RNN 모델의 Text Data 감성분석 Pipeline 제작 Project
TFX를 활용하여 텍스트데이터 감성분석을 수행하는 RNN 모델의 파이프라인을 구축하고 streamlit을 활용해 웹서비스화 해볼 것이다
박찬성님의 TFX Pipeline todo를 참고하였다.
- Notebook to prepare input dataset in
TFRecord
format - Upload the input dataset into the GCS bucket
- Implement and include RNN model in the pipeline
- Implement Streamlit app template
- Make a complete TFX pipeline with
ExampleGen
,SchemaGen
,Resolver
,Trainer
,Evaluator
, andPusher
components - Add necessary configurations to the configs.py
- Add
HFPusher
component to the TFX pipeline - Replace
SchemaGen
withImportSchemaGen
for better TFRecords parsing capability - (Optional) Integrate
Dataflow
inImportExampleGen
to handle a large amount of dataset. This feature is included in the code as a reference, but it is not used after we switched the Sidewalk to PETS dataset.