failed to load model.py file in handler

[SeanWangJS]

Please have a look at FAQ's and Troubleshooting guide, your query may be already addressed.

Your issue may already be reported! Please search on the issue tracker before creating one.

Context

• torchserve version: 0.4.2b20211115
• torch-model-archiver version: 0.4.2b20211115
• torch version:
• torchvision version [if any]:
• torchtext version [if any]:
• torchaudio version [if any]:
• java version:
• Operating System and version:

Your Environment

• Installed using source? [yes/no]: yes
• Are you planning to deploy it using docker container? [yes/no]:
• Is it a CPU or GPU environment?:
• Using a default/custom handler? [If possible upload/share custom handler/model]:
• What kind of model is it e.g. vision, text, audio?:
• Are you planning to use local models from model-store or public url being used e.g. from S3 bucket etc.? [If public url then provide link.]:
• Provide config.properties, logs [ts.log] and parameters used for model registration/update APIs:
• Link to your project [if any]:

Expected Behavior

load the model.py file properly.

Current Behavior

raise RuntimeError("Missing the model.py file") in vgg_handler example.

Possible Solution

let model_file in _load_pickled_model method be os.path.basename(model_file).

Steps to Reproduce

1. archive model by using relatve path, like torch-model-archiver --model-name vgg16 --version 0.1 --model-file .\examples\image_classifier\vgg_16\model.py --serialized-file .....cache\torch\hub\checkpoints\vgg16-397923af.pth --handler .\examples\image_classifier\vgg_16\vgg_handler.py --export-path .\model_store\ -f
2. start serve torchserve --start --ncs --model-store model_store --models vgg16.mar --ts-config .\examples\config.properties ...

Failure Logs [if any]

2021-11-16 10:51:41,653 [INFO ] W-9000-vgg16_0.1-stdout MODEL_LOG - manifest: {'createdOn': '15/11/2021 21:17:23', 'runtime': 'python', 'model': {'modelName': 'vgg16', 'serializedFile': 'vgg16-397923af.pth', 'handler': 'vgg_handler.py', 'modelFile': '.\examples\image_classifier\vgg_16\model.py', 'modelVersion': '0.1'}, 'archiverVersion': '0.4.2'} 2021-11-16 10:51:41,654 [INFO ] W-9000-vgg16_0.1-stdout MODEL_LOG - Loading eager model 2021-11-16 10:51:41,654 [INFO ] W-9000-vgg16_0.1-stdout MODEL_LOG - Loading model from C:\Users\wangx\AppData\Local\Temp\models\8cc14a8fd79b420ebb1d79bf67a388e3.\examples\image_classifier\vgg_16\model.py 2021-11-16 10:51:41,655 [INFO ] W-9000-vgg16_0.1-stdout MODEL_LOG - Backend worker process died. 2021-11-16 10:51:41,655 [INFO ] W-9000-vgg16_0.1-stdout MODEL_LOG - Traceback (most recent call last): 2021-11-16 10:51:41,655 [INFO ] W-9000-vgg16_0.1-stdout MODEL_LOG - File "C:\Users\wangx\anaconda3\Lib\site-packages\ts\model_service_worker.py", line 191, in 2021-11-16 10:51:41,655 [INFO ] W-9000-vgg16_0.1-stdout MODEL_LOG - worker.run_server() 2021-11-16 10:51:41,655 [INFO ] W-9000-vgg16_0.1-stdout MODEL_LOG - File "C:\Users\wangx\anaconda3\Lib\site-packages\ts\model_service_worker.py", line 163, in run_server 2021-11-16 10:51:41,656 [INFO ] W-9000-vgg16_0.1-stdout MODEL_LOG - self.handle_connection(cl_socket) 2021-11-16 10:51:41,656 [INFO ] W-9000-vgg16_0.1-stdout MODEL_LOG - File "C:\Users\wangx\anaconda3\Lib\site-packages\ts\model_service_worker.py", line 124, in handle_connection 2021-11-16 10:51:41,656 [INFO ] W-9000-vgg16_0.1-stdout MODEL_LOG - service, result, code = self.load_model(msg) 2021-11-16 10:51:41,656 [INFO ] W-9000-vgg16_0.1-stdout MODEL_LOG - File "C:\Users\wangx\anaconda3\Lib\site-packages\ts\model_service_worker.py", line 96, in load_model 2021-11-16 10:51:41,656 [INFO ] W-9000-vgg16_0.1-stdout MODEL_LOG - batch_size, envelope, limit_max_image_pixels) 2021-11-16 10:51:41,657 [INFO ] W-9000-vgg16_0.1-stdout MODEL_LOG - File "c:\users\wangx\anaconda3\lib\site-packages\ts\model_loader.py", line 112, in load 2021-11-16 10:51:41,657 [INFO ] W-9000-vgg16_0.1-stdout MODEL_LOG - initialize_fn(service.context) 2021-11-16 10:51:41,657 [INFO ] W-9000-vgg16_0.1-stdout MODEL_LOG - File "c:\users\wangx\anaconda3\lib\site-packages\ts\torch_handler\vision_handler.py", line 20, in initialize 2021-11-16 10:51:41,657 [INFO ] W-9000-vgg16_0.1-stdout MODEL_LOG - super().initialize(context) 2021-11-16 10:51:41,657 [INFO ] W-9000-vgg16_0.1-stdout MODEL_LOG - File "c:\users\wangx\anaconda3\lib\site-packages\ts\torch_handler\base_handler.py", line 66, in initialize 2021-11-16 10:51:41,658 [INFO ] W-9000-vgg16_0.1-stdout MODEL_LOG - self.model = self._load_pickled_model(model_dir, model_file, model_pt_path) 2021-11-16 10:51:41,658 [INFO ] W-9000-vgg16_0.1-stdout MODEL_LOG - File "C:\Users\wangx\AppData\Local\Temp\models\8cc14a8fd79b420ebb1d79bf67a388e3\vgg_handler.py", line 21, in _load_pickled_model 2021-11-16 10:51:41,658 [INFO ] W-9000-vgg16_0.1-stdout MODEL_LOG - raise RuntimeError("Missing the model.py file") 2021-11-16 10:51:41,658 [INFO ] W-9000-vgg16_0.1-stdout MODEL_LOG - RuntimeError: Missing the model.py file

[toretak]

Hi @SeanWangJS, can you try to copy model file into model-archiver workdir .. and pack it like that

torch-model-archiver --model-name vgg16 --version 0.1 --model-file model.py --serialized-file 
.....cache\torch\hub\checkpoints\vgg16-397923af.pth --handler .\examples\image_classifier\vgg_16\vgg_handler.py
 --export-path .\model_store\ -f

It seems that archiver doesn't support relative path for model file. (you can also specify platform - win10? and environment - no docker?) And you can inspect mar file directly (unpack it like a standard zipfile) and check if the model.py is there. if it's there, you can fix model path in MAR-INF/MANIFEST.json