Fixed issue where code-7b and other parameter sets were failing to load. Temp fix please review.

[JonathanC8]

These have been modified to use a version of python cpp that runs faster and seems to work without issues on code-7b and others. I have also made a change noted in the following issues thread which fixes those parameter packages from failing.
I don't expect this to get merged since it does not include sha security in the compose files. I am only asking for a review from the maintainers to fix this at some point. Thank you all for this project.
[https://github.com/getumbrel/llama-gpt/issues/98](url)

modified:   docker-compose-gguf.yml
modified:   docker-compose.yml