Update app.py

app.py

@@ -12,14 +12,29 @@ from transformers import TFAutoModelForSequenceClassification
 from transformers import AutoTokenizer
 from transformers_interpret import SequenceClassificationExplainer
 
+device = "cuda:0" if torch.cuda.is_available() else "cpu"
+
 tokenizer = AutoTokenizer.from_pretrained("paragon-analytics/ADRv1")  
-model = AutoModelForSequenceClassification.from_pretrained("paragon-analytics/ADRv1")
+model = AutoModelForSequenceClassification.from_pretrained("paragon-analytics/ADRv1").to(device)
+
+# build a pipeline object to do predictions
+pred = transformers.pipeline("text-classification", model=model, 
+                             tokenizer=tokenizer, return_all_scores=True)
+
+
+def interpretation_function(text):
+    explainer = shap.Explainer(pred)
+    shap_values = explainer([text])
+    scores = list(zip(shap_values.data[0], shap_values.values[0, :, 1]))
+    return scores
+
+# model = AutoModelForSequenceClassification.from_pretrained("paragon-analytics/ADRv1")
 # modelc = AutoModelForSequenceClassification.from_pretrained("paragon-analytics/ADRv1").cuda
 
 
-cls_explainer = SequenceClassificationExplainer(
-    model,
-    tokenizer)
+# cls_explainer = SequenceClassificationExplainer(
+#     model,
+#     tokenizer)
 
 # # define a prediction function
 # def f(x):
@@ -42,26 +57,26 @@ def adr_predict(x):
     # shap_values = explainer([x])
     # shap_plot = shap.plots.text(shap_values)
 
-    word_attributions = cls_explainer(str(x))
-    # scores = list(zip(shap_values.data[0], shap_values.values[0, :, 1]))
-    letter = []
-    score = []
-    for i in word_attributions:
-        if i[1]>0.5:
-            a = "++"
-        elif (i[1]<=0.5) and (i[1]>0.1):
-            a = "+"
-        elif (i[1]>=-0.5) and (i[1]<-0.1):
-            a = "-"
-        elif i[1]<-0.5:
-            a = "--"
-        else: 
-            a = "NA"
+    # word_attributions = cls_explainer(str(x))
+    # # scores = list(zip(shap_values.data[0], shap_values.values[0, :, 1]))
+    # letter = []
+    # score = []
+    # for i in word_attributions:
+    #     if i[1]>0.5:
+    #         a = "++"
+    #     elif (i[1]<=0.5) and (i[1]>0.1):
+    #         a = "+"
+    #     elif (i[1]>=-0.5) and (i[1]<-0.1):
+    #         a = "-"
+    #     elif i[1]<-0.5:
+    #         a = "--"
+    #     else: 
+    #         a = "NA"
     
-        letter.append(i[0])
-        score.append(a)
+    #     letter.append(i[0])
+    #     score.append(a)
     
-    word_attributions = [(letter[i], score[i]) for i in range(0, len(letter))]
+    # word_attributions = [(letter[i], score[i]) for i in range(0, len(letter))]
 
     # # SHAP:
     # # build an explainer using a token masker
@@ -70,10 +85,11 @@ def adr_predict(x):
     # scores = list(zip(shap_values.data[0], shap_values.values[0, :, 1]))
     # # plot the first sentence's explanation
     # # plt = shap.plots.text(shap_values[0],display=False)
+    shap_scores = interpretation_function(str(x).lower())
 
 
-    return {"Severe Reaction": float(scores.numpy()[1]), "Non-severe Reaction": float(scores.numpy()[0])}, word_attributions
-    # ,scores
+    return {"Severe Reaction": float(scores.numpy()[1]), "Non-severe Reaction": float(scores.numpy()[0])}, shap_scores
+    # , word_attributions ,scores
 
 def main(text):
     text = str(text).lower()
@@ -100,25 +116,27 @@ with gr.Blocks(title=title) as demo:
         # color_map={"+++": "royalblue","++": "cornflowerblue",
         #  "+": "lightsteelblue", "NA":"white"})
         # NER = gr.HTML(label = 'NER:')
-        intp = gr.HighlightedText(label="Word Scores",
-        combine_adjacent=False).style(color_map={"++": "darkred","+": "red", 
-                                                "--": "darkblue",
-                                                "-": "blue", "NA":"white"})
+        # intp = gr.HighlightedText(label="Word Scores",
+        # combine_adjacent=False).style(color_map={"++": "darkred","+": "red", 
+        #                                         "--": "darkblue",
+        #                                         "-": "blue", "NA":"white"})
 
-        # interpretation = gr.components.Interpretation(text)
+        interpretation = gr.components.Interpretation(text)
 
 
     submit_btn.click(
         main,
         [text],
-        [label,intp
-         # ,interpretation
+        [label
+         # ,intp
+         ,interpretation
         ], api_name="adr"
     )
 
     gr.Markdown("### Click on any of the examples below to see to what extent they contain resilience messaging:")
-    gr.Examples([["I have minor pain."],["I have severe pain."]], [text], [label,intp
-                                                                           # ,interpretation
+    gr.Examples([["I have minor pain."],["I have severe pain."]], [text], [label
+                                                                           # ,intp
+                                                                           ,interpretation
                                                                           ], main, cache_examples=True)
     
 demo.launch()