Quick question: is ibm-granite/granite-3.1-2b-instruct derived from ibm-granite/granite-3.1-2b-base?
#5
by dqdw - opened
Dear [Developer/Team],
I've been using ibm-granite/granite-3.1-2b-instruct recently, and I find it quite useful for my tasks.
Since I plan to extend this model, I want to confirm how it relates to ibm-granite/granite-3.1-2b-base according to Hugging Face:
Direct Fine-tuning: Is it derived directly from ibm-granite/granite-3.1-2b-base, or through other checkpoints?
Inheritance: Was there any merging or distillation involved?
This clarification would help me avoid compatibility issues.
Thank you for your time and support!
Hi @dqdw , the -instruct variants are direct fine tunes of the -base variants using full fine tuning and merging.