Mention update to PyTorch 2.1.1

Fix so --no-cache is always used

Update PyTorch to 2.1.1

Update README_Runpod_LocalLLMsUIandAPI.md

Update README_Runpod_LocalLLMsUIandAPI.md

Update README_Runpod_LocalLLMsUI.md

Update README.md

Remove `--api` CLI arg for now. Edit logging so it will appear in Runpod log viewer as well (`tee -a`). Change start script to ENTRYPOINT to hide pointless NVidia message.

Update README.md

Update for new API CLI. Update docs

Clarify pod started message

Turn caching back on

Update READMEs, remove outdated cuda11.8.0 textgen container

Update to CUDA 12.1.1, rebuild oneclick

Add chat Serverless Runpod template