Skip to content

Commit

Permalink
Fix feature-store-api and machine-learning-api links
Browse files Browse the repository at this point in the history
  • Loading branch information
aversey committed Nov 11, 2024
1 parent c6c56c9 commit 2c7d5d7
Show file tree
Hide file tree
Showing 24 changed files with 181 additions and 223 deletions.
17 changes: 6 additions & 11 deletions docs/js/inject-api-links.js
Original file line number Diff line number Diff line change
@@ -1,15 +1,12 @@
window.addEventListener("DOMContentLoaded", function () {
var windowPathNameSplits = window.location.pathname.split("/");
var majorVersionRegex = new RegExp("(\\d+[.]\\d+)")
var latestRegex = new RegExp("latest")
if (majorVersionRegex.test(windowPathNameSplits[1])) { // On landing page docs.hopsworks.api/3.0 - URL contains major version
var majorVersionRegex = new RegExp("(\\d+[.]\\d+)");
var latestRegex = new RegExp("latest");
if (majorVersionRegex.test(windowPathNameSplits[1])) { // On landing page docs.hopsworks.api/4.0 - URL contains major version
// Version API dropdown
document.getElementById("hopsworks_api_link").href = "https://docs.hopsworks.ai/hopsworks-api/" + windowPathNameSplits[1] + "/generated/api/login/";
document.getElementById("hsfs_api_link").href = "https://docs.hopsworks.ai/feature-store-api/" + windowPathNameSplits[1] + "/generated/api/connection_api/";
document.getElementById("hsfs_javadoc_link").href = "https://docs.hopsworks.ai/feature-store-api/" + windowPathNameSplits[1] + "/javadoc";
document.getElementById("hsml_api_link").href = "https://docs.hopsworks.ai/machine-learning-api/" + windowPathNameSplits[1] + "/generated/connection_api/";
} else { // on docs.hopsworks.api/feature-store-api/3.0 / docs.hopsworks.api/hopsworks-api/3.0 / docs.hopsworks.api/machine-learning-api/3.0

document.getElementById("hsfs_javadoc_link").href = "https://docs.hopsworks.ai/hopsworks-api/" + windowPathNameSplits[1] + "/javadoc";
} else { // on / docs.hopsworks.api/hopsworks-api/4.0
if (latestRegex.test(windowPathNameSplits[2]) || latestRegex.test(windowPathNameSplits[1])) {
var majorVersion = "latest";
} else {
Expand All @@ -26,8 +23,6 @@ window.addEventListener("DOMContentLoaded", function () {
document.getElementsByClassName("md-tabs__link")[6].href = "https://docs.hopsworks.ai/" + majorVersion + "/admin/";
// Version API dropdown
document.getElementById("hopsworks_api_link").href = "https://docs.hopsworks.ai/hopsworks-api/" + majorVersion + "/generated/api/login/";
document.getElementById("hsfs_api_link").href = "https://docs.hopsworks.ai/feature-store-api/" + majorVersion + "/generated/api/connection_api/";
document.getElementById("hsfs_javadoc_link").href = "https://docs.hopsworks.ai/feature-store-api/" + majorVersion + "/javadoc";
document.getElementById("hsml_api_link").href = "https://docs.hopsworks.ai/machine-learning-api/" + majorVersion + "/generated/connection_api/";
document.getElementById("hsfs_javadoc_link").href = "https://docs.hopsworks.ai/hopsworks-api/" + majorVersion + "/javadoc";
}
});
37 changes: 0 additions & 37 deletions docs/reference_guides/index.md

This file was deleted.

30 changes: 15 additions & 15 deletions docs/setup_installation/common/arrow_flight_duckdb.md
Original file line number Diff line number Diff line change
Expand Up @@ -3,16 +3,16 @@ By default, Hopsworks uses big data technologies (Spark or Hive) to create train
This is great for large datasets, but for small or moderately sized datasets (think of the size of data that would fit in a Pandas
DataFrame in your local Python environment), the overhead of starting a Spark or Hive job and doing distributed data processing can be significant.

