Skip to content

Update packaging-models.md #285

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Aug 6, 2019
Merged

Update packaging-models.md #285

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
32d1932
Update packaging-models.md
deliahu Aug 6, 2019
f8d55f9
Update packaging-models.md
ospillinger Aug 6, 2019
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
34 changes: 24 additions & 10 deletions docs/apis/packaging-models.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,22 +2,36 @@

## TensorFlow

Zip the exported estimator output in your checkpoint directory:
Export your trained model and zip the model directory. An example is shown below (here is the [complete example](https://github.com/cortexlabs/cortex/blob/master/examples/iris/models/tensorflow_model.py)):

```text
$ ls export/estimator/1560263597/
saved_model.pb variables/
```Python
import tensorflow as tf
import shutil
import os

...

classifier = tf.estimator.Estimator(
model_fn=my_model, model_dir="iris", params={"hidden_units": [10, 10], "n_classes": 3}
)

exporter = tf.estimator.FinalExporter("estimator", serving_input_fn, as_text=False)
train_spec = tf.estimator.TrainSpec(train_input_fn, max_steps=1000)
eval_spec = tf.estimator.EvalSpec(eval_input_fn, exporters=[exporter], name="estimator-eval")

tf.estimator.train_and_evaluate(classifier, train_spec, eval_spec)

$ zip -r model.zip export/estimator
# zip the estimator export dir (the exported path looks like iris/export/estimator/1562353043/)
shutil.make_archive("tensorflow", "zip", os.path.join("iris/export/estimator"))
```

Upload the zipped file to Amazon S3:
Upload the zipped file to Amazon S3 using the AWS web console or CLI:

```text
$ aws s3 cp model.zip s3://my-bucket/model.zip
```

Reference your `model` in an API:
Reference your model in an `api`:

```yaml
- kind: api
Expand Down Expand Up @@ -45,20 +59,20 @@ with open("sklearn.onnx", "wb") as f:
f.write(onnx_model.SerializeToString())
```

Here are examples of converting models from some of the common ML frameworks to ONNX:
Here are complete examples of converting models from some of the common ML frameworks to ONNX:

* [PyTorch](https://github.com/cortexlabs/cortex/blob/master/examples/iris/models/pytorch_model.py)
* [Sklearn](https://github.com/cortexlabs/cortex/blob/master/examples/iris/models/sklearn_model.py)
* [XGBoost](https://github.com/cortexlabs/cortex/blob/master/examples/iris/models/xgboost_model.py)
* [Keras](https://github.com/cortexlabs/cortex/blob/master/examples/iris/models/keras_model.py)

Upload your trained model in ONNX format to Amazon S3:
Upload your trained model in ONNX format to Amazon S3 using the AWS web console or CLI:

```text
$ aws s3 cp model.onnx s3://my-bucket/model.onnx
```

Reference your `model` in an API:
Reference your model in an `api`:

```yaml
- kind: api
Expand Down