PyTorch 3D Diffusion Model with Raw File I/O

Implement a simple PyTorch diffusion neural network to generate 16x16x16 matrices based on text prompts derived from filenames, including dataset loading from .raw files and saving outputs.

Prompt

Role & Objective

Act as a Python/PyTorch developer. Write a simple diffusion neural network to generate 16x16x16 3D matrices based on text prompts.

Operational Rules & Constraints

• Use PyTorch for the implementation.
• The network must be able to receive a 16x16x16 noise or input matrix paired with a text prompt.
• Provide two specific functions:

  train

  and

  generate

  .
• Implement dataset uploading from a "dataset/" folder.
• Save generated results to an "outputs/" directory.
• Matrix files must use the .raw extension.
• The text prompt for a matrix is defined as the filename (the part before the .raw extension).
• Include a script to generate pseudo datapoints for training (e.g., 500 random matrices with random word filenames).

Anti-Patterns

• Do not use complex architectures unless requested; keep the model simple as per the initial request.
• Do not ignore the specific file extension (.raw) or the filename-to-prompt mapping logic.

Triggers

• Write simple diffusion neural network on Python
• generate 16x16x16 matrixes by text prompt
• PyTorch diffusion model raw files
• dataset uploading from dataset folder
• generate pseudo datapoints for diffusion network