torchabc 0.2.0__tar.gz → 0.2.2__tar.gz

{torchabc-0.2.0 → torchabc-0.2.2}/PKG-INFO

@@ -1,7 +1,7 @@
 Metadata-Version: 2.3
 Name: torchabc
-Version: 0.2.0
-Summary: A minimal abstract class for training, evaluation, and inference of pytorch models.
+Version: 0.2.2
+Summary: A simple abstract class for training and inference in PyTorch.
 License: MIT
 Author: Emanuele Guidotti
 Author-email: emanuele.guidotti@usi.ch
@@ -18,21 +18,21 @@ Description-Content-Type: text/markdown
 
 # TorchABC
 
-[`TorchABC`](https://github.com/eguidotti/torchabc) is a minimal abstract class for training, evaluation, and inference of pytorch models that helps you keep your code organized. It depends on [`torch`](https://pypi.org/project/torch/) only and it consists of a simple self-contained [file](https://github.com/eguidotti/torchabc/blob/main/torchabc/__init__.py).
+[`TorchABC`](https://github.com/eguidotti/torchabc/blob/main/torchabc/__init__.py) is an abstract class for training and inference in PyTorch that helps you keep your code well organized. It is a minimalist version of [pytorch-lightning](https://pypi.org/project/pytorch-lightning/), it depends on [torch](https://pypi.org/project/torch/) only, and it consists of a simple self-contained [file](https://github.com/eguidotti/torchabc/blob/main/torchabc/__init__.py).
 
 ## Workflow
 
 ![diagram](https://github.com/user-attachments/assets/f3eac7aa-6a39-4a93-887c-7b7f8ac5f0f4)
 
-`TorchABC` implements the workflow illustrated above. The workflow begins with raw **DATA**, which undergoes a **preprocess** step. This preprocessing step transforms the raw data into **INPUT** samples and their corresponding **TARGET** labels.
+The `TorchABC` class implements the workflow illustrated above. The workflow begins with raw `data`, which undergoes a `preprocess` step. This step transforms the raw `data` into `input` samples and their corresponding `target` labels.
 
-Next, the individual **INPUT** samples are grouped into batches called **INPUTS** using a **collate** function. Similarly, the **TARGET** labels are batched into **TARGETS**. The **INPUTS** are then fed into the neural **network**, which produces **OUTPUTS**.
+Next, the individual `input` samples are grouped into batches called `inputs` using a `collate` function. Similarly, the `target` labels are batched into `targets`. The `inputs` are then fed into the `network`, which produces `outputs`.
 
-The **OUTPUTS** of the network are compared to the **TARGETS** using a **LOSS** function. The calculated loss quantifies the error between the model's predictions and the actual targets. This **LOSS** is then used by the **optimizer / scheduler** to update the parameters of the **network** to minimize the loss. The optimizer controls how the parameters are adjusted, while the scheduler can dynamically change the learning rate of the optimizer during training.
+The `outputs` are compared to the `targets` using a `loss` function which quantifies the error between the two. The `optimizer` updates the parameters of the `network` to minimize the `loss`. The `scheduler` can dynamically change the learning rate of the `optimizer` during training.
 
-Finally, the raw **OUTPUTS** from the network undergo a **postprocess** step to generate the final **PREDICTIONS**. This could involve converting probabilities to class labels, applying thresholds, or other task-specific transformations. 
+Finally, the raw `outputs` from the `network` undergo a `postprocess` step to generate the final `predictions`. This could involve converting probabilities to class labels, applying thresholds, or other task-specific transformations. 
 
-**The core logic blocks** are abstract. You define their specific behavior with maximum flexibility. 
+The core logic blocks are abstract. You define their specific behavior with maximum flexibility. 
 
