Fix/Improve a bunch of things

[xNul]

In this PR:

• Immediately stops installation on an error
• Properly activates the environment for each command
• Switches from a depreciated method of GPTQ installation to the recommended method
• Fixes the CUDA version mismatch error (still waiting on confirmation though)

  • 29

  • 38

  • https://github.com/oobabooga/text-generation-webui/issues/1440
  • https://github.com/oobabooga/text-generation-webui/issues/1456
• Fixes the "No CUDA runtime is found" error (still waiting on confirmation though)

  • 26

• Fixes the "Torch not compiled with CUDA enabled" error

  • 27

  • https://github.com/oobabooga/text-generation-webui/issues/1394
• Fixes the issue of installing to a separate Windows drive
  • https://github.com/oobabooga/text-generation-webui/issues/1351
• Possibly fixes this error
  • https://github.com/oobabooga/text-generation-webui/issues/1356

[jllllll]

Could you move the webui args to a variable towards the top of the script? I'm thinking that would make it easier for people to add or change arguments.

[xNul]

@jllllll oh sure, that's a good idea. Like this?

WEBUI_ARGS="--chat --model-menu"

[jllllll]

@xNul Yeah, that's good!

[oobabooga]

Thanks for the detailed update!