ArrowFlight Server with DuckDB significantly reduces the time that Python clients need to read feature groups
ArrowFlight Server with DuckDB significantly reduces the time that Python clients need to read feature groups
and batch inference data from the Feature Store, as well as creating moderately-sized in-memory training datasets.

When the service is enabled, clients will automatically use it for the following operations:

- [reading Feature Groups](https://docs.hopsworks.ai/feature-store-api/{{{ hopsworks_version }}}/generated/api/feature_group_api/#read)
- [reading Queries](https://docs.hopsworks.ai/feature-store-api/{{{ hopsworks_version }}}/generated/api/query_api/#read)
- [reading Training Datasets](https://docs.hopsworks.ai/feature-store-api/{{{ hopsworks_version }}}/generated/api/feature_view_api/#get_training_data)
- [creating In-Memory Training Datasets](https://docs.hopsworks.ai/feature-store-api/{{{ hopsworks_version }}}/generated/api/feature_view_api/#training_data)
- [reading Batch Inference Data](https://docs.hopsworks.ai/feature-store-api/{{{ hopsworks_version }}}/generated/api/feature_view_api/#get_batch_data)
- [reading Feature Groups](https://docs.hopsworks.ai/hopsworks-api/{{{ hopsworks_version }}}/generated/api/feature_group_api/#read)
- [reading Queries](https://docs.hopsworks.ai/hopsworks-api/{{{ hopsworks_version }}}/generated/api/query_api/#read)
- [reading Training Datasets](https://docs.hopsworks.ai/hopsworks-api/{{{ hopsworks_version }}}/generated/api/feature_view_api/#get_training_data)
- [creating In-Memory Training Datasets](https://docs.hopsworks.ai/hopsworks-api/{{{ hopsworks_version }}}/generated/api/feature_view_api/#training_data)
- [reading Batch Inference Data](https://docs.hopsworks.ai/hopsworks-api/{{{ hopsworks_version }}}/generated/api/feature_view_api/#get_batch_data)

For larger datasets, clients can still make use of the Spark/Hive backend by explicitly setting
`read_options={"use_hive": True}`.
Expand All @@ -21,9 +21,9 @@ For larger datasets, clients can still make use of the Spark/Hive backend by exp

!!! note
Supported only on AWS at the moment.

The ArrowFlight Server is co-located with RonDB in the Hopsworks cluster.
If the ArrowFlight Server is activated, RonDB and ArrowFlight Server can each use up to 50%
If the ArrowFlight Server is activated, RonDB and ArrowFlight Server can each use up to 50%
of the available resources on the node, so they can co-exist without impacting each other.
Just like RonDB, the ArrowFlight Server can be replicated across multiple nodes to serve more clients at lower latency.
To guarantee high performance, each individual ArrowFlight Server instance processes client requests sequentially.
Expand All @@ -42,12 +42,12 @@ To deploy ArrowFlight Server on a cluster:
2. Select an instance type with at least 16GB of memory and 4 cores. (*)
3. Tick the checkbox `Enable ArrowFlight Server`.

(*) The service should have at least the 2x the amount of memory available that a typical Python client would have.
Because RonDB and ArrowFlight Server share the same node we recommend selecting an instance type with at least 4x the
client memory. For example, if the service serves Python clients with typically 4GB of memory,
an instance with at least 16GB of memory should be selected.
An instance with 16GB of memory will be able to read feature groups and training datasets of up to 10-100M rows,
depending on the number of columns and size of the features (~2GB in parquet). The same instance will be able to create
point-in-time correct training datasets with 1-10M rows, also depending on the number and the size of the features.
(*) The service should have at least the 2x the amount of memory available that a typical Python client would have.
Because RonDB and ArrowFlight Server share the same node we recommend selecting an instance type with at least 4x the
client memory. For example, if the service serves Python clients with typically 4GB of memory,
an instance with at least 16GB of memory should be selected.
An instance with 16GB of memory will be able to read feature groups and training datasets of up to 10-100M rows,
depending on the number of columns and size of the features (~2GB in parquet). The same instance will be able to create
point-in-time correct training datasets with 1-10M rows, also depending on the number and the size of the features.
Larger instances are able to handle larger datasets. The numbers scale roughly linearly with the instance size.

