Stable Diffusion x4 upscaler

Stable Diffusion x4 upscaler - is a diffusion model for 4x image resolution upscaling based on text prompts. It takes as input a low-resolution image and a text prompt, along with the noise_level parameter.

Model type: Diffusion-based text-to-image generation model
Text encoder: Pre-trained OpenCLIP-ViT/H model.
Training: 1.25 million steps on 32 NVIDIA A100 GPUs, using the AdamW optimizer with a learning rate of 0.0001.
Input dimensions: 512x512 images and English-language text prompts.

Limitations:

• Does not provide photorealism.
• Does not generate readable text.
• Performs poorly with complex compositions in images (e.g., "a red cube on a blue sphere").

License: CreativeML Open RAIL++-M. Use is prohibited for creating harmful content (e.g., violence, sexual violence, or the spread of stereotypes).

---

The model is a component of the image generation pipeline, consisting of:

• CLIP text encoder: ~340M parameters,
• UNET: ~473M parameters,
• VAE: ~55M parameters.

Total: ~869M parameters

---

The 'VRAM requirements' line is calculated based on the size of the weights in the specified precision, with an additional +10% coefficient for inference:
  model_parameters_count * bytes_per_parameter * 1.1 / 1024^3 
where model_parameters_count is number of model parameters, bytes_per_parameter is number of bytes used to store one parameter. For 16 bit = 2, 8 bit = 1, 4 bit = 0.5.

Announce Date: 23.11.2022
Parameters: 0.869B
Context: 77
VRAM requirements: 0.4 GB using 4 bits quantization, 0.9 GB using 8 bits quantization, 1.8 GB using 16 bits quantization
Developer: Stability AI
License: OpenRail++

Use our pre-built public endpoints for free to test inference and explore Stable Diffusion x4 upscaler capabilities. You can obtain an API access token on the token management page after registration and verification.

Rent your own physically dedicated instance with hourly or long-term monthly billing.

We recommend deploying private instances in the following scenarios:

• maximize endpoint performance,
• enable full context for long sequences,
• ensure top-tier security for data processing in an isolated, dedicated environment,
• use custom weights, such as fine-tuned models or LoRA adapters.

Contact our dedicated neural networks support team at nn@immers.cloud or send your request to the sales department at sale@immers.cloud.