MinerU
MinerU copied to clipboard
Package too large
The package is too large. If built into a docker image, including the model, it will weigh more than 20g. If GPU support is added, it will double. There are also many redundancies in the project. For example, paddle ocr repeatedly installs torch and gpu-related libraries, and the output utilization efficiency of layoutmlv3 model is not high, and the information utilization efficiency of pdf itself is also not high