Update app.py
Browse files
app.py
CHANGED
|
@@ -296,36 +296,28 @@ async def analyze_with_som(embedding, predicted_family, filename):
|
|
| 296 |
try:
|
| 297 |
print("\n🗺️ Performing SOM analysis (New Logic)...")
|
| 298 |
|
| 299 |
-
# 1.
|
| 300 |
-
|
| 301 |
-
|
| 302 |
-
|
| 303 |
-
data_source = ""
|
| 304 |
-
|
| 305 |
-
if hf_embeddings and len(hf_embeddings) > 0:
|
| 306 |
-
# 2. 如果找到了 -> 計算平均 embedding
|
| 307 |
-
print(f"✅ Found {len(hf_embeddings)} embeddings from HF dataset. Calculating average.")
|
| 308 |
-
embedding_to_use = np.mean(hf_embeddings, axis=0, dtype=np.float32)
|
| 309 |
-
data_source = "hf_dataset_avg"
|
| 310 |
else:
|
| 311 |
-
|
| 312 |
-
|
| 313 |
-
if isinstance(embedding, list):
|
| 314 |
-
embedding_to_use = np.array(embedding, dtype=np.float32)
|
| 315 |
-
else:
|
| 316 |
-
embedding_to_use = embedding
|
| 317 |
-
data_source = "realtime_model" # (之前這裡是 background_data,現在更正)
|
| 318 |
-
|
| 319 |
# 4. 根據選定的 embedding 決定座標
|
| 320 |
winner_row, winner_col = som_analyzer.find_winner_neuron(embedding_to_use)
|
| 321 |
-
print(f" 📍 Winner neuron: ({winner_row}, {winner_col}) (Source:
|
|
|
|
|
|
|
|
|
|
|
|
|
| 322 |
|
| 323 |
# 5. 分析座標周圍 (使用已抓取的 hf_embeddings 或 fallback)
|
| 324 |
if hf_embeddings and len(hf_embeddings) > 0:
|
| 325 |
-
|
| 326 |
features_prob = analyze_som_region_probability(
|
| 327 |
som_analyzer, winner_row, winner_col, hf_embeddings, radius=2
|
| 328 |
)
|
|
|
|
| 329 |
else:
|
| 330 |
# (使用原有的 fallback 邏輯)
|
| 331 |
print(f"⚠️ No HF data for prob, using background data")
|
|
@@ -333,8 +325,7 @@ async def analyze_with_som(embedding, predicted_family, filename):
|
|
| 333 |
winner_row, winner_col, radius=1
|
| 334 |
)
|
| 335 |
features_prob = {k: float(v) for k, v in features_dict.items()}
|
| 336 |
-
|
| 337 |
-
data_source = "background_data_prob" if data_source == "realtime_model" else data_source
|
| 338 |
|
| 339 |
# 6. 組合最終結果 (此部分不變)
|
| 340 |
features_binary = {
|
|
@@ -357,6 +348,12 @@ async def analyze_with_som(embedding, predicted_family, filename):
|
|
| 357 |
import traceback
|
| 358 |
traceback.print_exc()
|
| 359 |
return None
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 360 |
|
| 361 |
|
| 362 |
@app.post("/predict")
|
|
|
|
| 296 |
try:
|
| 297 |
print("\n🗺️ Performing SOM analysis (New Logic)...")
|
| 298 |
|
| 299 |
+
# 1. 決定用於 "定位" 的 Embedding
|
| 300 |
+
# (我們假設傳入的 'embedding' 已經是最好的了,例如高注意力的那一個)
|
| 301 |
+
if isinstance(embedding, list):
|
| 302 |
+
embedding_to_use = np.array(embedding, dtype=np.float32)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 303 |
else:
|
| 304 |
+
embedding_to_use = embedding
|
| 305 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 306 |
# 4. 根據選定的 embedding 決定座標
|
| 307 |
winner_row, winner_col = som_analyzer.find_winner_neuron(embedding_to_use)
|
| 308 |
+
print(f" 📍 Winner neuron: ({winner_row}, {winner_col}) (Source: high-attention segment)")
|
| 309 |
+
|
| 310 |
+
# 1. (原步驟1) 現在去抓取 "所有" 相關的 embeddings,用於 "鄰域分析"
|
| 311 |
+
hf_embeddings = await get_embeddings_from_hf_dataset(filename)
|
| 312 |
+
data_source = "high-attention (pos) + " # 標記數據源
|
| 313 |
|
| 314 |
# 5. 分析座標周圍 (使用已抓取的 hf_embeddings 或 fallback)
|
| 315 |
if hf_embeddings and len(hf_embeddings) > 0:
|
| 316 |
+
print(f" Analysing region using {len(hf_embeddings)} HF embeddings...")
|
| 317 |
features_prob = analyze_som_region_probability(
|
| 318 |
som_analyzer, winner_row, winner_col, hf_embeddings, radius=2
|
| 319 |
)
|
| 320 |
+
data_source += "hf_dataset (prob)"
|
| 321 |
else:
|
| 322 |
# (使用原有的 fallback 邏輯)
|
| 323 |
print(f"⚠️ No HF data for prob, using background data")
|
|
|
|
| 325 |
winner_row, winner_col, radius=1
|
| 326 |
)
|
| 327 |
features_prob = {k: float(v) for k, v in features_dict.items()}
|
| 328 |
+
data_source += "background_data (prob)"
|
|
|
|
| 329 |
|
| 330 |
# 6. 組合最終結果 (此部分不變)
|
| 331 |
features_binary = {
|
|
|
|
| 348 |
import traceback
|
| 349 |
traceback.print_exc()
|
| 350 |
return None
|
| 351 |
+
|
| 352 |
+
except Exception as e:
|
| 353 |
+
print(f"❌ SOM analysis error: {e}")
|
| 354 |
+
import traceback
|
| 355 |
+
traceback.print_exc()
|
| 356 |
+
return None
|
| 357 |
|
| 358 |
|
| 359 |
@app.post("/predict")
|