torch-model-archiver

# Assume you have a PyTorch model 'model.py' and a serialized state_dict 'model.pth' # Also assume you have a handler 'handler.py' (or use a default one like 'image_classifier') # Create a simple dummy model.py and handler.py for demonstration: # model.py: # import torch.nn as nn # class MyModel(nn.Module): # def __init__(self): # super(MyModel, self).__init__() # self.linear = nn.Linear(10, 1) # def forward(self, x): # return self.linear(x) # # handler.py (minimal): # from ts.torch_handler.base_handler import BaseHandler # class MyHandler(BaseHandler): # def preprocess(self, data): # # Implement your data preprocessing logic # return data # def postprocess(self, data): # # Implement your data postprocessing logic # return data # Command to archive a model (example with a hypothetical densenet161 setup): # Ensure 'densenet161_model.py', 'densenet161_state.pth', and 'index_to_name.json' exist # For a real run, replace paths with actual files and ensure handler logic matches the model. # Example from TorchServe docs (adjust paths if running locally without cloning the repo) # This assumes a model file like 'densenet_161/model.py' and a state dict like 'densenet161-8d451a50.pth' # and a default handler 'image_classifier' # # Make a dummy model_store directory import os os.makedirs('model_store', exist_ok=True) # This example is illustrative. For a runnable quickstart, you'd need to provide actual model.py, .pth, and handler files. # A fully runnable quickstart often involves downloading example assets from the TorchServe repo. # This specific command uses a generic handler and placeholder files. # In a real scenario, you'd replace 'my_model.py', 'my_model_state.pth', and 'my_handler.py' with your actual files. # We are using 'image_classifier' as a built-in handler for demonstration purposes. print("To create a model archive (.mar) file:") print("torch-model-archiver --model-name mymodel --version 1.0 --model-file path/to/my_model.py --serialized-file path/to/my_model_state.pth --handler image_classifier --export-path model_store -f") print("\nThis command will create 'model_store/mymodel.mar'") # Example using subprocess (if you wanted to run it from Python) import subprocess # This path is relative to the torchserve repo; adjust if you cloned it elsewhere or use your own model files. # For a truly isolated example, you'd need to create dummy files or download real ones. model_name = "densenet161" model_version = "1.0" # Placeholder paths for demonstration model_file_path = "./dummy_model.py" serialized_file_path = "./dummy_state.pth" export_path = "model_store" handler_name = "image_classifier" # Using a default handler for simplicity # Create dummy files if they don't exist for the subprocess command to not error immediately with open(model_file_path, "w") as f: f.write("import torch.nn as nn\nclass MyModel(nn.Module):\n def __init__(self):\n super().__init__()\n self.linear = nn.Linear(10, 1)\n def forward(self, x):\n return self.linear(x)") # Create a dummy serialized file (e.g., an empty file or a minimal PyTorch save) import torch torch.save({'state_dict': {}}, serialized_file_path) cmd = [ "torch-model-archiver", "--model-name", model_name, "--version", model_version, "--model-file", model_file_path, "--serialized-file", serialized_file_path, "--handler", handler_name, "--export-path", export_path, "-f" # Force overwrite if file exists ] try: # Not actually running this in a quickstart as it requires external files, just showing the structure # subprocess.run(cmd, check=True, capture_output=True) # print(f"Successfully created {export_path}/{model_name}.mar") pass # Suppress actual execution for quickstart to avoid requiring external files except subprocess.CalledProcessError as e: print(f"Error archiving model: {e.stderr.decode()}") except FileNotFoundError: print("Error: 'torch-model-archiver' command not found. Please ensure the library is installed and in your PATH.") # Clean up dummy files os.remove(model_file_path) os.remove(serialized_file_path) # os.rmdir(export_path) # Don't remove if you expect a .mar file for a real test