mirror of
https://github.com/GoogleCloudPlatform/vertex-ai-samples.git
synced 2026-09-26 14:42:04 +00:00
feat: autodiscover (#775)
* feat: import automl tabular model * feat: import automl tabular model * feat: HPT for non-TF * feat: HPT for non-TF * fix: split guidelines from template * fix: split guidelines from template * fix: split guidelines from template * upgrade: updates for new release * upgrade: updates for new release * update: tune title * update: tune title * feat: auto-discover * fix: title * fix: title * fix: title * fix: title * fix: title * fix: title * feat: autodiscover * feat: autodiscover * fix: title * fix: title * feat: autodiscover * fix: title * fix: title * feat: autodiscover * fix: title * fix: title * feat: autodiscover * fix: title * fix: title * fix: title * fix: title * feat: autodiscover
This commit is contained in:
Andrew Ferlitsch
@@ -29,11 +29,49 @@ This stage may be done entirely by MLOps. We recommend:
|
||||
|
||||
### Get Started
|
||||
|
||||
[Get started with Custom Prediction Routine](get_started_with_cpr.ipynb)
|
||||
|
||||
[Get started with TensorFlow serving functions with Vertex AI Prediction](get_started_with_tf_serving_function.ipynb)
|
||||
|
||||
```
|
||||
The steps performed include:
|
||||
- Download a pretrained image classification model from TensorFlow Hub.
|
||||
- Create a serving function to receive compressed image data, and output decomopressed preprocessed data for the model input.
|
||||
- Upload the TensorFlow Hub model and serving function as a `Vertex AI Model` resource.
|
||||
- Creating an `Endpoint` resource.
|
||||
- Deploying the `Model` resource to an `Endpoint` resource.
|
||||
- Make an online prediction to the `Model` resource instance deployed to the `Endpoint` resource.
|
||||
```
|
||||
|
||||
[Get started with FastAPI with Vertex AI Prediction](get_started_with_fastapi.ipynb)
|
||||
|
||||
```
|
||||
The steps performed include:
|
||||
- Download a pretrained image classification model from TensorFlow Hub.
|
||||
- Create a serving function to receive compressed image data, and output decomopressed preprocessed data for the model input.
|
||||
- Upload the TensorFlow Hub model and serving function as a `Vertex AI Model` resource.
|
||||
- Creating an `Endpoint` resource.
|
||||
- Deploying the `Model` resource to an `Endpoint` resource with `FastAPI` custom serving binary.
|
||||
- Make an online prediction to the `Model` resource instance deployed to the `Endpoint` resource.
|
||||
```
|
||||
|
||||
[Get started with Nvidia Triton server](get_started_with_nvidia_triton_serving.ipynb)
|
||||
|
||||
```
|
||||
The steps performed in this tutorial include:
|
||||
- Download the model artifacts from TensorFlow Hub.
|
||||
- Create Triton serving configuration file for the model.
|
||||
- Construct a custom container, with Triton serving image, for model deployment.
|
||||
- Upload the model as a `Vertex AI Model` resource.
|
||||
- Deploy the `Vertex AI Model` resource to a `Vertex AI Endpoint` resource.
|
||||
- Make a prediction request
|
||||
- Undeploy the `Model` resource and delete the `Endpoint`
|
||||
|
||||
```
|
||||
|
||||
[Get started with Custom Prediction Routine (CPR)](get_started_with_cpr.ipynb)
|
||||
|
||||
```
|
||||
The steps performed include:
|
||||
- Write a custom data preprocessor.
|
||||
- Train the model.
|
||||
- Build a custom scikit-learn serving container with custom data preprocessing using the Custom Prediction Routine model server.
|
||||
@@ -57,140 +95,20 @@ The steps performed include:
|
||||
- Make a prediction request.
|
||||
```
|
||||
|
||||
[Get started with Vertex AI Raw Prediction](get_started_with_raw_predict.ipynb)
|
||||
[Get started with re-importing AutoML tabular models](get_started_automl_tabular_exported_deploy.ipynb)
|
||||
|
||||
```
|
||||
The steps performed include:
|
||||
|
||||
- Download a pretrained tabular classification model artifacts for a TensorFlow 1.x estimator.
|
||||
- Upload the TensorFlow estimator model as a `Vertex AI Model` resource.
|
||||
- Creating an `Endpoint` resource.
