Updates for November 15

[bfhealy]

Pull requests

• Checkpoint feature downloads #157
• Add readme instruction for generating docs locally #158
• Save multiple feature files per field #159
• Loop over multiple parquet feature files during inference #162
• Add WandB group name to training runs #163
• Restructure features lists in config defaults #164 (supersedes Remove ra/dec/field/ccd/quad from training #142)
• Set feature datatypes and resume interrupted downloads #165
• Save inference results in HDF5 format #166
• Add method to generate inference script #167
• Change flake8 repo link to fix failing tests #168

Issues

• Add WandB group name to training runs #136
• Save inference results in HDF5 format #160
• Code to generate inference script #161

Going forward

With improvements to the inference pipeline nearly complete, we are in good position to run inference on the 20-field sample within the next week. After a little more experimentation with model training, we will move forward with inference and begin uploading high-confidence results to Fritz.