haykgrigorian

/

TimeCapsuleLLM-v1-phi1_5-700M

like 23

Hi! Very cool model(s) you've got here. Might I suggest changing the naming convention so users can simply plug in the model hash and know about what they're running.

eg, separate LLMs, named by size, version, and architecture:

• haykgrigorian/TimeCapsuleLLM-v0-nanoGPT-124m
• haykgrigorian/TimeCapsuleLLM-v0_5-nanoGPT-124m
• haykgrigorian/TimeCapsuleLLM-v1-phi1_5-1B
• haykgrigorian/TimeCapsuleLLM-v2-<arch>-200m

Reasoning: A user like me knows the version, size, and architecture right away

vllm serve haykgrigorian/TimeCapsuleLLM-v1-phi1_5-1B #... and get a working model

Adding them to a collection "TimeCapsuleLLM" would be great after that..

Food for thought, if you choose to expand this to different eras, maybe add those to the names 1800s-Britian,.etc