𝑺𝒐𝒉𝒚𝒖𝒏

# fixing random seeddef seed_everything(seed:int = 1004): random.seed(seed) np.random.seed(seed) os.environ["PYTHONHASHSEED"] = str(seed) torch.manual_seed(seed) torch.cuda.manual_seed(seed) # current gpu seed torch.cuda.manual_seed_all(seed) # All gpu seed torch.backends.cudnn.deterministic = True # type: ignore torch.backends.cudnn.benchmark = False seed = seed_everyt..

In Python, "self" refers to the specific instance that is calling the method. So, "self" is used to assign or update attributes for the specific instance.# Defining a Class using "self"class greatatcoding: def __init__(self, name, age): self.name = name # "self.name" is an instance attribute; "name" is a local parameter self.age = age def hello(self): print(f"..

Here are some useful code lines for debugging using Pdb..(Pdb) print(UNet2DConditionModel)(Pdb) import inspect(Pdb) print(inspect.getfile(UNet2DConditionModel))/home/aiv-du/anaconda3/envs/text2videozero/lib/python3.9/site-packages/diffusers/models/unet_2d_condition.py (Pdb) import torchvision.utils as tu(Pdb) tu.save_image(xT[0], 'xT[0].png')

1. Using @torch.no_grad() It's a decorator used in Pytorch that disables automatic differentiation. This means that gradients are not calculated within the scope of this function.Since the goal is not to update model weights but rather generate results, it is primarily used during the inference phase.2. torch.randn() vs torch.randn_like() ~ to be continued ~ ~ gonna keep updating ~