix_factor": "blocks.40.mix_factor",867 "model.diffusion_model.input_blocks.5.1.time_pos_embed.0.bias": "blocks.39.positional_embedding_proj.0.bias",868 "model.diffusion_model.input_blocks.5.1.time_pos_embed.0.weight": "blocks.39.positional_embedding_proj.0.weight",869 "model.diffusion_model.input_blocks.5.1.time_pos_embed.2.bias": "blocks.39.positional_embedding_proj.2.bias",870 "model.diffusion_model.input_blocks.5.1.time_pos_embed.2.weight": "blocks.39.positional_embedding_proj.2.weight",871 "model.diffusion_model.input_blocks.5.1.time_stack.0.attn1.to_k.weight": "blocks.39.attn1.to_k.weight",872 "model.diffusion_model.input_blocks.5.1.time_stack.0.attn1.to_out.0.bias": "blocks.39.attn1.to_out.bias",873 "model.diffusion_model.input_blocks.5.1.time_stack.0.attn1.to_out.0.weight": "blocks.39.attn1.to_out.weight",874 "model.diffusion_model.input_blocks.5.1.time_stack.0.attn1.to_q.weight": "blocks.39.attn1.to_q.weight",875 "model.diffusion_model.input_blocks.5.1.time_stack.0.attn1.to_v.weight": "blocks.39.attn1.to_v.weight",876 "model.diffusion_model.input_blocks.5.1.time_stack.0.attn2.to_k.weight": "blocks.39.attn2.to_k.weight",877 "model.diffusion_model.input_blocks.5.1.time_stack.0.attn2.to_out.0.bias": "blocks.39.attn2.to_out.bias",878 "model.diffusion_model.input_blocks.5.1.time_stack.0.attn2.to_out.0.weight": "blocks.39.attn2.to_out.weight",879 "model.diffusion_model.input_blocks.5.1.time_stack.0.attn2.to_q.weight": "blocks.39.attn2.to_q.weight",880 "model.diffusion_model.input_blocks.5.1.time_stack.0.attn2.to_v.weight": "blocks.39.attn2.to_v.weight",881 "model.diffusion_model.input_blocks.5.1.time_stack.0.ff.net.0.proj.bias": "blocks.39.act_fn_out.proj.bias",882 "model.diffusion_model.input_blocks.5.1.time_stack.0.ff.net.0.proj.weight": "blocks.39.act_fn_out.proj.weight",883 "model.diffusion_model.input_blocks.5.1.time_stack.0.ff.net.2.bias": "blocks.39.ff_out.bias",884 "model.diffusion_model.input_blocks.5.1.time_stack.0.ff.net.2.weight": "blocks.39.ff_out.weight",885 "model.diffusion_model.input_blocks.5.1.time_stack.0.ff_in.net.0.proj.bias": "blocks.39.act_fn_in.proj.bias",886 "model.diffusion_model.input_blocks.5.1.time_stack.0.ff_in.net.0.proj.weight": "blocks.39.act_fn_in.proj.weight",887 "model.diffusion_model.input_blocks.5.1.time_stack.0.ff_in.net.2.bias": "blocks.39.ff_in.bias",888 "model.diffusion_model.input_blocks.5.1.time_stack.0.ff_in.net.2.weight": "blocks.39.ff_in.weight",889 "model.diffusion_model.input_blocks.5.1.time_stack.0.norm1.bias": "blocks.39.norm1.bias",890 "model.diffusion_model.input_blocks.5.1.time_stack.0.norm1.weight": "blocks.39.norm1.weight",891 "model.diffusion_model.input_blocks.5.1.time_stack.0.norm2.bias": "blocks.39.norm2.bias",892 "model.diffusion_model.input_blocks.5.1.time_stack.0.norm2.weight": "blocks.39.norm2.weight",893 "model.diffusion_model.input_blocks.5.1.time_stack.0.norm3.bias": "blocks.39.norm_out.bias",894 "model.diffusion_model.input_blocks.5.1.time_stack.0.norm3.weight": "blocks.39.norm_out.weight",895 "model.diffusion_model.input_blocks.5.1.time_stack.0.norm_in.bias": "blocks.39.norm_in.bias",896 "model.diffusion_model.input_blocks.5.1.time_stack.0.norm_in.weight": "blocks.39.norm_in.weight",897 "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn1.to_k.weight": "blocks.37.transformer_blocks.0.attn1.to_k.weight",898 "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn1.to_out.0.bias": "blocks.37.transformer_blocks.0.attn1.to_out.bias",899 "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn1.to_out.0.weight": "blocks.37.transformer_blocks.0.attn1.to_out.weight",900 "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn1.to_q.weight": "blocks.37.transformer_blocks.0.attn1.to_q.weight",901 "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn1.to_v.weight": "blocks.37.transformer_blocks.0.attn1.to_v.weight",902 "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn2.to_k.weight": "blocks.37.transformer_blocks.0.attn2.to_k.weight",903 "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn2.to_out.0.bias": "blocks.37.transformer_blocks.0.attn2.to_out.bias",904 "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn2.to_out.0.weight": "blocks.37.transformer_blocks.0.attn2.to_out.weight",905 "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn2.to_q.weight": "blocks.37.transformer_blocks.0.attn2.to_q.weight",906 "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn2.to_v.weight": "blocks.37.transformer_blocks.0.attn2.to_v.weight",907 "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.ff.net.0.proj.bias": "blocks.37.transformer_blocks.0.act_fn.proj.bias",908 "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.ff.net.0.proj.weight": "blocks.37.transformer_blocks.0.act_fn.proj.weight",909 "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.ff.net.2.bias": "blocks.37.transformer_blocks.0.ff.bias",910 "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.ff.net.2.weight": "blocks.37.transformer_blocks.0.ff.weight",911 "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.norm1.bias": "blocks.37.transformer_blocks.0.norm1.bias",912 "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.norm1.weight": "blocks.37.transformer_blocks.0.norm1.weight",913 "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.norm2.bias": "blocks.37.transformer_blocks.0.norm2.bias",914 "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.norm2.weight": "blocks.37.transformer_blocks.0.norm2.weight",915 "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.norm3.bias": "blocks.37.transformer_blocks.0.norm3.bias",916 "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.norm3.weight": "blocks.37.transformer_blocks.0.norm3.weight",917 "model.diffusion_model.input_blocks.6.0.op.bias": "blocks.42.conv.bias",918 "model.diffusion_model.input_blocks.6.0.op.weight": "blocks.42.conv.weight",919 "model.diffusion_model.input_blocks.7.0.emb_layers.1.bias": "blocks.44.time_emb_proj.bias",920 "model.diffusion_model.input_blocks.7.0.emb_layers.1.weight": "blocks.44.time_emb_proj.weight",921 "model.diffusion_model.input_blocks.7.0.in_layers.0.bias": "blocks.44.norm1.bias",922 "model.diffusion_model.input_blocks.7.0.in_layers.0.weight": "blocks.44.norm1.weight",923 "model.diffusion_model.input_blocks.7.0.in_layers.2.bias": "blocks.44.conv1.bias",924 "model.diffusion_model.input_blocks.7.0.in_layers.2.weight": "blocks.44.conv1.weight",925 "model.diffusion_model.input_blocks.7.0.out_layers.0.bias": "blocks.44.norm2.bias",926 "model.diffusion_model.input_blocks.7.0.out_layers.0.weight": "blocks.44.norm2.weight",927 "model.diffusion_model.input_blocks.7.0.out_layers.3.bias": "blocks.44.conv2.bias",928 "model.diffusion_model.input_blocks.7.0.out_layers.3.weight": "blocks.44.conv2.weight",929 "model.diffusion_model.input_blocks.7.0.skip_connection.bias": "blocks.44.conv_shortcut.bias",930 "model.diffusion_model.input_blocks.7.0.skip_connection.weight": "blocks.44.conv_shortcut.weight",931 "model.diffusion_model.input_blocks.7.0.time_mixer.mix_factor": "blocks.47.mix_factor",932 "model.diffusion_model.input_blocks.7.0.time_stack.emb_layers.1.bias": "blocks.46.time_emb_proj.bias",933 "model.diffusion_model.input_blocks.7.0.time_stack.emb_layers.1.weight": "blocks.46.time_emb_proj.weight",934 "model.diffusion_model.input_blocks.7.0.time_stack.in_layers.0.bias": "blocks.46.norm1.bias",935 "model.diffusion_model.input_blocks.7.0.time_stack.in_layers.0.weight": "blocks.46.norm1.weight",936 "model.diffusion_model.input_blocks.7.0.time_stack.in_layers.2.bias": "blocks.46.conv1.bias",937 "model.diffusion_model.input_blocks.7.0.time_stack.in_layers.2.weight": "blocks.46.conv1.weight",938 "model.diffusion_model.input_blocks.7.0.time_stack.out_layers.0.bias": "blocks.46.norm2.bias",939 "model.diffusion_model.input_blocks.7.0.time_stack.out_layers.0.weight": "blocks.46.norm2.weight",940 "model.diffusion_model.input_blocks.7.0.time_stack.out_layers.3.bias": "blocks.46.conv2.bias",941 "model.diffusion_model.input_blocks.7.0.time_stack.out_layers.3.weight": "blocks.46.conv2.weight",942 "model.diffusion_model.input_blocks.7.1.norm.bias": "blocks.49.norm.bias",943 "model.diffusion_model.input_blocks.7.1.norm.weight": "blocks.49.norm.weight",944 "model.diffusion_model.input_blocks.7.1.proj_in.bias": "blocks.49.proj_in.bias",945 "model.diffusion_model.input_blocks.7.1.proj_in.weight": "blocks.49.proj_in.weight",946 "model.diffusion_model.input_blocks.7.1.proj_out.bias": "blocks.52.proj.bias",947 "model.diffusion_model.input_blocks.7.1.proj_out.weight": "blocks.52.proj.weight",948 "model.diffusion_model.input_blocks.7.1.time_mixer.mix_factor": "blocks.52.mix_factor",949 "model.diffusion_model.input_blocks.7.1.time_pos_embed.0.bias": "blocks.51.positional_embedding_proj.0.bias",950 "model.diffusion_model.input_blocks.7.1.time_pos_embed.0.weight": "blocks.51.positional_embedding_proj.0.weight",951 "model.diffusion_model.input_blocks.7.1.time_pos_embed.2.bias": "blocks.51.positional_embedding_proj.2.bias",952 "model.diffusion_model.input_blocks.7.1.time_pos_embed.2.weight": "blocks.51.positional_embedding_proj.2.weight",953 "model.diffusion_model.input_blocks.7.1.time_stack.0.attn1.to_k.weight": "blocks.51.attn1.to_k.weight",954 "model.diffusion_model.input_blocks.7.1.time_stack.0.attn1.to_out.0.bias": "blocks.51.attn1.to_out.bias",955 "model.diffusion_model.input_blocks.7.1.time_stack.0.attn1.to_out.0.weight": "blocks.51.attn1.to_out.weight",956 "model.diffusion_model.input_blocks.7.1.time_stack.0.attn1.to_q.weight": "blocks.51.attn1.to_q.weight",957 "model.diffusion_model.input_blocks.7.1.time_stack.0.attn1.to_v.weight": "blocks.51.attn1.to_v.weight",958 "model.diffusion_model.input_blocks.7.1.time_stack.0.attn2.to_k.weight": "blocks.51.attn2.to_k.weight",959 "model.diffusion_model.input_blocks.7.1.time_stack.0.attn2.to_out.0.bias": "blocks.51.attn2.to_out.bias",960 "model.diffusion_model.input_blocks.7.1.time_stack.0.attn2.to_out.0.weight": "blocks.51.attn2.to_out.weight",961 "model.diffusion_model.input_blocks.7.1.time_stack.0.attn2.to_q.weight": "blocks.51.attn2.to_q.weight",962 "model.diffusion_model.input_blocks.7.1.time_stack.0.attn2.to_v.weight": "blocks.51.attn2.to_v.weight",963 "model.diffusion_model.input_blocks.7.1.time_stack.0.ff.net.0.proj.bias": "blocks.51.act_fn_out.proj.bias",964 "model.diffusion_model.input_blocks.7.1.time_stack.0.ff.net.0.proj.weight": "blocks.51.act_fn_out.proj.weight",965 "model.diffusion_model.input_blocks.7.1.time_stack.0.ff.net.2.bias": "blocks.51.ff_out.bias",966 "model.diffusion_model.input_blocks.7.1.time_stack.0.ff.net.2.weight": "blocks.51.ff_out.weight",967 "model.diffusion_model.input_blocks.7.1.time_stack.0.ff_in.net.0.proj.bias": "blocks.51.act_fn_in.proj.bias",968 "model.diffusion_model.input_blocks.7.1.time_stack.0.ff_in.net.0.proj.weight": "blocks.51.act_fn_in.proj.weight",969 "model.diffusion_model.input_blocks.7.1.time_stack.0.ff_in.net.2.bias": "blocks.51.ff_in.bias",970 "model.diffusion_model.input_blocks.7.1.time_stack.0.ff_in.net.2.weight": "blocks.51.ff_in.weight",971 "model.diffusion_model.input_blocks.7.1.time_stack.0.norm1.bias": "blocks.51.norm1.bias",972 "model.diffusion_model.input_blocks.7.1.time_stack.0.norm1.weight": "blocks.51.norm1.weight",973 "model.diffusion_model.input_blocks.7.1.time_stack.0.norm2.bias": "blocks.51.norm2.bias",974 "model.diffusion_model.input_blocks.7.1.time_stack.0.norm2.weight": "blocks.51.norm2.weight",975 "model.diffusion_model.input_blocks.7.1.time_stack.0.norm3.bias": "blocks.51.norm_out.bias",976 "model.diffusion_model.input_blocks.7.1.time_stack.0.norm3.weight": "blocks.51.norm_out.weight",977 "model.diffusion_model.input_blocks.7.1.time_stack.0.norm_in.bias": "blocks.51.norm_in.bias",978 "model.diffusion_model.input_blocks.7.1.time_stack.0.norm_in.weight": "blocks.51.norm_in.weight",979 "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn1.to_k.weight": "blocks.49.transformer_blocks.0.attn1.to_k.weight",980 "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn1.to_out.0.bias": "blocks.49.transformer_blocks.0.attn1.to_out.bias",981 "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn1.to_out.0.weight": "blocks.49.transformer_blocks.0.attn1.to_out.weight",982 "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn1.to_q.weight": "blocks.49.transformer_blocks.0.attn1.to_q.weight",983 "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn1.to_v.weight": "blocks.49.transformer_blocks.0.attn1.to_v.weight",984 "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn2.to_k.weight": "blocks.49.transformer_blocks.0.attn2.to_k.weight",985 "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn2.to_out.0.bias": "blocks.49.transformer_blocks.0.attn2.to_out.bias",986 "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn2.to_out.0.weight": "blocks.49.transformer_blocks.0.attn2.to_out.weight",987 "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn2.to_q.weight": "blocks.49.transformer_blocks.0.attn2.to_q.weight",988 "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn2.to_v.weight": "blocks.49.transformer_blocks.0.attn2.to_v.weight",989 "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.ff.net.0.proj.bias": "blocks.49.transformer_blocks.0.act_fn.proj.bias",990 "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.ff.net.0.proj.weight": "blocks.49.transformer_blocks.0.act_fn.proj.weight",991 "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.ff.net.2.bias": "blocks.49.transformer_blocks.0.ff.bias",992 "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.ff.net.2.weight": "blocks.49.transformer_blocks.0.ff.weight",993 "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.norm1.bias": "blocks.49.transformer_blocks.0.norm1.bias",994 "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.norm1.weight": "blocks.49.transformer_blocks.0.norm1.weight",995 "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.norm2.bias": "blocks.49.transformer_blocks.0.norm2.bias",996 "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.norm2.weight": "blocks.49.transformer_blocks.0.norm2.weight",997 "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.norm3.bias": "blocks.49.transformer_blocks.0.norm3.bias",998 "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.norm3.weight": "blocks.49.transformer_blocks.0.norm3.weight",999 "model.diffusion_model.input_blocks.8.0.emb_layers.1.bias": "blocks.54.time_emb_proj.bias",1000 "model.diffusion_model.input_blocks.8.0.emb_layers.1.weight": "blocks.54.time_emb_proj.weight",1001 "model.diffusion_model.input_blocks.8.0.in_layers.0.bias": "blocks.54.norm1.bias",1002 "model.diffusion_model.input_blocks.8.0.in_layers.0.weight": "blocks.54.norm1.weight",1003 "model.diffusion_model.input_blocks.8.0.in_layers.2.bias": "blocks.54.conv1.bias",1004 "model.diffusion_model.input_blocks.8.0.in_layers.2.weight": "blocks.54.conv1.weight",1005 "model.diffusion_model.input_blocks.8.0.out_layers.0.bias": "blocks.54.norm2.bias",1006 "model.diffusion_model.input_blocks.8.0.out_layers.0.weight": "blocks.54.norm2.weight",1007 "model.diffusion_model.input_blocks.8.0.out_layers.3.bias": "blocks.54.conv2.bias",1008 "model.diffusion_model.input_blocks.8.0.out_layers.3.weight": "blocks.54.conv2.weight",1009 "model.diffusion_model.input_blocks.8.0.time_mixer.mix_factor": "blocks.57.mix_factor",1010 "model.diffusion_model.input_blocks.8.0.time_stack.emb_layers.1.bias": "blocks.56.time_emb_proj.bias",1011 "model.diffusion_model.input_blocks.8.0.time_stack.emb_layers.1.weight": "blocks.56.time_emb_proj.weight",1012 "model.diffusion_model.input_blocks.8.0.time_stack.in_layers.0.bias": "blocks.56.norm1.bias",1013 "model.diffusion_model.input_blocks.8.0.time_stack.in_layers.0.weight": "blocks.56.norm1.weight",1014 "model.diffusion_model.input_blocks.8.0.time_stack.in_layers.2.bias": "blocks.56.conv1.bias",1015 "model.diffusion_model.input_blocks.8.0.time_stack.in_layers.2.weight": "blocks.56.conv1.weight",1016 "model.diffusion_model.input_blocks.8.0.time_stack.out_layers.0.bias": "blocks.56.norm2.bias",1017 "model.diffusion_model.input_blocks.8.0.time_stack.out_layers.0.weight": "blocks.56.norm2.weight",1018 "model.diffusion_model.input_blocks.8.0.time_stack.out_layers.3.bias": "blocks.56.conv2.bias",1019 "model.diffusion_model.input_blocks.8.0.time_stack.out_layers.3.weight": "blocks.56.conv2.weight",1020 "model.diffusion_model.input_blocks.8.1.norm.bias": "blocks.59.norm.bias",1021 "model.diffusion_model.input_blocks.8.1.norm.weight": "blocks.59.norm.weight",1022 "model.diffusion_model.input_blocks.8.1.proj_in.bias": "blocks.59.proj_in.bias",1023 "model.diffusion_model.input_blocks.8.1.proj_in.weight": "blocks.59.proj_in.weight",1024 "model.diffusion_model.input_blocks.8.1.proj_out.bias": "blocks.62.proj.bias",1025 "model.diffusion_model.input_blocks.8.1.proj_out.weight": "blocks.62.proj.weight",1026 "model.diffusion_model.input_blocks.8.1.time_mixer.mix_factor": "blocks.62.mix_factor",1027 "model.diffusion_model.input_blocks.8.1.time_pos_embed.0.bias": "blocks.61.positional_embedding_proj.0.bias",1028 "model.diffusion_model.input_blocks.8.1.time_pos_embed.0.weight": "blocks.61.positional_embedding_proj.0.weight",1029 "model.diffusion_model.input_blocks.8.1.time_pos_embed.2.bias": "blocks.61.positional_embedding_proj.2.bias",1030 "model.diffusion_model.input_blocks.8.1.time_pos_embed.2.weight": "blocks.61.positional_embedding_proj.2.weight",1031 "model.diffusion_model.input_blocks.8.1.time_stack.0.attn1.to_k.weight": "blocks.61.attn1.to_k.weight",1032 "model.diffusion_model.input_blocks.8.1.time_stack.0.attn1.to_out.0.bias": "blocks.61.attn1.to_out.bias",1033 "model.diffusion_model.input_blocks.8.1.time_stack.0.attn1.to_out.0.weight": "blocks.61.attn1.to_out.weight",1034 "model.diffusion_model.input_blocks.8.1.time_stack.0.attn1.to_q.weight": "blocks.61.attn1.to_q.weight",1035 "model.diffusion_model.input_blocks.8.1.time_stack.0.attn1.to_v.weight": "blocks.61.attn1.to_v.weight",1036 "model.diffusion_model.input_blocks.8.1.time_stack.0.attn2.to_k.weight": "blocks.61.attn2.to_k.weight",1037 "model.diffusion_model.input_blocks.8.1.time_stack.0.attn2.to_out.0.bias": "blocks.61.attn2.to_out.bias",1038 "model.diffusion_model.input_blocks.8.1.time_stack.0.attn2.to_out.0.weight": "blocks.61.attn2.to_out.weight",1039 "model.diffusion_model.input_blocks.8.1.time_stack.0.attn2.to_q.weight": "blocks.61.attn2.to_q.weight",1040 "model.diffusion_model.input_blocks.8.1.time_stack.0.attn2.to_v.weight": "blocks.61.attn2.to_v.weight",1041 "model.diffusion_model.input_blocks.8.1.time_stack.0.ff.net.0.proj.bias": "blocks.61.act_fn_out.proj.bias",1042 "model.diffusion_model.input_blocks.8.1.time_stack.0.ff.net.0.proj.weight": "blocks.61.act_fn_out.proj.weight",1043 "model.diffusion_model.input_blocks.8.1.time_stack.0.ff.net.2.bias": "blocks.61.ff_out.bias",1044 "model.diffusion_model.input_blocks.8.1.time_stack.0.ff.net.2.weight": "blocks.61.ff_out.weight",1045 "model.diffusion_model.input_blocks.8.1.time_stack.0.ff_in.net.0.proj.bias": "blocks.61.act_fn_in.proj.bias",1046 "model.diffusion_model.input_blocks.8.1.time_stack.0.ff_in.net.0.proj.weight": "blocks.61.act_fn_in.proj.weight",1047 "model.diffusion_model.input_blocks.8.1.time_stack.0.ff_in.net.2.bias": "blocks.61.ff_in.bias",1048 "model.diffusion_model.input_blocks.8.1.time_stack.0.ff_in.net.2.weight": "blocks.61.ff_in.weight",1049 "model.diffusion_model.input_blocks.8.1.time_stack.0.norm1.bias": "blocks.61.norm1.bias",1050 "model.diffusion_model.input_blocks.8.1.time_stack.0.norm1.weight": "blocks.61.norm1.weight",1051 "model.diffusion_model.input_blocks.8.1.time_stack.0.norm2.bias": "blocks.61.norm2.bias",1052 "model.diffusion_model.input_blocks.8.1.time_stack.0.norm2.weight": "blocks.61.norm2.weight",1053 "model.diffusion_model.input_blocks.8.1.time_stack.0.norm3.bias": "blocks.61.norm_out.bias",1054 "model.diffusion_model.input_blocks.8.1.time_stack.0.norm3.weight": "blocks.61.norm_out.weight",1055 "model.diffusion_model.input_blocks.8.1.time_stack.0.norm_in.bias": "blocks.61.norm_in.bias",1056 "model.diffusion_model.input_blocks.8.1.time_stack.0.norm_in.weight": "blocks.61.norm_in.weight",1057 "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.attn1.to_k.weight": "blocks.59.transformer_blocks.0.attn1.to_k.weight",1058 "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.attn1.to_out.0.bias": "blocks.59.transformer_blocks.0.attn1.to_out.bias",1059 "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.attn1.to_out.0.weight": "blocks.59.transformer_blocks.0.attn1.to_out.weight",1060 "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.attn1.to_q.weight": "blocks.59.transformer_blocks.0.attn1.to_q.weight",1061 "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.attn1.to_v.weight": "blocks.59.transformer_blocks.0.attn1.to_v.weight",1062 "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.attn2.to_k.weight": "blocks.59.transformer_blocks.0.attn2.to_k.weight",1063 "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.attn2.to_out.0.bias": "blocks.59.transformer_blocks.0.attn2.to_out.bias",1064 "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.attn2.to_out.0.weight": "blocks.59.transformer_blocks.0.attn2.to_out.weight",1065 "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.attn2.to_q.weight": "blocks.59.transformer_blocks.0.attn2.to_q.weight",1066 "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.attn2.to_v.weight": "blocks.59.transformer_blocks.0.attn2.to_v.weight",1067 "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.ff.net.0.proj.bias": "blocks.59.transformer_blocks.0.act_fn.proj.bias",1068 "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.ff.net.0.proj.weight": "blocks.59.transformer_blocks.0.act_fn.proj.weight",1069 "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.ff.net.2.bias": "blocks.59.transformer_blocks.0.ff.bias",1070 "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.ff.net.2.weight": "blocks.59.transformer_blocks.0.ff.weight",1071 "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.norm1.bias": "blocks.59.transformer_blocks.0.norm1.bias",1072 "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.norm1.weight": "blocks.59.transformer_blocks.0.norm1.weight",1073 "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.norm2.bias": "blocks.59.transformer_blocks.0.norm2.bias",1074 "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.norm2.weight": "blocks.59.transformer_blocks.0.norm2.weight",1075 "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.norm3.bias": "blocks.59.transformer_blocks.0.norm3.bias",1076 "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.norm3.weight": "blocks.59.transformer_blocks.0.norm3.weight",1077 "model.diffusion_model.input_blocks.9.0.op.bias": "blocks.64.conv.bias",1078 "model.diffusion_model.input_blocks.9.0.op.weight": "blocks.64.conv.weight",1079 "model.diffusion_model.label_emb.0.0.bias": "add_time_embedding.0.bias",1080 "model.diffusion_model.label_emb.0.0.weight": "add_time_embedding.0.weight",1081 "model.diffusion_model.label_emb.0.2.bias": "add_time_embedding.2.bias",1082 "model.diffusion_model.label_emb.0.2.weight": "add_time_embedding.2.weight",1083 "model.diffusion_model.middle_block.0.emb_layers.1.bias": "blocks.76.time_emb_proj.bias",1084 "model.diffusion_model.middle_block.0.emb_layers.1.weight": "blocks.76.time_emb_proj.weight",1085 "model.diffusion_model.middle_block.0.in_layers.0.bias": "blocks.76.norm1.bias",1086 "model.diffusion_model.middle_block.0.in_layers.0.weight": "blocks.76.norm1.weight",1087 "model.diffusion_model.middle_block.0.in_layers.2.bias": "blocks.76.conv1.bias",1088 "model.diffusion_model.middle_block.0.in_layers.2.weight": "blocks.76.conv1.weight",1089 "model.diffusion_model.middle_block.0.out_layers.0.bias": "blocks.76.norm2.bias",1090 "model.diffusion_model.middle_block.0.out_layers.0.weight": "blocks.76.norm2.weight",1091 "model.diffusion_model.middle_block.0.out_layers.3.bias": "blocks.76.conv2.bias",1092 "model.diffusion_model.middle_block.0.out_layers.3.weight": "blocks.76.conv2.weight",1093 "model.diffusion_model.middle_block.0.time_mixer.mix_factor": "blocks.79.mix_factor",1094 "model.diffusion_model.middle_block.0.time_stack.emb_layers.1.bias": "blocks.78.time_emb_proj.bias",1095 "model.diffusion_model.middle_block.0.time_stack.emb_layers.1.weight": "blocks.78.time_emb_proj.weight",1096 "model.diffusion_model.middle_block.0.time_stack.in_layers.0.bias": "blocks.78.norm1.bias",1097 "model.diffusion_model.middle_block.0.time_stack.in_layers.0.weight": "blocks.78.norm1.weight",1098 "model.diffusion_model.middle_block.0.time_stack.in_layers.2.bias": "blocks.78.conv1.bias",1099 "model.diffusion_model.middle_block.0.time_stack.in_layers.2.weight": "blocks.78.conv1.weight",1100 "model.diffusion_model.middle_block.0.time_stack.out_layers.0.bias": "blocks.78.norm2.bias",1101 "model.diffusion_model.middle_block.0.time_stack.out_layers.0.weight": "blocks.78.norm2.weight",1102 "model.diffusion_model.middle_block.0.time_stack.out_layers.3.bias": "blocks.78.conv2.bias",1103 "model.diffusion_model.middle_block.0.time_stack.out_layers.3.weight": "blocks.78.conv2.weight",1104 "model.diffusion_model.middle_block.1.norm.bias": "blocks.81.norm.bias",1105 "model.diffusion_model.middle_block.1.norm.weight": "blocks.81.norm.weight",1106 "model.diffusion_model.middle_block.1.proj_in.bias": "blocks.81.proj_in.bias",1107 "model.diffusion_model.middle_block.1.proj_in.weight": "blocks.81.proj_in.weight",1108 "model.diffusion_model.middle_block.1.proj_out.bias": "blocks.84.proj.bias",1109 "model.diffusion_model.middle_block.1.proj_out.weight": "blocks.84.proj.weight",1110 "model.diffusion_model.middle_block.1.time_mixer.mix_factor": "blocks.84.mix_factor",1111 "model.diffusion_model.middle_block.1.time_pos_embed.0.bias": "blocks.83.positional_embedding_proj.0.bias",1112 "model.diffusion_model.middle_block.1.time_pos_embed.0.weight": "blocks.83.positional_embedding_proj.0.weight",1113 "model.diffusion_model.middle_block.1.time_pos_embed.2.bias": "blocks.83.positional_embedding_proj.2.bias",1114 "model.diffusion_model.middle_block.1.time_pos_embed.2.weight": "blocks.83.positional_embedding_proj.2.weight",1115 "model.diffusion_model.middle_block.1.time_stack.0.attn1.to_k.weight": "blocks.83.attn1.to_k.weight",1116 "model.diffusion_model.middle_block.1.time_stack.0.attn1.to_out.0.bias": "blocks.83.attn1.to_out.bias",1117 "model.diffusion_model.middle_block.1.time_stack.0.attn1.to_out.0.weight": "blocks.83.attn1.to_out.weight",1118 "model.diffusion_model.middle_block.1.time_stack.0.attn1.to_q.weight": "blocks.83.attn1.to_q.weight",1119 "model.diffusion_model.middle_block.1.time_stack.0.attn1.to_v.weight": "blocks.83.attn1.to_v.weight",1120 "model.diffusion_model.middle_block.1.time_stack.0.attn2.to_k.weight": "blocks.83.attn2.to_k.weight",1121 "model.diffusion_model.middle_block.1.time_stack.0.attn2.to_out.0.bias": "blocks.83.attn2.to_out.bias",1122 "model.diffusion_model.middle_block.1.time_stack.0.attn2.to_out.0.weight": "blocks.83.attn2.to_out.weight",1123 "model.diffusion_model.middle_block.1.time_stack.0.attn2.to_q.weight": "blocks.83.attn2.to_q.weight",1124 "model.diffusion_model.middle_block.1.time_stack.0.attn2.to_v.weight": "blocks.83.attn2.to_v.weight",1125 "model.diffusion_model.middle_block.1.time_stack.0.ff.net.0.proj.bias": "blocks.83.act_fn_out.proj.bias",1126 "model.diffusion_model.middle_block.1.time_stack.0.ff.net.0.proj.weight": "blocks.83.act_fn_out.proj.weight",1127 "model.diffusion_model.middle_block.1.time_stack.0.ff.net.2.bias": "blocks.83.ff_out.bias",1128 "model.diffusion_model.middle_block.1.time_stack.0.ff.net.2.weight": "blocks.83.ff_out.weight",1129 "model.diffusion_model.middle_block.1.time_stack.0.ff_in.net.0.proj.bias": "blocks.83.act_fn_in.proj.bias",1130 "model.diffusion_model.middle_block.1.time_stack.0.ff_in.net.0.proj.weight": "blocks.83.act_fn_in.proj.weight",1131 "model.diffusion_model.middle_block.1.time_stack.0.ff_in.net.2.bias": "blocks.83.ff_in.bias",1132 "model.diffusion_model.middle_block.1.time_stack.0.ff_in.net.2.weight": "blocks.83.ff_in.weight",1133 "model.diffusion_model.middle_block.1.time_stack.0.norm1.bias": "blocks.83.norm1.bias",1134 "model.diffusion_model.middle_block.1.time_stack.0.norm1.weight": "blocks.83.norm1.weight",1135 "model.diffusion_model.middle_block.1.time_stack.0.norm2.bias": "blocks.83.norm2.bias",1136 "model.diffusion_model.middle_block.1.time_stack.0.norm2.weight": "blocks.83.norm2.weight",1137 "model.diffusion_model.middle_block.1.time_stack.0.norm3.bias": "blocks.83.norm_out.bias",1138 "model.diffusion_model.middle_block.1.time_stack.0.norm3.weight": "blocks.83.norm_out.weight",1139 "model.diffusion_model.middle_block.1.time_stack.0.norm_in.bias": "blocks.83.norm_in.bias",1140 "model.diffusion_model.middle_block.1.time_stack.0.norm_in.weight": "blocks.83.norm_in.weight",1141 "model.diffusion_model.middle_block.1.transformer_blocks.0.attn1.to_k.weight": "blocks.81.transformer_blocks.0.attn1.to_k.weight",1142 "model.diffusion_model.middle_block.1.transformer_blocks.0.attn1.to_out.0.bias": "blocks.81.transformer_blocks.0.attn1.to_out.bias",1143 "model.diffusion_model.middle_block.1.transformer_blocks.0.attn1.to_out.0.weight": "blocks.81.transformer_blocks.0.attn1.to_out.weight",1144 "model.diffusion_model.middle_block.1.transformer_blocks.0.attn1.to_q.weight": "blocks.81.transformer_blocks.0.attn1.to_q.weight",1145 "model.diffusion_model.middle_block.1.transformer_blocks.0.attn1.to_v.weight": "blocks.81.transformer_blocks.0.attn1.to_v.weight",1146 "model.diffusion_model.middle_block.1.transformer_blocks.0.attn2.to_k.weight": "blocks.81.transformer_blocks.0.attn2.to_k.weight",1147 "model.diffusion_model.middle_block.1.transformer_blocks.0.attn2.to_out.0.bias": "blocks.81.transformer_blocks.0.attn2.to_out.bias",1148 "model.diffusion_model.middle_block.1.transformer_blocks.0.attn2.to_out.0.weight": "blocks.81.transformer_blocks.0.attn2.to_out.weight",1149 "model.diffusion_model.middle_block.1.transformer_blocks.0.attn2.to_q.weight": "blocks.81.transformer_blocks.0.attn2.to_q.weight",1150 "model.diffusion_model.middle_block.1.transformer_blocks.0.attn2.to_v.weight": "blocks.81.transformer_blocks.0.attn2.to_v.weight",1151 "model.diffusion_model.middle_block.1.transformer_blocks.0.ff.net.0.proj.bias": "blocks.81.transformer_blocks.0.act_fn.proj.bias",1152 "model.diffusion_model.middle_block.1.transformer_blocks.0.ff.net.0.proj.weight": "blocks.81.transformer_blocks.0.act_fn.proj.weight",1153 "model.diffusion_model.middle_block.1.transformer_blocks.0.ff.net.2.bias": "blocks.81.transformer_blocks.0.ff.bias",1154 "model.diffusion_model.middle_block.1.transformer_blocks.0.ff.net.2.weight": "blocks.81.transformer_blocks.0.ff.weight",1155 "model.diffusion_model.middle_block.1.transformer_blocks.0.norm1.bias": "blocks.81.transformer_blocks.0.norm1.bias",1156 "model.diffusion_model.middle_block.1.transformer_blocks.0.norm1.weight": "blocks.81.transformer_blocks.0.norm1.weight",1157 "model.diffusion_model.middle_block.1.transformer_blocks.0.norm2.bias": "blocks.81.transformer_blocks.0.norm2.bias",1158 "model.diffusion_model.middle_block.1.transformer_blocks.0.norm2.weight": "blocks.81.transformer_blocks.0.norm2.weight",1159 "model.diffusion_model.middle_block.1.transformer_blocks.0.norm3.bias": "blocks.81.transformer_blocks.0.norm3.bias",1160 "model.diffusion_model.middle_block.1.transformer_blocks.0.norm3.weight": "blocks.81.transformer_blocks.0.norm3.weight",1161 "model.diffusion_model.middle_block.2.emb_layers.1.bias": "blocks.85.time_emb_proj.bias",1162 "model.diffusion_model.middle_block.2.emb_layers.1.weight": "blocks.85.time_emb_proj.weight",1163 "model.diffusion_model.middle_block.2.in_layers.0.bias": "blocks.85.norm1.bias",1164 "model.diffusion_model.middle_block.2.in_layers.0.weight": "blocks.85.norm1.weight",1165 "model.diffusion_model.middle_block.2.in_layers.2.bias": "blocks.85.conv1.bias",1166 "model.diffusion_model.middle_block.2.in_layers.2.weight": "blocks.85.conv1.weight",1167 "model.diffusion_model.middle_block.2.out_layers.0.bias": "blocks.85.norm2.bias",1168 "model.diffusion_model.middle_block.2.out_layers.0.weight": "blocks.85.norm2.weight",1169 "model.diffusion_model.middle_block.2.out_layers.3.bias": "blocks.85.conv2.bias",1170 "model.diffusion_model.middle_block.2.out_layers.3.weight": "blocks.85.conv2.weight",1171 "model.diffusion_model.middle_block.2.time_mixer.mix_factor": "blocks.88.mix_factor",1172 "model.diffusion_model.middle_block.2.time_stack.emb_layers.1.bias": "blocks.87.time_emb_proj.bias",1173 "model.diffusion_model.middle_block.2.time_stack.emb_layers.1.weight": "blocks.87.time_emb_proj.weight",1174 "model.diffusion_model.middle_block.2.time_stack.in_layers.0.bias": "blocks.87.norm1.bias",1175 "model.diffusion_model.middle_block.2.time_stack.in_layers.0.weight": "blocks.87.norm1.weight",1176 "model.diffusion_model.middle_block.2.time_stack.in_layers.2.bias": "blocks.87.conv1.bias",1177 "model.diffusion_model.middle_block.2.time_stack.in_layers.2.weight": "blocks.87.conv1.weight",1178 "model.diffusion_model.middle_block.2.time_stack.out_layers.0.bias": "blocks.87.norm2.bias",1179 "model.diffusion_model.middle_block.2.time_stack.out_layers.0.weight": "blocks.87.norm2.weight",1180 "model.diffusion_model.middle_block.2.time_stack.out_layers.3.bias": "blocks.87.conv2.bias",1181 "model.diffusion_model.middle_block.2.time_stack.out_layers.3.weight": "blocks.87.conv2.weight",1182 "model.diffusion_model.out.0.bias": "conv_norm_out.bias",1183 "model.diffusion_model.out.0.weight": "conv_norm_out.weight",1184 "model.diffusion_model.out.2.bias": "conv_out.bias",1185 "model.diffusion_model.out.2.weight": "conv_out.weight",1186 "model.diffusion_model.output_blocks.0.0.emb_layers.1.bias": "blocks.90.time_emb_proj.bias",1187 "model.diffusion_model.output_blocks.0.0.emb_layers.1.weight": "blocks.90.time_emb_proj.weight",1188 "model.diffusion_model.output_blocks.0.0.in_layers.0.bias": "blocks.90.norm1.bias",1189 "model.diffusion_model.output_blocks.0.0.in_layers.0.weight": "blocks.90.norm1.weight",1190 "model.diffusion_model.output_blocks.0.0.in_layers.2.bias": "blocks.90.conv1.bias",1191 "model.diffusion_model.output_blocks.0.0.in_layers.2.weight": "blocks.90.conv1.weight",1192 "model.diffusion_model.output_blocks.0.0.out_layers.0.bias": "blocks.90.norm2.bias",1193 "model.diffusion_model.output_blocks.0.0.out_layers.0.weight": "blocks.90.norm2.weight",1194 "model.diffusion_model.output_blocks.0.0.out_layers.3.bias": "blocks.90.conv2.bias",1195 "model.diffusion_model.output_blocks.0.0.out_layers.3.weight": "blocks.90.conv2.weight",1196 "model.diffusion_model.output_blocks.0.0.skip_connection.bias": "blocks.90.conv_shortcut.bias",1197 "model.diffusion_model.output_blocks.0.0.skip_connection.weight": "blocks.90.conv_shortcut.weight",1198 "model.diffusion_model.output_blocks.0.0.time_mixer.mix_factor": "blocks.93.mix_factor",1199 "model.diffusion_model.output_blocks.0.0.time_stack.emb_layers.1.bias": "blocks.92.time_emb_proj.bias",1200 "model.diffusion_model.output_blocks.0.0.time_stack.emb_layers.1.weight": "blocks.92.time_emb_proj.weight",