5.14 예상 출력 예시

아래 출력은 환경에 따른 예상 출력 예시다.

python3 code/normalization.py

train episodes: [0, 1, 2, 3, 4, 6, 7, 8]
validation episodes: [5, 9]
train mean: [-0.0757  0.3933 -0.219 ]
validation normalized mean: [ 0.226  -0.0607  0.1596]
round-trip max error: 1.1102230246251565e-16

python3 code/temporal_ensemble.py

step=0: action=[0. 0.]
step=1: action=[0.3401 0.1401]
step=2: action=[0.674 0.274]
step=3: action=[1.074 0.474]

python3 code/lora_budget.py --hidden 1024 --rank 16 --layers 16

Simplified LoRA estimate: 1,048,576
corresponding full matrices: 33,554,432
ratio: 3.125%
scaling example (alpha=32): 2.0
Use the framework's trainable-parameter report as the authoritative count.