KSEB 3기 04.12 오프라인 수업 후기

 

gpt와 llama의 차이점: llama는 새로운 모델 자체를 오픈함 --> 무료로 customizing해서 사용 가능함, gpt는 과거 모델들을 한정적으로 오픈하고 playground에서 쓰도록 한정함
>>llama 모델이 조금 더 낫다

gpt는 lambda함수를 많이 쓰는 경향이 큼 그러나 lambda 함수는 그리 선호되는 방법이 아님 재활용이 어렵고 속도가 느리기 때문에 좋지 않은 특성이 있음

pandas.DataFrame.select_dtypes 참고

sklearn.preprocessing.LabelEncoder.fit_transform 참고 -> df를 원복시키는 방법 역시 중요

keras.layers.Dense(
    units,
    activation=None,
    use_bias=True,
    kernel_initializer="glorot_uniform",
    bias_initializer="zeros",
    kernel_regularizer=None,
    bias_regularizer=None,
    activity_regularizer=None,
    kernel_constraint=None,
    bias_constraint=None,
    lora_rank=None,
    **kwargs
)

tensorflow.keras.Model.compile()

scaler의 의미? 정규화와 유사함