Merge pull request #30 from eubinecto/issue_28

[#27] run_deploy.py done. with some formatting

cleanrnns/fetchers.py

@@ -76,6 +76,7 @@ def fetch_model_for_ner():
 
 def fetch_pipeline_for_classification(entity: str, name: str, run: Run = None) -> PipelineForClassification:
     model = fetch_model_for_classification(entity, name, run)
+    model.eval()
     tokenizer = fetch_tokenizer(entity, run)
     config = fetch_config()[name]
     pipeline = PipelineForClassification(model, tokenizer, config)

explore/explore_cleanrnns_fetch_pipeline_for_classification.py

@@ -9,6 +9,10 @@ def main():
     pipeline = fetch_pipeline_for_classification("eubinecto", "lstm_for_classification")
     print(pipeline("너무 좋다"))  # just a sanity check
     print(pipeline("재미없다"))  # just a sanity check
+    # 향후 테스트를 진행할 계획
+    pipeline = fetch_pipeline_for_classification("eubinecto", "bilstm_for_classification")
+    print(pipeline("너무 좋다"))  # just a sanity check
+    print(pipeline("재미없다"))  # just a sanity check
     shutil.rmtree("artifacts")  # clear the cache after testing
 
 

run_deploy.py

@@ -1,38 +1,46 @@
 """
 use streamlit to deploy them
 """
-from typing import Tuple
-
 import pandas as pd
 import streamlit as st
+from typing import Tuple
 from cleanrnns.fetchers import fetch_pipeline_for_classification
 from cleanrnns.pipelines import PipelineForClassification
 
 
 @st.cache(allow_output_mutation=True)
-def cache_pipeline() -> Tuple[PipelineForClassification, PipelineForClassification]:
+def cache_pipeline() -> Tuple[PipelineForClassification,
+                              PipelineForClassification,
+                              PipelineForClassification]:
     rnn = fetch_pipeline_for_classification("eubinecto", "rnn_for_classification")
     lstm = fetch_pipeline_for_classification("eubinecto", "lstm_for_classification")
-    return rnn, lstm
+    bilstm = fetch_pipeline_for_classification("eubinecto", "bilstm_for_classification")
+    return rnn, lstm, bilstm
 
 
 def main():
     # fetch a pre-trained model
-    rnn, lstm = cache_pipeline()
+    rnn, lstm, bilstm = cache_pipeline()
     st.title("The Clean Rnns - 긍 / 부정 감성분석")
-    text = st.text_input("문장을 입력하세요", value="난 너가 정말 좋으면서도 싫다")
+    text = st.text_input("문장을 입력하세요", value="제목은 시선을 끌지만 줄거리가 애매모호하다")
     if st.button(label="분석하기"):
-        with st.spinner("Please wait..."):
+        with st.spinner("로딩중..."):
             # prediction with RNN
             table = list()
             pred, probs = rnn(text)
-            sentiment = "`긍정`" if pred else "`부정`"
+            sentiment = "🟢(긍정)" if pred else "🔴(부정)"
+            probs = ["{:.4f}".format(prob) for prob in probs]
             table.append(["RNN", sentiment, str(probs)])
             # prediction with LSTM
             pred, probs = lstm(text)
-            sentiment = "`긍정`" if pred else "`부정`"
+            probs = ["{:.4f}".format(prob) for prob in probs]
+            sentiment = "🟢(긍정)" if pred else "🔴(부정)"
             table.append(["LSTM", sentiment, str(probs)])
-            df = pd.DataFrame(table, columns=["모델", "예측", "확률분포"])
+            pred, probs = bilstm(text)
+            sentiment = "🟢(긍정)" if pred else "🔴(부정)"
+            probs = ["{:.4f}".format(prob) for prob in probs]
+            table.append(["RNN", sentiment, str(probs)])
+            df = pd.DataFrame(table, columns=["모델", "예측", "확률분포 [부정, 긍정]"])
             st.markdown(df.to_markdown(index=False))