20 changes: 10 additions & 10 deletions docs/user_guides/fs/compute_engines.md
Original file line number Diff line number Diff line change
Expand Up @@ -20,15 +20,15 @@ Hopsworks is aiming to provide functional parity between the computational engin

| Functionality | Method | Spark | Python | Flink | Beam | Comment |
| ----------------------------------------------------------------- | ------ | ----- | ------ | ------ | ------ | ------- |
| Feature Group Creation from dataframes | [`FeatureGroup.create_feature_group()`](https://docs.hopsworks.ai/feature-store-api/{{{ hopsworks_version }}}/generated/api/feature_group_api/#create_feature_group) | :white_check_mark: | :white_check_mark: | - | - | Currently Flink/Beam doesn't support registering feature group metadata. Thus it needs to be pre-registered before you can write real time features computed by Flink/Beam.|
| Training Dataset Creation from dataframes | [`TrainingDataset.save()`](https://docs.hopsworks.ai/feature-store-api/{{{ hopsworks_version }}}/generated/api/training_dataset_api/#save) | :white_check_mark: | - | - | - | Functionality was deprecated in version 3.0 |
| Data validation using Great Expectations for streaming dataframes | [`FeatureGroup.validate()`](https://docs.hopsworks.ai/feature-store-api/{{{ hopsworks_version }}}/generated/api/feature_group_api/#validate) [`FeatureGroup.insert_stream()`](https://docs.hopsworks.ai/feature-store-api/{{{ hopsworks_version }}}/generated/api/feature_group_api/#insert_stream) | - | - | - | - | `insert_stream` does not perform any data validation even when a expectation suite is attached. |
| Stream ingestion | [`FeatureGroup.insert_stream()`](https://docs.hopsworks.ai/feature-store-api/{{{ hopsworks_version }}}/generated/api/feature_group_api/#insert_stream) | :white_check_mark: | - | :white_check_mark: | :white_check_mark: | Python/Pandas/Polars has currently no notion of streaming. |
| Stream ingestion | [`FeatureGroup.insert_stream()`](https://docs.hopsworks.ai/feature-store-api/{{{ hopsworks_version }}}/generated/api/feature_group_api/#insert_stream) | :white_check_mark: | - | :white_check_mark: | :white_check_mark: | Python/Pandas/Polars has currently no notion of streaming. |
| Reading from Streaming Storage Connectors | [`KafkaConnector.read_stream()`](https://docs.hopsworks.ai/feature-store-api/{{{ hopsworks_version }}}/generated/api/storage_connector_api/#read_stream) | :white_check_mark: | - | - | - | Python/Pandas/Polars has currently no notion of streaming. For Flink/Beam only write operations are supported |
| Reading training data from external storage other than S3 | [`FeatureView.get_training_data()`](https://docs.hopsworks.ai/feature-store-api/{{{ hopsworks_version }}}/generated/api/feature_view_api/#get_training_data) | :white_check_mark: | - | - | - | Reading training data that was written to external storage using a Storage Connector other than S3 can currently not be read using HSFS APIs, instead you will have to use the storage's native client. |
| Reading External Feature Groups into Dataframe | [`ExternalFeatureGroup.read()`](https://docs.hopsworks.ai/feature-store-api/{{{ hopsworks_version }}}/generated/api/external_feature_group_api/#read) | :white_check_mark: | - | - | - | Reading an External Feature Group directly into a Pandas/Polars Dataframe is not supported, however, you can use the [Query API](https://docs.hopsworks.ai/feature-store-api/{{{ hopsworks_version }}}/generated/api/query_api/) to create Feature Views/Training Data containing External Feature Groups. |
| Read Queries containing External Feature Groups into Dataframe | [`Query.read()`](https://docs.hopsworks.ai/feature-store-api/{{{ hopsworks_version }}}/generated/api/query_api/#read) | :white_check_mark: | - | - | - | Reading a Query containing an External Feature Group directly into a Pandas/Polars Dataframe is not supported, however, you can use the Query to create Feature Views/Training Data and write the data to a Storage Connector, from where you can read up the data into a Pandas/Polars Dataframe. |
