---
license: mit
metrics:
- mse
library_name: diffusers
tags:
- diffusion
---
## Generate some cu~~rsed~~te kitty cats!

![So cute, right?](https://huggingface.co/nroggendorff/cats/resolve/main/samples/0009.png)

- `train_batch_size`: 16
- `eval_batch_size`: 16
- `num_epochs`: 50
- `gradient_accumulation_steps`: 1
- `learning_rate`: 1e-4
- `lr_warmup_steps`: 500
- `mixed_precision`: "fp16"
- `eval_metric`: "mean_squared_error"

```
import torch
from diffusers import UNet2DModel, DDPMPipeline, DDPMScheduler
import safetensors

# Load model state dictionary
model_path = "ddpm-cats/model_epoch_9.safetensors"
state_dict = safetensors.torch.load_file(model_path)
image_size = 128

# Initialize UNet model
model = UNet2DModel(
    sample_size=image_size,
    in_channels=3,
    out_channels=3,
    layers_per_block=2,
    block_out_channels=(128, 128, 256, 256, 512, 512),
    down_block_types=("DownBlock2D", "DownBlock2D", "DownBlock2D", "DownBlock2D", "AttnDownBlock2D", "DownBlock2D"),
    up_block_types=("UpBlock2D", "AttnUpBlock2D", "UpBlock2D", "UpBlock2D", "UpBlock2D", "UpBlock2D"),
)
model.load_state_dict(state_dict)
model.eval()

# Initialize scheduler
scheduler = DDPMScheduler(
    num_train_timesteps=1000,
    beta_start=0.0001,
    beta_end=0.02,
    beta_schedule="linear",
)

# Initialize pipeline
pipeline = DDPMPipeline(
    unet=model,
    scheduler=scheduler,
)

# Generate image
image = pipeline(batch_size=1, generator=torch.manual_seed(0)).images[0].save("generated_image_128.png")
```