Proposal: A way to skip LLMs

[jeremyfowers]

Problem:

• running turnkey transformers/*.py is a nice way to sweep over lots of transformers
• that folder now contains a lot of LLMs
• LLMs will melt your RAM and hard drive if they run on insufficient hardware

Why this is needed: if I run a job on a machine with limit RAM and HDD, I want to skip over LLMs so that my system doesn't crash.

Possible solutions:

• Quick/sad: move the LLMs out of transformers into a different folder (eg, models/llm)
• Moderate: add an task::llm or similar label to the LLMs, and a way to exclude models based on their labels (eg, turnkey transformers/*.py --labels ~task::llm
• Advanced: filter models based on parameter count like turnkey transformers/*.py --max-params 1B

cc @danielholanda @ramkrishna2910