f

Author: winskuo-quic

.github/workflows/pull.yml

@@ -565,7 +565,7 @@ jobs:
       docker-image: ci-image:executorch-ubuntu-22.04-qnn-sdk
       submodules: 'recursive'
       ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
-      timeout: 900
+      timeout: 90
       script: |
         # The generic Linux job chooses to use base env, not the one setup by the image
         CONDA_ENV=$(conda env list --json | jq -r ".envs | .[-1]")

examples/qualcomm/oss_scripts/llama/decoder_utils.py

@@ -114,7 +114,7 @@ def __init__(
         ],
         runtime_tokenizer_path,
     ):
-        logging.info("init start~~~")
+        logging.info("Win log: init start~~~")
         self.args = args
         self.pte_path = pte_path
         self.enable_x86_64 = args.enable_x86_64
@@ -196,7 +196,7 @@ def __init__(
         # n seq len = n-1 cache len, so we len(inps) = n-1 during _model_call
         # pyre-ignore
         super().__init__(None, tokenizer, self.max_seq_length - 1)
-        logging.info("init end~~~")
+        logging.info("Win log: init end~~~")
 
     def _model_call(self, inps):
         logging.info(f"Start 1 iteration~~~~~~~~~~~~~~")
@@ -225,7 +225,7 @@ def post_process():
                 with open(f"{self.args.artifact}/{performance_output_path}", "r") as f:
                     self.inference_speed = float(f.read())
 
-        logging.info(f"Finish 1 iteration~~~~~~~~~~~~~~")
+        
         if self.enable_x86_64:
             qnn_sdk = os.getenv("QNN_SDK_ROOT")
             target = "x86_64-linux-clang"
@@ -252,7 +252,9 @@ def post_process():
                 executable="/bin/bash",
                 capture_output=True,
             )
+            logging.info(f"Finish 1 iteration~~~~~~~~~~~~~~")
             post_process()
+            logging.info(f"Finish 1 iteration post process~~~~~~~~~~~~~~")
 
         else:
             runner_cmd = " ".join(