|
||||
- Deploying the `Model` resource to an `Endpoint` resource.
|
||||
- Make an online raw prediction to the `Model` resource instance deployed to the `Endpoint` resource.
|
||||
- Importing a pretrained AutoML tabular exported model artifacts, as a `Model` resource.
|
||||
- Create an `Endpoint` resource.
|
||||
- Deploy the `Model` resource to the `Endpoint` resource.
|
||||
- Make a prediction.
|
||||
```
|
||||
|
||||
|
||||
[Get started with serving functions for TensorFlow model](get_started_with_tf_serving_function.ipynb)
|
||||
[Get started with Vertex Explainable AI using custom deployment container](get_started_with_xai_and_custom_server.ipynb)
|
||||
|
||||
```
|
||||
The steps performed include:
|
||||
|
||||
- Download a pretrained image classification model from TensorFlow Hub.
|
||||
- Create a serving function to receive compressed image data, and output decomopressed preprocessed data for the model input.
|
||||
- Upload the TensorFlow Hub model and serving function as a `Vertex AI Model` resource.
|
||||
- Creating an `Endpoint` resource.
|
||||
- Deploying the `Model` resource to an `Endpoint` resource.
|
||||
- Make an online prediction to the `Model` resource instance deployed to the `Endpoint` resource.
|
||||
```
|
||||
|
||||
[Get started with TensorFlow Serving](get_started_with_tf_serving.ipynb)
|
||||
|
||||
```
|
||||
The steps performed include:
|
||||
|
||||
- Download a pretrained image classification model from TensorFlow Hub.
|
||||
- Create a serving function to receive compressed image data, and output decomopressed preprocessed data for the model input.
|
||||
- Upload the TensorFlow Hub model and serving function as a `Vertex AI Model` resource.
|
||||
- Creating an `Endpoint` resource.
|
||||
- Deploying the `Model` resource to an `Endpoint` resource with `TensorFlow Serving` serving binary.
|
||||
- Make an online prediction to the `Model` resource instance deployed to the `Endpoint` resource.
|
||||
```
|
||||
|
||||
[Get started with FastAPI serving binary](get_started_with_fastapi.ipynb)
|
||||
|
||||
```
|
||||
The steps performed include:
|
||||
|
||||
- Download a pretrained image classification model from TensorFlow Hub.
|
||||
- Create a serving function to receive compressed image data, and output decomopressed preprocessed data for the model input.
|
||||
- Upload the TensorFlow Hub model and serving function as a `Vertex AI Model` resource.
|
||||
- Creating an `Endpoint` resource.
|
||||
- Deploying the `Model` resource to an `Endpoint` resource with `FastAPI` custom serving binary.
|
||||
- Make an online prediction to the `Model` resource instance deployed to the `Endpoint` resource.
|
||||
```
|
||||
|
||||
[Get started with Vertex AI Prediction optimized TensorFlow run-time container](get_started_with_optimized_tfe_bert.ipynb)
|
||||
|
||||
```
|
||||
The steps performed include:
|
||||
|
||||
- Download a pretrained BERT model from TensorFlow Hub.
|
||||
- Fine-tune (transfer learning) the BERT model as a binary classifier.
|
||||
- Upload the TensorFlow Hub model as a `Vertex AI Model` resource, with standard TensorFlow serving container.
|
||||
- Upload the TensorFlow Hub model as a `Vertex AI Model` resource, with `Vertex AI Prediction optimized TensorFlow runtime` container
|
||||
- Create two `Endpoint` resources.
|
||||
- Deploying both `Model` resources to separate `Endpoint` resources.
|
||||
- Make the same online prediction requests to both `Model` resource instances deployed to the `Endpoint` resources.
|
||||
- Compare the prediction accuracy between the two deployed `Model` resources.
|
||||
- Configuring container settings for fine-tune control of optimizations.
|
||||
- Create a `Private Endpoint` resource.
|
||||
- Deploy the `Model` resoure with then `Vertex AI Prediction optimized TensorFlow runtime` to the `Private Endpoint` resource.
|
||||
- Make an online prediction request to the `Private Endpoint` resource.
|
||||
```
|
||||
|
||||
[Get started with Nvidia Triton Server](get_started_with_nvidia_triton_server.ipynb)
|
||||
|
||||
```
|
||||
The steps performed in this tutorial include:
|
||||
|
||||
- Download the model artifacts from TensorFlow Hub.
