Skip to content

Latest commit

 

History

History
149 lines (116 loc) · 7.22 KB

README.md

File metadata and controls

149 lines (116 loc) · 7.22 KB
Raw

Sample commands to create a resnet-18 eager mode model archive, register it on TorchServe and run image prediction

Run the commands given in following steps from the parent directory of the root of the repository. For example, if you cloned the repository into /home/my_path/serve, run the steps from /home/my_path/serve

wget https://download.pytorch.org/models/resnet18-f37072fd.pth
torch-model-archiver --model-name resnet-18 --version 1.0 --model-file ./examples/image_classifier/resnet_18/model.py --serialized-file resnet18-f37072fd.pth --handler image_classifier --extra-files ./examples/image_classifier/index_to_name.json
mkdir model_store
mv resnet-18.mar model_store/
torchserve --start --model-store model_store --models resnet-18=resnet-18.mar --disable-token-auth  --enable-model-api
curl http://127.0.0.1:8080/predictions/resnet-18 -T ./examples/image_classifier/kitten.jpg

TorchServe inference with torch.compile of Resnet18 model

This example shows how to take eager model of Resnet18, configure TorchServe to use torch.compile and run inference using torch.compile.

Change directory to the examples directory Ex: cd examples/image_classifier/resnet_18

torch.compile config

torch.compile supports a variety of config and the performance you get can vary based on the config. You can find the various options here.

In this example , we use the following config

echo "pt2:
  compile:
    enable: True
    backend: inductor
    mode: reduce-overhead" > model-config.yaml
Sample commands to create a Resnet18 torch.compile model archive, register it on TorchServe and run image prediction
wget https://download.pytorch.org/models/resnet18-f37072fd.pth
torch-model-archiver --model-name resnet-18 --version 1.0 --model-file model.py --serialized-file resnet18-f37072fd.pth --handler image_classifier --extra-files ../index_to_name.json --config-file model-config.yaml
mkdir model_store
mv resnet-18.mar model_store/
torchserve --start --model-store model_store --models resnet-18=resnet-18.mar --disable-token-auth  --enable-model-api
curl http://127.0.0.1:8080/predictions/resnet-18 -T ../kitten.jpg

produces the output

{
  "tabby": 0.40966343879699707,
  "tiger_cat": 0.346704363822937,
  "Egyptian_cat": 0.13002890348434448,
  "lynx": 0.023919545114040375,
  "bucket": 0.011532172560691833
}

TorchScript example using Resnet18 image classifier:

  • Save the Resnet18 model in as an executable script module or a traced script:

  1. Save model using scripting

    #scripted mode
from torchvision import models
import torch
model = models.resnet18(pretrained=True)
sm = torch.jit.script(model)
sm.save("resnet-18.pt")

  2. Save model using tracing

    #traced mode
from torchvision import models
import torch
model = models.resnet18(pretrained=True)
model.eval()
example_input = torch.rand(1, 3, 224, 224)
traced_script_module = torch.jit.trace(model, example_input)
traced_script_module.save("resnet-18.pt")

  • Use following commands to register Resnet18 torchscript model on TorchServe and run image prediction

    torch-model-archiver --model-name resnet-18 --version 1.0  --serialized-file resnet-18.pt --extra-files ./serve/examples/image_classifier/index_to_name.json --handler image_classifier
mkdir model_store
mv resnet-18.mar model_store/
torchserve --start --model-store model_store --models resnet-18=resnet-18.mar --disable-token-auth  --enable-model-api
curl http://127.0.0.1:8080/predictions/resnet-18 -T ./serve/examples/image_classifier/kitten.jpg

Debug TorchServe Backend

If you want to test your handler code, you can use the example in debugging_backend/test_handler.py

python debugging_backend/test_handler.py --batch_size 2

results in

Torch TensorRT not enabled
DEBUG:ts.torch_handler.base_handler:Model file /home/ubuntu/serve/examples/image_classifier/resnet_18/resnet-18.pt loaded successfully
INFO:__main__:Result is [{'tabby': 0.4096629023551941, 'tiger_cat': 0.34670525789260864, 'Egyptian_cat': 0.13002872467041016, 'lynx': 0.02391958236694336, 'bucket': 0.011532173492014408}, {'tabby': 0.4096629023551941, 'tiger_cat': 0.34670525789260864, 'Egyptian_cat': 0.13002872467041016, 'lynx': 0.02391958236694336, 'bucket': 0.011532173492014408}]

If this doesn't work, you can use a debugger to find the problem in your backend handler code. Once you are confident this works, you can use your handler to deploy the model using TorchServe

Below is a screenshot of debugger running with this handler

image info

You can also use this with pytest

pytest debugging_backend/test_handler.py

results in

================================================================================== test session starts ===================================================================================
platform linux -- Python 3.8.18, pytest-7.3.1, pluggy-1.0.0
rootdir: /home/ubuntu/serve
plugins: mock-3.10.0, anyio-3.6.1, cov-4.1.0, hypothesis-6.54.3
collected 1 item

debugging_backend/test_handler.py .                                                                                                                                                [100%]

==================================================================================== warnings summary ====================================================================================
../../../../anaconda3/envs/torchserve/lib/python3.8/site-packages/ts/torch_handler/base_handler.py:13
  /home/ubuntu/anaconda3/envs/torchserve/lib/python3.8/site-packages/ts/torch_handler/base_handler.py:13: DeprecationWarning: pkg_resources is deprecated as an API. See https://setuptools.pypa.io/en/latest/pkg_resources.html
    from pkg_resources import packaging

../../../../anaconda3/envs/torchserve/lib/python3.8/site-packages/pkg_resources/__init__.py:2871
  /home/ubuntu/anaconda3/envs/torchserve/lib/python3.8/site-packages/pkg_resources/__init__.py:2871: DeprecationWarning: Deprecated call to `pkg_resources.declare_namespace('mpl_toolkits')`.
  Implementing implicit namespace packages (as specified in PEP 420) is preferred to `pkg_resources.declare_namespace`. See https://setuptools.pypa.io/en/latest/references/keywords.html#keyword-namespace-packages
    declare_namespace(pkg)

../../../../anaconda3/envs/torchserve/lib/python3.8/site-packages/pkg_resources/__init__.py:2871
../../../../anaconda3/envs/torchserve/lib/python3.8/site-packages/pkg_resources/__init__.py:2871
  /home/ubuntu/anaconda3/envs/torchserve/lib/python3.8/site-packages/pkg_resources/__init__.py:2871: DeprecationWarning: Deprecated call to `pkg_resources.declare_namespace('ruamel')`.
  Implementing implicit namespace packages (as specified in PEP 420) is preferred to `pkg_resources.declare_namespace`. See https://setuptools.pypa.io/en/latest/references/keywords.html#keyword-namespace-packages
    declare_namespace(pkg)

-- Docs: https://docs.pytest.org/en/stable/how-to/capture-warnings.html
============================================================================= 1 passed, 4 warnings in 2.29s ==============================================================================