Update app.py

app.py

@@ -3,38 +3,31 @@ from fastai.tabular.all import *
 import pandas as pd
 import torch
 
-model = load_learner('learn_model.pkl')
+df_framework = pd.read_csv('/kaggle/input/stroke-prediction-dataset/healthcare-dataset-stroke-data.csv')
+splits = RandomSplitter(seed=42)(df_framework)
 
-def convert(vocab, data):
-    tensor =  torch.zeros(len(vocab), len(data))
-    for i, char in enumerate(vocab):
-        tensor[i][ord(char) - ord('a')] = i
+dls = TabularPandas(
+    df_framework, splits=splits,
+    procs = [Categorify, FillMissing, Normalize],
+    cat_names=["gender","ever_married","work_type","Residence_type", "smoking_status"],
+    cont_names=['age', 'hypertension', 'heart_disease', 'avg_glucose_level', 'bmi'],
+    y_names="stroke", y_block = CategoryBlock(),
+).dataloaders(path=".")
+
+learn = tabular_learner(dls, metrics=accuracy, layers=[10,10])
+learn.fit(16, lr=0.025)
 
-    return tensor
-    
 def predict(age, hypertension, heart_disease, avg_glucose_level, bmi, gender, married, work_type, residence_type, smoking_status):
-    
+
+    data = [[gender, age, hypertension, heart_disease, married, work_type, residence_type, avg_glucose_level, bmi, smoking_status]]
     columns_df = ['gender', 'age', 'hypertension', 'heart_disease', 'ever_married', 'work_type', 'Residence_type', 'avg_glucose_level', 'bmi','smoking_status']
-    tensor = []
-    
-    data = [gender, age, hypertension, heart_disease, married, work_type, residence_type, avg_glucose_level, bmi, smoking_status]
+    df_row = pd.DataFrame(data,columns=columns_df)
 
+    dl = learn.dls.test_dl(df_row)
+    preds,_ = learn.get_preds(dl=dl)
     
-    # df = pd.Dataframe(data, columns=columns_df)
-    age_tensor = torch.tensor(age, dtype=torch.int64)
-    hypertension_tensor = convert(hypertension, data)
-    heart_disease_tensor = convert(heart_disease, data)
-    avg_glucose_level_tensor = torch.tensor(avg_glucose_level, dtype=torch.float)
-    bmi_tensor = torch.tensor(bmi, dtype=torch.float)
-    gender_tensor = convert(hypertension, data)
-    married_tensor = convert(married, data)
-    work_type_tensor = convert(work_type, data)
-    residence_type_tensor = convert(residence_type, data)
-    smoking_status_tensor = convert(smoking_status, data)
-    
-    tensor = torch.cat([gender_tensor, married_tensor, work_type_tensor, residence_type_tensor, smoking_status_tensor, bmi_tensor, age_tensor, hypertension_tensor, heart_disease_tensor, avg_glucose_level_tensor])
-    prediction = model.predict(tensor)
-    return "O paciente tem a seguinte possibilidade de infarto: " + str(predictions[0])
+    return "O paciente tem a seguinte possibilidade de infarto: " + str(preds[0])
+
 
 gr.Interface(
     fn=predict,
@@ -46,7 +39,7 @@ gr.Interface(
         gr.Dropdown(["Yes", "No"], label="Hypertension"),
         gr.Dropdown(["Yes", "No"], label="Heart disease"),
         gr.inputs.Number(default=100, label="Average glucose level"),
-        gr.inputs.Number(default=28.8, label="Body mass index (BMI)"),
+        gr.inputs.Number(default=28.8, label="Body Mass Index (BMI)"),
         gr.inputs.Radio(choices=["Female", "Male", "Other"], default="Female", label="Gender"),
         gr.Dropdown(["No", "Yes"], label="Married"),
         gr.Dropdown(["Governamental", "Never worked", "Private", "Self-employed", "Children"], label="Work type"),