fix: replaced st.form call with simpler workflow

dashboard_featurefinder.py

@@ -11,6 +11,7 @@ from torchvision.transforms import Compose, Resize, ToPILImage
 from transformers import CLIPProcessor, FlaxCLIPModel
 from PIL import Image
 
+import utils
 
 BASELINE_MODEL = "openai/clip-vit-base-patch32"
 # MODEL_PATH = "/home/shared/models/clip-rsicd/bs128x8-lr5e-6-adam/ckpt-1"
@@ -23,32 +24,6 @@ IMAGE_VECTOR_FILE = "./vectors/test-bs128x8-lr5e-6-adam-ckpt-1.tsv"
 # IMAGES_DIR = "/home/shared/data/rsicd_images"
 IMAGES_DIR = "./images"
 
-2
-# @st.cache(allow_output_mutation=True)
-# def load_index():
-#     filenames, image_vecs = [], []
-#     fvec = open(IMAGE_VECTOR_FILE, "r")
-#     for line in fvec:
-#         cols = line.strip().split('\t')
-#         filename = cols[0]
-#         image_vec = np.array([float(x) for x in cols[1].split(',')])
-#         filenames.append(filename)
-#         image_vecs.append(image_vec)
-#     V = np.array(image_vecs)
-#     index = nmslib.init(method='hnsw', space='cosinesimil')
-#     index.addDataPointBatch(V)
-#     index.createIndex({'post': 2}, print_progress=True)
-#     return filenames, index
-
-
-@st.cache(allow_output_mutation=True)
-def load_model():
-        # model = FlaxCLIPModel.from_pretrained(MODEL_PATH)
-        # processor = CLIPProcessor.from_pretrained(BASELINE_MODEL)
-    model = FlaxCLIPModel.from_pretrained("flax-community/clip-rsicd-v2")
-    processor = CLIPProcessor.from_pretrained("flax-community/clip-rsicd-v2")
-    return model, processor
-
 
 def split_image(X):
   num_rows = X.shape[0] // 224
@@ -86,7 +61,7 @@ def get_image_ranks(probs):
 
 
 def app():
-    model, processor = load_model()
+    model, processor = utils.load_model(MODEL_PATH, BASELINE_MODEL)
 
     st.title("Find Features in Images")
     st.markdown("""
@@ -114,12 +89,9 @@ def app():
         aerial photographs. You will need to download from Unsplash to your 
         computer and upload it to the demo app.
     """)
-    with st.form(key="form_3"):
-        buf = st.file_uploader("Upload Image for Analysis")
-        searched_feature = st.text_input(label="Feature to find")
-        submit_button = st.form_submit_button("Find")
-
-    if submit_button:
+    buf = st.file_uploader("Upload Image for Analysis")
+    searched_feature = st.text_input("Feature to find")
+    if st.button("Find"):
         ftmp = NamedTemporaryFile()
         ftmp.write(buf.getvalue())
         image = plt.imread(ftmp.name)