| Feature Group Creation from dataframes | [`FeatureGroup.create_feature_group()`](https://docs.hopsworks.ai/hopsworks-api/{{{ hopsworks_version }}}/generated/api/feature_group_api/#create_feature_group) | :white_check_mark: | :white_check_mark: | - | - | Currently Flink/Beam doesn't support registering feature group metadata. Thus it needs to be pre-registered before you can write real time features computed by Flink/Beam.|
| Training Dataset Creation from dataframes | [`TrainingDataset.save()`](https://docs.hopsworks.ai/hopsworks-api/{{{ hopsworks_version }}}/generated/api/training_dataset_api/#save) | :white_check_mark: | - | - | - | Functionality was deprecated in version 3.0 |
| Data validation using Great Expectations for streaming dataframes | [`FeatureGroup.validate()`](https://docs.hopsworks.ai/hopsworks-api/{{{ hopsworks_version }}}/generated/api/feature_group_api/#validate) [`FeatureGroup.insert_stream()`](https://docs.hopsworks.ai/hopsworks-api/{{{ hopsworks_version }}}/generated/api/feature_group_api/#insert_stream) | - | - | - | - | `insert_stream` does not perform any data validation even when a expectation suite is attached. |
| Stream ingestion | [`FeatureGroup.insert_stream()`](https://docs.hopsworks.ai/hopsworks-api/{{{ hopsworks_version }}}/generated/api/feature_group_api/#insert_stream) | :white_check_mark: | - | :white_check_mark: | :white_check_mark: | Python/Pandas/Polars has currently no notion of streaming. |
| Stream ingestion | [`FeatureGroup.insert_stream()`](https://docs.hopsworks.ai/hopsworks-api/{{{ hopsworks_version }}}/generated/api/feature_group_api/#insert_stream) | :white_check_mark: | - | :white_check_mark: | :white_check_mark: | Python/Pandas/Polars has currently no notion of streaming. |
| Reading from Streaming Storage Connectors | [`KafkaConnector.read_stream()`](https://docs.hopsworks.ai/hopsworks-api/{{{ hopsworks_version }}}/generated/api/storage_connector_api/#read_stream) | :white_check_mark: | - | - | - | Python/Pandas/Polars has currently no notion of streaming. For Flink/Beam only write operations are supported |
| Reading training data from external storage other than S3 | [`FeatureView.get_training_data()`](https://docs.hopsworks.ai/hopsworks-api/{{{ hopsworks_version }}}/generated/api/feature_view_api/#get_training_data) | :white_check_mark: | - | - | - | Reading training data that was written to external storage using a Storage Connector other than S3 can currently not be read using HSFS APIs, instead you will have to use the storage's native client. |
| Reading External Feature Groups into Dataframe | [`ExternalFeatureGroup.read()`](https://docs.hopsworks.ai/hopsworks-api/{{{ hopsworks_version }}}/generated/api/external_feature_group_api/#read) | :white_check_mark: | - | - | - | Reading an External Feature Group directly into a Pandas/Polars Dataframe is not supported, however, you can use the [Query API](https://docs.hopsworks.ai/hopsworks-api/{{{ hopsworks_version }}}/generated/api/query_api/) to create Feature Views/Training Data containing External Feature Groups. |
| Read Queries containing External Feature Groups into Dataframe | [`Query.read()`](https://docs.hopsworks.ai/hopsworks-api/{{{ hopsworks_version }}}/generated/api/query_api/#read) | :white_check_mark: | - | - | - | Reading a Query containing an External Feature Group directly into a Pandas/Polars Dataframe is not supported, however, you can use the Query to create Feature Views/Training Data and write the data to a Storage Connector, from where you can read up the data into a Pandas/Polars Dataframe. |

## Python

Expand Down Expand Up @@ -64,7 +64,7 @@ Connecting to the Feature Store from an external Flink cluster, such as GCP Data

### Inside Hopsworks

Beam is only supported as an external client.
Beam is only supported as an external client.

### Outside Hopsworks

Expand Down
Loading

0 comments on commit 2c7d5d7

Please sign in to comment.