File size: 8,403 Bytes
eb94d89 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 |
import types
from typing import List, Optional
import torch
from torch import nn
from einops import rearrange
from utils.scheduler import SchedulerInterface, FlowMatchScheduler
from wan.modules.tokenizers import HuggingfaceTokenizer
from wan.modules.model import WanModel #, RegisterTokens, GanAttentionBlock
from wan.modules.vae import _video_vae
# from wan.modules.t5 import umt5_xxl
from wan.modules.causal_model import CausalWanModel
class WanVAEWrapper(torch.nn.Module): # todo
def __init__(self):
super().__init__()
mean = [
-0.7571, -0.7089, -0.9113, 0.1075, -0.1745, 0.9653, -0.1517, 1.5508,
0.4134, -0.0715, 0.5517, -0.3632, -0.1922, -0.9497, 0.2503, -0.2921
]
std = [
2.8184, 1.4541, 2.3275, 2.6558, 1.2196, 1.7708, 2.6052, 2.0743,
3.2687, 2.1526, 2.8652, 1.5579, 1.6382, 1.1253, 2.8251, 1.9160
]
self.mean = torch.tensor(mean, dtype=torch.float32)
self.std = torch.tensor(std, dtype=torch.float32)
# init model
self.model = _video_vae(
pretrained_path="skyreels_models/Wan2.1-T2V-1.3B/Wan2.1_VAE.pth",
z_dim=16,
).eval().requires_grad_(False)
def encode_to_latent(self, pixel: torch.Tensor) -> torch.Tensor:
# pixel: [batch_size, num_channels, num_frames, height, width]
device, dtype = pixel.device, pixel.dtype
scale = [self.mean.to(device=device, dtype=dtype),
1.0 / self.std.to(device=device, dtype=dtype)]
output = [
self.model.encode(u.unsqueeze(0), scale).float().squeeze(0)
for u in pixel
]
output = torch.stack(output, dim=0)
return output
def decode_to_pixel(self, latent: torch.Tensor, use_cache: bool = False) -> torch.Tensor:
if use_cache:
assert latent.shape[0] == 1, "Batch size must be 1 when using cache"
device, dtype = latent.device, latent.dtype
scale = [self.mean.to(device=device, dtype=dtype),
1.0 / self.std.to(device=device, dtype=dtype)]
if use_cache:
decode_function = self.model.cached_decode
else:
decode_function = self.model.decode
output = []
for u in zs:
output.append(decode_function(u.unsqueeze(0), scale).float().clamp_(-1, 1).squeeze(0))
output = torch.stack(output, dim=0)
return output
class WanDiffusionWrapper(torch.nn.Module):
def __init__(
self,
model_config="",
timestep_shift=5.0,
is_causal=True,
):
super().__init__()
print(model_config)
self.model = CausalWanModel.from_config(model_config)
self.model.eval()
# For non-causal diffusion, all frames share the same timestep
self.uniform_timestep = not is_causal
self.scheduler = FlowMatchScheduler(
shift=timestep_shift, sigma_min=0.0, extra_one_step=True
)
self.scheduler.set_timesteps(1000, training=True)
self.seq_len = 15 * 880 # 32760 # [1, 15, 16, 60, 104]
self.post_init()
def enable_gradient_checkpointing(self) -> None:
self.model.enable_gradient_checkpointing()
def _convert_flow_pred_to_x0(self, flow_pred: torch.Tensor, xt: torch.Tensor, timestep: torch.Tensor) -> torch.Tensor:
"""
Convert flow matching's prediction to x0 prediction.
flow_pred: the prediction with shape [B, C, H, W]
xt: the input noisy data with shape [B, C, H, W]
timestep: the timestep with shape [B]
pred = noise - x0
x_t = (1-sigma_t) * x0 + sigma_t * noise
we have x0 = x_t - sigma_t * pred
see derivations https://chatgpt.com/share/67bf8589-3d04-8008-bc6e-4cf1a24e2d0e
"""
# use higher precision for calculations
original_dtype = flow_pred.dtype
flow_pred, xt, sigmas, timesteps = map(
lambda x: x.double().to(flow_pred.device), [flow_pred, xt,
self.scheduler.sigmas,
self.scheduler.timesteps]
)
timestep_id = torch.argmin(
(timesteps.unsqueeze(0) - timestep.unsqueeze(1)).abs(), dim=1)
sigma_t = sigmas[timestep_id].reshape(-1, 1, 1, 1)
x0_pred = xt - sigma_t * flow_pred
return x0_pred.to(original_dtype)
@staticmethod
def _convert_x0_to_flow_pred(scheduler, x0_pred: torch.Tensor, xt: torch.Tensor, timestep: torch.Tensor) -> torch.Tensor:
"""
Convert x0 prediction to flow matching's prediction.
