configs/micro_llama_3b.yml

run-title: micro-llama-3b
model: micro-llama-3b

base-model: meta-llama/Llama-3.2-3B
tokenizer: meta-llama/Llama-3.2-3B-Instruct
num-experts: 4
top-k-experts: 1
jitter-noise: 0
use-router: True
mask-input: True
max-length: 8192

trainable:
  - model