fix: Fix broken default installation (#127)

* fix: Add guard for onnx related package import

* build: Update version and changelog

* docs: Update minor errors

* docs: Minor docs improvements

* docs: Minor doc fix

* docs: Update changelog

* refactor: Disable interactive backend for matplotlib

* feat: Make quadra fail if mixed precision export doesn't terminate correctly

* docs: Update changelog

CHANGELOG.md

@@ -2,6 +2,18 @@
 # Changelog
 All notable changes to this project will be documented in this file.
 
+### [2.2.2]
+
+#### Updated
+
+- Terminate quadra with error if it's not possible to export an ONNX model with automatic mixed precision instead of falling back to full precision
+
+#### Fixed
+
+- Fix default quadra installation requiring extra dependencies incorrectly
+- Fix matplotlib using interactive backend by default
+- Fix documentation errors
+
 ### [2.2.1]
 
 #### Updated

README.md

@@ -172,7 +172,7 @@ On the same dataset we can run a simple SSL training using the BYOL algorithm wi
 quadra experiment=generic/imagenette/ssl/byol logger=csv
 ```
 
-BYOL is not the only SSL algorithm available, you can find a list of all the available algorithms under `quadra/experiment/generic/imagenette/ssl` folder.
+BYOL is not the only SSL algorithm available, you can find a list of all the available algorithms under `quadra/configs/experiment/generic/imagenette/ssl` folder.
 
 
 ### Anomaly Detection Training
@@ -185,11 +185,11 @@ quadra experiment=generic/mnist/anomaly/padim logger=csv
 
 This will run an anomaly detection considering on of the classes as good (default is the number 9) and the rest as anomalies.
 
-This will make use of the [anomalib](https://github.com/openvinotoolkit/anomalib) library to train the model. Many different algorithms are available, you can find them under `quadra/experiment/generic/mnist/anomaly` folder.
+This will make use of the [anomalib](https://github.com/openvinotoolkit/anomalib) library to train the model. Many different algorithms are available, you can find them under `quadra/configs/experiment/generic/mnist/anomaly` folder.
 
 ## Running with Custom Datasets
 
-Each task comes with a default configuration file that can be customized for your needs. Each example experiment we have seen so far uses a default configuration file that can be found under `quadra/experiment/base/<task>/<config_name>.yaml`. 
+Each task comes with a default configuration file that can be customized for your needs. Each example experiment we have seen so far uses a default configuration file that can be found under `quadra/configs/experiment/base/<task>/<config_name>.yaml`. 
 
 Let's see how we can customize the configuration file to run the classification experiment on a custom dataset.
 

docs/tutorials/configurations.md

@@ -231,7 +231,7 @@ callbacks:
 
 In the experiment configuration we aggregate the various building blocks of the framework using the `defaults` key. In this case we are using the `classification` datamodule, a `resnet18` backbone, the `cross_entropy` loss, the `classification` model (Lightning Module), `adam` as optimizer, the `classification` task, the reduce on plateau (`rop`) scheduler and the `default_resize` transform.
 
- We can also see that we are overriding some of the parameters of the different modules. For example, we are overriding the `lr_scheduler_interval` of the backbone to be `epoch` instead of `step`. We are also overriding the `max_epochs` of the trainer to be 200 instead of the default value.
+ We can also see that we are overriding some of the parameters of the different modules. For example, we are overriding the `lr_scheduler_interval` of the backbone to be `step` instead of `epoch`. We are also overriding the `max_epochs` of the trainer to be 200 instead of the default value.
 
 The experiment is the most important configuration file as it is the one actually telling the framework what to do!
 
@@ -307,7 +307,7 @@ classifier:
   out_features: ${model.num_classes}
 module:
   _target_: quadra.modules.classification.ClassificationModule
-  lr_scheduler_interval: "epoch"
+  lr_scheduler_interval: "step"
   criterion: ${loss}
   gradcam: true
 ```

docs/tutorials/examples/anomaly_detection.md

@@ -31,7 +31,7 @@ MNIST/
 
 MNIST doesn't have ground truth masks for defects, by default we will use empty masks for good images and full white masks for anomalies.
 
-The standard datamodule configuration for anomaly is found under `datamodule/anomaly/base.yaml`.
+The standard datamodule configuration for anomaly is found under `datamodule/base/anomaly.yaml`.
 
 ```yaml
 _target_: quadra.datamodules.AnomalyDataModule

docs/tutorials/examples/segmentation.md

@@ -218,15 +218,15 @@ defaults:
 Assuming that you have created a virtual environment and installed the `quadra` library, you can run the experiment by running the following command:
 
 ```bash
-quadra experiment=custom_experiment/smp_multiclass
+quadra experiment=segmentation/custom_experiment/smp_multiclass
 ```
 
 ### Run (Advanced)
 
 Lets assume that you would like to run the experiment with different models and with/without freezing the encoder part of the model, thanks to `hydra` multi-run option we can run as follows:
 
 ```bash
-quadra experiment=custom_experiment/smp_multiclass \
+quadra experiment=segmentation/custom_experiment/smp_multiclass \
 backbone.model.arch=unet,unetplusplus \
 backbone.model.encoder_name=resnet18,resnet50 \
 backbone.model.freeze_encoder=False,True \

docs/tutorials/examples/ssl.md

@@ -63,7 +63,7 @@ We will make some changes to the datamodule in the experiment configuration file
 
 ### Experiment
 
-First, let's check how base experiment configuration file is defined for BYOL algorithm located in `configs/experiment/base/ssl/BYOL.yaml`.
+First, let's check how base experiment configuration file is defined for BYOL algorithm located in `configs/experiment/base/ssl/byol.yaml`.
 
 ```yaml
 # @package _global_

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "quadra"
-version = "2.2.1"
+version = "2.2.2"
 description = "Deep Learning experiment orchestration library"
 authors = [
 	"Federico Belotti <[email protected]>",

quadra/__init__.py

@@ -1,4 +1,4 @@
-__version__ = "2.2.1"
+__version__ = "2.2.2"
 
 
 def get_version():

quadra/main.py

@@ -1,6 +1,7 @@
 import time
 
 import hydra
+import matplotlib
 from omegaconf import DictConfig
 from pytorch_lightning import seed_everything
 
@@ -12,6 +13,8 @@
 load_envs()
 register_resolvers()
 
+
+matplotlib.use("Agg")
 log = get_logger(__name__)
 
 

quadra/utils/export.py

@@ -7,7 +7,6 @@
 import torch
 from anomalib.models.cflow import CflowLightning
 from omegaconf import DictConfig, ListConfig, OmegaConf
-from onnxconverter_common import auto_convert_mixed_precision
 from torch import nn
 
 from quadra.models.base import ModelSignatureWrapper
@@ -22,6 +21,7 @@
 try:
     import onnx  # noqa
     from onnxsim import simplify as onnx_simplify  # noqa
+    from onnxconverter_common import auto_convert_mixed_precision  # noqa
 
     ONNX_AVAILABLE = True
 except ImportError:
@@ -385,8 +385,9 @@ def _safe_export_half_precision_onnx(
             onnx.checker.check_model(onnx_model)
             return True
         except Exception as e:
-            log.debug("Failed to export model with mixed precision with error: %s", e)
-            return False
+            raise RuntimeError(
+                "Failed to export model with automatic mixed precision, check your model or disable ONNX export"
+            ) from e
     else:
         log.info("Exported half precision ONNX model does not contain NaN values, model is stable")
         return True