Bulat comments

Repositories
Issues
Comments

Results 2 comments of


                                            Bulat

Train my own dataset on handwritten digits

Change in create_lmdb_dataset.py: ```env = lmdb.open(outputPath, map_size=1099511627776)``` to ```env = lmdb.open(outputPath, map_size=1073741824)``` worked for me.

RTX3070+CUDA11.3+pytorch1.11.0

Hello. I is this library is not deprecated, the needed support for CUDA >= 11.4 versions. Build with CUDA 11.4 gives error: Dockerfile to reproduce: ```Dockerfile FROM nvcr.io/nvidia/pytorch:22.09-py3 RUN apt-get...