dssm train

recommend-model-produce/src/main/python/tools/inferv2.py

@@ -40,6 +40,9 @@ def process_file(file_path, model_file, params_file):
         left_features = [float(x) for x in left_features_str.split(',')]
         input_data[vid] = left_features
 
+
+    i=0
+    count=len(input_data)
     result = []
     for k, v in input_data.items():
         v2 = np.array([v], dtype=np.float32)
@@ -49,6 +52,9 @@ def process_file(file_path, model_file, params_file):
         output_handle = predictor.get_output_handle(predictor.get_output_names()[0])
         output_data = output_handle.copy_to_cpu()
         result.append(k + "\t" + str(output_data.tolist()[0]))
+        i=i+1
+        if i % 1000 == 0:
+            print(f"Thread {name}: write batch {i}/{count}")
 
     return result
 
@@ -79,9 +85,10 @@ def main():
     params_file = os.path.join(init_model_path, "dssm.pdiparams")
 
     max_workers = 2
+
     split_file_list = [
         ['/dw/recommend/model/56_dssm_i2i_itempredData/20241206/part-00017.gz'],
-        ['/dw/recommend/model/56_dssm_i2i_itempredData/20241206/part-00017.gz']
+        ['/dw/recommend/model/56_dssm_i2i_itempredData/20241206/part-00018.gz']
     ]
     future_list = []
     with ThreadPoolExecutor(max_workers=max_workers) as executor: