Upload app.py with huggingface_hub

app.py

@@ -146,18 +146,19 @@ def _run_bur_embedding():
 
     st.info(f"Embedding {n:,} kompetencji (batch={BATCH_SIZE})…")
     embeddings_list = []
+    status_placeholder = st.empty()
     pbar = st.progress(0.0, text=f"0 / {n:,} (0%)")
-    batch_step = 50  # aktualizacja co 50 batchy – mniej overhead, stały postęp
     for i in range(0, n, BATCH_SIZE):
         batch = [(labels[j] or " ").strip() for j in range(i, min(i + BATCH_SIZE, n))]
         emb = encode_fn(batch, normalize_embeddings=True, batch_size=BATCH_SIZE, show_progress_bar=False)
         for k in range(len(batch)):
             embeddings_list.append(np.asarray(emb[k], dtype=np.float32))
         done = min(i + len(batch), n)
-        if (i // BATCH_SIZE) % batch_step == 0 or done == n:
-            pct = 100 * done / n
-            pbar.progress(done / n, text=f"{done:,} / {n:,} ({pct:.1f}%)")
-            time.sleep(0.05)  # pozwala Streamlit wysłać aktualizację do frontendu
+        pct = 100 * done / n
+        status_placeholder.caption(f"📊 {done:,} / {n:,} ({pct:.1f}%)")
+        pbar.progress(done / n, text=f"{done:,} / {n:,} ({pct:.1f}%)")
+        time.sleep(0.08)  # pozwala Streamlit wysłać aktualizację na HF (wolne WebSocket)
+    status_placeholder.empty()
     pbar.progress(1.0, text=f"Gotowe: {n:,} (100%)")
 
     arr = np.stack(embeddings_list, axis=0)
@@ -211,8 +212,7 @@ def main():
             st.download_button("⬇️ bur_competencies_kalm_metadata.pkl", data=BUR_META.read_bytes(), file_name="bur_competencies_kalm_metadata.pkl", mime="application/octet-stream", key="bur_meta")
         else:
             if st.button("Uruchom embedding BUR", type="primary", key="bur_btn"):
-                with st.spinner("Trwa embedding (~410k kompetencji, może 2–3 h)…"):
-                    _run_bur_embedding()
+                _run_bur_embedding()
                 st.rerun()