This article describes the supported versions and compatibility with supported ML frameworks. For some frameworks, links to example repositories on Gitlab are included. These example repositories can be used to create example deployments on Deeploy.
Deeploy | 1.29.x - 1.31.x | 1.32.x + |
---|
KServe | 0.9.0 | 0.11.0 |
Python | 3.7 - 3.9 | 3.8 - 3.11 |
Model Frameworks
The versions described in the following sections are the recommended versions on which the model images have been built. If you deviate from the version, check for potential compatibility issues.
PyTorch
Pytorch default image | pytorch/torchserve-kfs:0.6.0 | pytorch/torchserve-kfs:0.8.0 |
---|
torch | 1.7.1 | 2.0.0 |
torchserve | 0.7.0 | 0.8.0 |
torch-model-archiver | 0.7.0 | 0.8.0 |
Tensorflow
Tensorflow default image | tensorflow/serving:2.6.2 | tensorflow/serving:2.6.2 |
---|
tensorflow | 2.6.2 | 2.6.2 |
XGBoost
XGBoost default image | kserve/xgbserver:v0.9.0 | kserve/xgbserver:v0.11.0 |
---|
xgboost | 1.5.0 | 1.7.5 |
Scikit-learn
Scikit-learn default image | kserve/sklearnserver:v0.9.0 | kserve/sklearnserver:v0.11.0 |
---|
scikit-learn | 1.0.1 | 1.3.0 |
joblib | 1.1.0 | 1.3.1 |
LightGBM
LightGBM default image | kserve/lgbserver:v0.9.0 | kserve/lgbserver:v0.11.0 |
---|
lightgbm | 3.3.2 | 3.3.5 |
Custom model Docker images
Custom model | image-repo/image-name>:tag | image-repo/image-name:tag |
---|
deeploy-cli | - | 0.1.0 |
kserve | 0.9.0 | 0.11.0 |
Standard Explainers
Tree SHAP
Deploy a tree SHAP explainer without training it yourself. Available only for tree-based classification models when using XGBoost, Scikit-learn, or LightGBM model frameworks.
In the case of multiple inputs in single explanation request, the explanation values are returned only for the first input.
For LightGBM models where probabilities are returned, the explanation is for the class with the highest probability.
Saliency
Deploy a saliency based explainer without training it yourself. Available only for text generation and text-to-text generation models when using huggingface model framework. The explainer can be utilized to obtain token importances for generated tokens.
It is derived from following research work Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps (Simonyan et al., 2013)
Attention
Deploy a attention based explainer without training it yourself. Available only for text generation and text-to-text generation models when using huggingface model framework. The explainer can be utilized to obtain token importances for generated tokens.
It is derived from following research work on Attention Weight Attribution, from Neural Machine Translation by Jointly Learning to Align and Translate (Bahdanau et al., 2014)
Trained Explainer Frameworks
The versions described in the following sections are the recommended versions on which the model images have been built. If you deviate from the version, check for potential compatibility issues.
Shap
Shap kernel default image | deeployml/alibi-explainer:v0.9.0-deeploy-1.0.0 | deeployml/alibi-explainer:v0.11.0-deeploy-1.0.0 |
---|
shap | 0.36.0 | 0.42.1 |
dill | 0.3.3 | 0.3.7 |
Anchors
Anchor tabular, text, image default image | deeployml/alibi-explainer:v0.9.0-deeploy-1.0.0 | deeployml/alibi-explainer:v0.11.0-deeploy-1.0.0 |
---|
alibi | 0.6.4 | 0.9.4 |
dill | 0.3.3 | 0.3.7 |
MACE
- Expected pre-trained explainer artefact: dill (explainer.dill)
Anchor tabular, text, image default image | deeployml/alibi-explainer:v0.9.0-deeploy-1.0.0 | deeployml/alibi-explainer:v0.11.0-deeploy-1.0.0 |
---|
omnixai | 1.1.4 | 1.3.1 |
dill | 0.3.3 | 0.3.7 |
PDP
- Expected pre-trained explainer artefact: dill (explainer.dill)
Anchor tabular, text, image default image | deeployml/alibi-explainer:v0.9.0-deeploy-1.0.0 | deeployml/alibi-explainer:v0.11.0-deeploy-1.0.0 |
---|
omnixai | 1.1.4 | 1.3.1 |
dill | 0.3.3 | 0.3.7 |
Custom explainer Docker images
Custom model | image-repo/image-name>:tag | image-repo/image-name:tag |
---|
deeploy | - | 1.3.0 |
kserve | 0.9.0 | 0.11.0 |
Custom docker | image-repo/image-name>:tag | image-repo/image-name:tag |
---|
deeploy | - | 1.3.0 |
kserve | 0.9.0 | 0.11.0 |