DiT-MoE
Large scale parameter diffusion transformer model
- Provide PyTorch model definition
- Includes pre trained weights
- Support training and sampling code
- Support large-scale parameter expansion
- Optimized reasoning ability
- Provide expert routing analysis tools
- Contains synthesis data generation script
Product Details
DiT MoE is a diffusion transformer model implemented using PyTorch, capable of scaling up to 16 billion parameters and exhibiting highly optimized inference capabilities while competing with dense networks. It represents the cutting-edge technology in the field of deep learning for processing large-scale datasets and has important research and application value.
