Added check for model before running prediction

predict.py

@@ -6,11 +6,12 @@
 PROJECT_DIR = Path("kitchenware_classifier")
 DATA_DIR = PROJECT_DIR / "data"
 TEST_FILE = DATA_DIR / "test.csv"
-TEST_DF = pd.read_csv(TEST_FILE)
 IMG_DIR = DATA_DIR / "images"
+TEST_DF = pd.read_csv(TEST_FILE)
 TEST_DF["image"] = TEST_DF["Id"].map(lambda x: f"{IMG_DIR}/{x:0>4}.jpg")
-MODEL_FILE = "fastai_model.pkl"
-OUTPUT_FILE = "submission.csv"
+
+MODEL_FILE = Path("fastai_model.pkl")
+EXPORT_FILE = Path("submission.csv")
 
 
 def process_images(df, model):
@@ -26,14 +27,18 @@ def generate_submission(tta, dls):
     vocab = np.array(dls.vocab)
     sub = pd.read_csv(TEST_FILE)
     sub["label"] = vocab[idxs]
-    sub.to_csv(OUTPUT_FILE, index=False)
+    sub.to_csv(EXPORT_FILE, index=False)
     return sub
 
 
 def main():
-    tta, dls = process_images(TEST_DF, MODEL_FILE)
-    sub = generate_submission(tta, dls)
-    print(sub.head())
+    if not MODEL_FILE.exists():
+        print(f"Model {MODEL_FILE} not found!")
+        print("Please run train.py first.")
+    else:
+        tta, dls = process_images(TEST_DF, MODEL_FILE)
+        sub = generate_submission(tta, dls)
+        print(sub.head())
 
 
 if __name__ == "__main__":