dnn test

recommend-model-produce/src/main/python/models/dnn/config.yaml

@@ -36,7 +36,7 @@ runner:
   # model_init_path: "output_model_dnn/2" # init model
   use_inference: True
   save_inference_feed_varnames: ["1","2","3","4","5","6","7","8","9","10","11","12","13","14","15","16","17","18","19","20","21","22","23","24","25","26","dense_input"]
-  save_inference_fetch_varnames: ["sigmoid_0.tmp_0"]
+  save_inference_fetch_varnames: []
 
   # distribute_config
   sync_mode: "async"

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

@@ -213,11 +213,6 @@ def main(args):
                 model_save_path,
                 epoch_id,
                 prefix='rec_static')
-        if(upload_oss):
-            compress.compress_tar(model_save_path, model_save_path + ".tar.gz")
-            client = HangZhouOSSClient("art-recommend")
-            client.put_object_from_file(oss_object_name, model_save_path + ".tar.gz")
-            logger.info("file {} upload success".format(model_save_path + ".tar.gz"))
         if use_save_data:
             save_data(fetch_batch_var, model_save_path)
 
@@ -251,8 +246,13 @@ def main(args):
                     fetchvars.append(paddle.static.default_main_program()
                                      .global_block().vars[var_name])
 
-            save_inference_model(model_save_path, epoch_id, feedvars,
+            inference_model_path = save_inference_model(model_save_path, epoch_id, feedvars,
                                  fetchvars, exe)
+            if(upload_oss):
+                compress.compress_tar(inference_model_path, "model.tar.gz")
+                client = HangZhouOSSClient("art-recommend")
+                client.put_object_from_file(oss_object_name, "model.tar.gz")
+                logger.info("file {} upload success".format(inference_model_path + ".tar.gz"))
 
 
 def dataset_train(epoch_id, dataset, fetch_vars, exe, config):