x0_pred: the x0 prediction with shape [B, C, H, W]
xt: the input noisy data with shape [B, C, H, W]
timestep: the timestep with shape [B]
pred = (x_t - x_0) / sigma_t
"""
# use higher precision for calculations
original_dtype = x0_pred.dtype
x0_pred, xt, sigmas, timesteps = map(
lambda x: x.double().to(x0_pred.device), [x0_pred, xt,
scheduler.sigmas,
scheduler.timesteps]
)
timestep_id = torch.argmin(
(timesteps.unsqueeze(0) - timestep.unsqueeze(1)).abs(), dim=1)
sigma_t = sigmas[timestep_id].reshape(-1, 1, 1, 1)
flow_pred = (xt - x0_pred) / sigma_t
return flow_pred.to(original_dtype)
def forward(
self,
noisy_image_or_video: torch.Tensor, conditional_dict: dict,
timestep: torch.Tensor, kv_cache: Optional[List[dict]] = None, kv_cache_mouse: Optional[List[dict]] = None, kv_cache_keyboard: Optional[List[dict]] = None,
crossattn_cache: Optional[List[dict]] = None,
current_start: Optional[int] = None,
cache_start: Optional[int] = None
) -> torch.Tensor:
assert noisy_image_or_video.shape[1] == 16
# [B, F] -> [B]
if self.uniform_timestep:
input_timestep = timestep[:, 0]
else:
input_timestep = timestep
logits = None
if kv_cache is not None:
flow_pred = self.model(
noisy_image_or_video.to(self.model.dtype),#.permute(0, 2, 1, 3, 4),
t=input_timestep, **conditional_dict,
# seq_len=self.seq_len,
kv_cache=kv_cache,
kv_cache_mouse=kv_cache_mouse, kv_cache_keyboard=kv_cache_keyboard,
crossattn_cache=crossattn_cache,
current_start=current_start,
cache_start=cache_start
)#.permute(0, 2, 1, 3, 4)
else:
flow_pred = self.model(
noisy_image_or_video.to(self.model.dtype),#.permute(0, 2, 1, 3, 4),
t=input_timestep, **conditional_dict)
#.permute(0, 2, 1, 3, 4)
pred_x0 = self._convert_flow_pred_to_x0(
flow_pred=rearrange(flow_pred, 'b c f h w -> (b f) c h w'),#.flatten(0, 1),
xt=rearrange(noisy_image_or_video, 'b c f h w -> (b f) c h w'),#.flatten(0, 1),
timestep=timestep.flatten(0, 1)
)# .unflatten(0, flow_pred.shape[:2])
pred_x0 = rearrange(pred_x0, '(b f) c h w -> b c f h w', b=flow_pred.shape[0])
if logits is not None:
return flow_pred, pred_x0, logits
return flow_pred, pred_x0
def get_scheduler(self) -> SchedulerInterface:
"""
Update the current scheduler with the interface's static method
"""
scheduler = self.scheduler
scheduler.convert_x0_to_noise = types.MethodType(
SchedulerInterface.convert_x0_to_noise, scheduler)
scheduler.convert_noise_to_x0 = types.MethodType(
SchedulerInterface.convert_noise_to_x0, scheduler)
scheduler.convert_velocity_to_x0 = types.MethodType(
SchedulerInterface.convert_velocity_to_x0, scheduler)
self.scheduler = scheduler
return scheduler
def post_init(self):
"""
A few custom initialization steps that should be called after the object is created.
Currently, the only one we have is to bind a few methods to scheduler.
We can gradually add more methods here if needed.
"""
self.get_scheduler()
|