Add docker image link

Use None default for test-data-path

Add file input for generate_samples.py

Add lock for loading checkpoint

Add script for small model

Update README for model parallel

Add change_mp.py

Add no-load-lr-scheduler argument

Fix model parallel

Remove hierarchical generation

Remove socket name

Merge branch 'xinhua'

Set train data

Update README.md

Implement xinhua finetune