|
||||
- Create Triton serving configuration file for the model.
|
||||
- Construct a custom container, with Triton serving image, for model deployment.
|
||||
- Locally test the custom container.
|
||||
- Upload the model as a Vertex AI Model resource.
|
||||
- Deploy the Vertex AI Model resource to a Vertex AI Endpoint resource.
|
||||
- Make a prediction request
|
||||
- Undeploy the Model resource and delete the Endpoint
|
||||
```
|
||||
|
||||
[Get started with Vertex AI Matching Engine](get_started_with_matching_engine.ipynb)
|
||||
|
||||
```
|
||||
The steps performed include:
|
||||
|
||||
- Create ANN Index.
|
||||
- Create an IndexEndpoint with VPC Network
|
||||
- Deploy ANN Index
|
||||
- Perform online query
|
||||
- Deploy brute force Index.
|
||||
- Perform calibration between ANN and brute force index.
|
||||
```
|
||||
|
||||
[Get started with Vertex AI Matching Engine using Two Towers builtin algorithm](get_started_with_matchine_engine_twotowers.ipynb)
|
||||
|
||||
```
|
||||
The steps performed include:
|
||||
|
||||
1. Train the `Two-Tower` algorithm to generate embeddings (encoder) for the dataset.
|
||||
2. Hyperparameter tune the trained `Two-Tower` encoder.
|
||||
3. Make example predictions (embeddings) from then trained encoder.
|
||||
4. Generate embeddings using the trained `Two-Tower` builtin algorithm.
|
||||
5. Store embeddings to format supported by `Matching Engine`.
|
||||
6. Create a `Matching Engine Index` for the embeddings.
|
||||
7. Deploy the `Matching Engine Index` to a `Index Endpoint`.
|
||||
8. Make a matching engine prediction request.
|
||||
```
|
||||
|
||||
[Get started with Vertex AI Matching Engine using Swivel builtin algorithm](get_started_with_matchine_swivel.ipynb)
|
||||
|
||||
```
|
||||
The steps performed include:
|
||||
|
||||
1. Train the `Swivel` algorithm to generate embeddings (encoder) for the dataset.
|
||||
2. Hyperparameter tune the trained `Swivel` encoder.
|
||||
3. Make example predictions (embeddings) from then trained encoder.
|
||||
4. Generate embeddings using the trained `Swivel` builtin algorithm.
|
||||
5. Store embeddings to format supported by `Matching Engine`.
|
||||
6. Create a `Matching Engine Index` for the embeddings.
|
||||
7. Deploy the `Matching Engine Index` to a `Index Endpoint`.
|
||||
8. Make a matching engine prediction request.
|
||||
```
|
||||
|
||||
[Get started with Explainable AI and custom model server](get_started_with_xai_and_custom_server.ipynb)
|
||||
|
||||
```
|
||||
The steps performed include:
|
||||
|
||||
- Locally train a Pytorch tabular classifier.
|
||||
- Locally test the trained model.
|
||||
- Build a HTTP server using FastAPI.
|
||||
@@ -201,15 +119,89 @@ The steps performed include:
|
||||
- Deploy the `Model` resource to an `Endpoint` resource.
|
||||
- Make a prediction request to the deployed custom serving container.
|
||||
- Make an explanation request to the deployed custom serving container.
|
||||
|
||||
```
|
||||
|
||||
[Get started with reimporting an exported AutoML Tabular model](get_started_automl_tabular_exported_deploy.ipynb)
|
||||
[Get started with TensorFlow serving functions with Vertex AI Raw Prediction](get_started_with_raw_predict.ipynb)
|
||||
|
||||
```
|
||||
The steps performed include:
|
||||
|
||||
- Importing a pretrained AutoML tabular exported model artifacts, as a `Model` resource.
|
||||
- Create an `Endpoint` resource.
|
||||
- Deploy the `Model` resource to the `Endpoint` resource.
|
||||
- Make a prediction.
|
||||
- Download a pretrained tabular classification model artifacts for a TensorFlow 1.x estimator.
|
||||
- Upload the TensorFlow estimator model as a `Vertex AI Model` resource.
|
||||
- Creating an `Endpoint` resource.
|
||||
- Deploying the `Model` resource to an `Endpoint` resource.
