1 回答
- 最新
- 投票最多
- 评论最多
1
Hi,
Take a look at this implementation using scikit_learn
They are all implemented as part of the HandlerService
object.
input_fn
for preprocessing when data comes inpredict_fn
for making a predictionoutput_fn
for postprocessing
Generally, input_fn
and output_fn
have default implementations provided by SageMaker, but predict_fn
does not.
已回答 2 年前
相关内容
- AWS 官方已更新 1 年前
- AWS 官方已更新 1 年前
- AWS 官方已更新 6 个月前
Hi, thanks for the answer, so it means model_fn needs to be mandatorily implemented (overriden) and we can ignore overriding/defining input_fn, predict_fn and output_fn methods while writing entry point script?