- Extracting chunk data from a Minecraft server.
- Training various models on the chunk data.
- Insert model output back into the Minecraft server
- ????
- Profit
- Extracts .mca region files from Minecraft server docker container
- Creates a backup of region file in container, then inserts source region file into Minecraft docker container
- Extracts Minecraft NBT chunk data from .mca files and plots the chunks as voxels.
- Extracts Minecraft NBT chunk data from .mca files and plots the blocks as voxels.
- Takes Minecraft NBT chunks and compresses and stores them in .mca region format to be consumed by a Minecraft server
- Simple FFN model that takes an input of a NxNxN numpy array.
- Splits the NxNxN array into sub chunks.
- Generate dataset where the input is a sub chunk and the output is the up to 6 sub chunks it shares a face with. The sub chunk is rotated such that input chunk is always facing its output chunk.
- Training the model to overfit on the dataset to get an idea of number of parameters and settings of hyperparameters.
- Training a FFN model on actual chunk data.
- Splitting a 16x16x16 into 2x2x2 chunks.
- Checking to see how many one-to-many relationships this causes as with sub chunks this small its very possible to have exact chunks have a large variance in adjacent chunks.
- Training a FFN model on actual chunk data.
- Splitting a 16x16x16 into 4x4x4 chunks.
- Increased sub chunk size reduces the number of one-to-many relationships at the cost of model size.
- Loads region files, extracts NBT chunks.
- Creates a combined palette that contains all block_state + biome combinations.
- Creates a 16x16x16 numpy array for each sub chunk, where the value represents an index in the combined block+biome palette.