|
||||
- Make an online raw prediction to the `Model` resource instance deployed to the `Endpoint` resource.
|
||||
```
|
||||
|
||||
[Get started with Vertex AI Matching Engine and Two Towers builtin algorithm](get_started_with_matching_engine_twotowers.ipynb)
|
||||
|
||||
```
|
||||
The steps performed include:
|
||||
1. Train the `Two-Tower` algorithm to generate embeddings (encoder) for the dataset.
|
||||
2. Hyperparameter tune the trained `Two-Tower` encoder.
|
||||
3. Make example predictions (embeddings) from then trained encoder.
|
||||
4. Generate embeddings using the trained `Two-Tower` builtin algorithm.
|
||||
5. Store embeddings to format supported by `Matching Engine`.
|
||||
6. Create a `Matching Engine Index` for the embeddings.
|
||||
7. Deploy the `Matching Engine Index` to a `Index Endpoint`.
|
||||
8. Make a matching engine prediction request.
|
||||
|
||||
```
|
||||
|
||||
[Get started with Optimized TensorFlow Enterprise container with Vertex AI Prediction / text models](get_started_with_optimized_tfe_bert.ipynb)
|
||||
|
||||
```
|
||||
The steps performed include:
|
||||
- Download a pretrained BERT model from TensorFlow Hub.
|
||||
- Fine-tune (transfer learning) the BERT model as a binary classifier.
|
||||
- Upload the TensorFlow Hub model as a `Vertex AI Model` resource, with standard TensorFlow serving container.
|
||||
- Upload the TensorFlow Hub model as a `Vertex AI Model` resource, with TensorFlow Enterprise Optimized container
|
||||
- Create two `Endpoint` resources.
|
||||
- Deploying both `Model` resources to separate `Endpoint` resources.
|
||||
- Make the same online prediction requests to both `Model` resource instances deployed to the `Endpoint` resources.
|
||||
- Compare the prediction accuracy between the two deployed `Model` resources.
|
||||
- Configuring container settings for fine-tune control of optimizations.
|
||||
- Create a `Private Endpoint` resource.
|
||||
- Deploy the `Model` resoure with then `TensorFlow Enterprise Optimized` to the `Private Endpoint` resource.
|
||||
- Make an online prediction request to the `Private Endpoint` resource.
|
||||
|
||||
```
|
||||
|
||||
[Get started with Vertex AI Matching Engine](get_started_with_matching_engine.ipynb)
|
||||
|
||||
```
|
||||
The steps performed include:
|
||||
- Create ANN Index.
|
||||
- Create an IndexEndpoint with VPC Network
|
||||
- Deploy ANN Index
|
||||
- Perform online query
|
||||
- Deploy brute force Index.
|
||||
- Perform calibration between ANN and brute force index.
|
||||
|
||||
```
|
||||
|
||||
[Get started with Vertex AI Matching Engine and Swivel builtin algorithm](get_started_with_matching_engine_swivel.ipynb)
|
||||
|
||||
```
|
||||
The steps performed include:
|
||||
1. Train the `Swivel` algorithm to generate embeddings (encoder) for the dataset.
|
||||
2. Make example predictions (embeddings) from then trained encoder.
|
||||
3. Generate embeddings using the trained `Swivel` builtin algorithm.
|
||||
4. Store embeddings to format supported by `Matching Engine`.
|
||||
5. Create a `Matching Engine Index` for the embeddings.
|
||||
6. Deploy the `Matching Engine Index` to a `Index Endpoint`.
|
||||
7. Make a matching engine prediction request.
|
||||
|
||||
```
|
||||
|
||||
[Get started with TensorFlow serving with Vertex AI Prediction](get_started_with_tf_serving.ipynb)
|
||||
|
||||
```
|
||||
The steps performed include:
|
||||
- Download a pretrained image classification model from TensorFlow Hub.
|
||||
- Create a serving function to receive compressed image data, and output decomopressed preprocessed data for the model input.
|
||||
- Upload the TensorFlow Hub model and serving function as a `Vertex AI Model` resource.
|
||||
- Creating an `Endpoint` resource.
|
||||
- Deploying the `Model` resource to an `Endpoint` resource with `TensorFlow Serving` serving binary.
|
||||
- Make an online prediction to the `Model` resource instance deployed to the `Endpoint` resource.
|
||||
```
|
||||
Reference in New Issue
Block a user