 ## Quick start
 
@@ -82,7 +82,7 @@ class ClassName(TorchABC):
         return outputs
 ```
 
-Fill out the template with the dataloaders, preprocessing and postprocessing steps, the neural network, optimizer, scheduler, loss, and evaluation metrics. 
+Fill out the template with the dataloaders, preprocessing and postprocessing steps, the neural network, optimizer, scheduler, loss, and optional evaluation metrics. 
 
 #### `dataloaders`: the dataloaders for training and evaluation
 

{torchabc-0.2.0 → torchabc-0.2.2}/README.md

@@ -1,20 +1,20 @@
 # TorchABC
 
-[`TorchABC`](https://github.com/eguidotti/torchabc) is a minimal abstract class for training, evaluation, and inference of pytorch models that helps you keep your code organized. It depends on [`torch`](https://pypi.org/project/torch/) only and it consists of a simple self-contained [file](https://github.com/eguidotti/torchabc/blob/main/torchabc/__init__.py).
+[`TorchABC`](https://github.com/eguidotti/torchabc/blob/main/torchabc/__init__.py) is an abstract class for training and inference in PyTorch that helps you keep your code well organized. It is a minimalist version of [pytorch-lightning](https://pypi.org/project/pytorch-lightning/), it depends on [torch](https://pypi.org/project/torch/) only, and it consists of a simple self-contained [file](https://github.com/eguidotti/torchabc/blob/main/torchabc/__init__.py).
 
 ## Workflow
 
 ![diagram](https://github.com/user-attachments/assets/f3eac7aa-6a39-4a93-887c-7b7f8ac5f0f4)
 
-`TorchABC` implements the workflow illustrated above. The workflow begins with raw **DATA**, which undergoes a **preprocess** step. This preprocessing step transforms the raw data into **INPUT** samples and their corresponding **TARGET** labels.
+The `TorchABC` class implements the workflow illustrated above. The workflow begins with raw `data`, which undergoes a `preprocess` step. This step transforms the raw `data` into `input` samples and their corresponding `target` labels.
 
-Next, the individual **INPUT** samples are grouped into batches called **INPUTS** using a **collate** function. Similarly, the **TARGET** labels are batched into **TARGETS**. The **INPUTS** are then fed into the neural **network**, which produces **OUTPUTS**.
+Next, the individual `input` samples are grouped into batches called `inputs` using a `collate` function. Similarly, the `target` labels are batched into `targets`. The `inputs` are then fed into the `network`, which produces `outputs`.
 
-The **OUTPUTS** of the network are compared to the **TARGETS** using a **LOSS** function. The calculated loss quantifies the error between the model's predictions and the actual targets. This **LOSS** is then used by the **optimizer / scheduler** to update the parameters of the **network** to minimize the loss. The optimizer controls how the parameters are adjusted, while the scheduler can dynamically change the learning rate of the optimizer during training.
+The `outputs` are compared to the `targets` using a `loss` function which quantifies the error between the two. The `optimizer` updates the parameters of the `network` to minimize the `loss`. The `scheduler` can dynamically change the learning rate of the `optimizer` during training.
 
-Finally, the raw **OUTPUTS** from the network undergo a **postprocess** step to generate the final **PREDICTIONS**. This could involve converting probabilities to class labels, applying thresholds, or other task-specific transformations. 
+Finally, the raw `outputs` from the `network` undergo a `postprocess` step to generate the final `predictions`. This could involve converting probabilities to class labels, applying thresholds, or other task-specific transformations. 
 
-**The core logic blocks** are abstract. You define their specific behavior with maximum flexibility. 
+The core logic blocks are abstract. You define their specific behavior with maximum flexibility. 
 
 ## Quick start
 
@@ -64,7 +64,7 @@ class ClassName(TorchABC):
         return outputs
 ```
 
-Fill out the template with the dataloaders, preprocessing and postprocessing steps, the neural network, optimizer, scheduler, loss, and evaluation metrics. 
+Fill out the template with the dataloaders, preprocessing and postprocessing steps, the neural network, optimizer, scheduler, loss, and optional evaluation metrics. 
 
 #### `dataloaders`: the dataloaders for training and evaluation
 

{torchabc-0.2.0 → torchabc-0.2.2}/pyproject.toml

@@ -1,7 +1,7 @@
 [project]
 name = "torchabc"
-version = "0.2.0"
-description = "A minimal abstract class for training, evaluation, and inference of pytorch models."
+version = "0.2.2"
+description = "A simple abstract class for training and inference in PyTorch."
 authors = [
     {name = "Emanuele Guidotti",email = "emanuele.guidotti@usi.ch"}
 ]

{torchabc-0.2.0 → torchabc-0.2.2}/torchabc/__init__.py

@@ -7,7 +7,7 @@ from typing import Any, Union, Dict, List, Callable
 
 class TorchABC(abc.ABC):
     """
-    A minimal abstract class for training, evaluation, and inference of pytorch models.
+    A simple abstract class for training and inference in PyTorch.
     """
 
     def __init__(self, device: Union[str, torch.device] = None, logger: Callable = print, **hparams):

{torchabc-0.2.0 → torchabc-0.2.2}/torchabc/cli.py

@@ -7,7 +7,7 @@ from . import TorchABC
 def main():
     parser = argparse.ArgumentParser(description="Generate template.")
     parser.add_argument('--create', type=str, required=True, help='The path to create the template file.')
-    parser.add_argument('--minimal', action='store_true', help='Generate a minimal template without docstrings.')
+    parser.add_argument('--min', action='store_true', help='Generate a minimalistic template without docstrings.')
     args = parser.parse_args()
 
     cached_properties = {}
@@ -36,9 +36,9 @@ from typing import Any, Dict
 
 class ClassName(TorchABC):"""
 
-    if not args.minimal:
+    if not args.min:
         template += """
-    \"\"\"A concrete implementation of the TorchABC abstract class.
+    \"\"\"A concrete subclass of the TorchABC abstract class.
 
     Use this template to implement your own model by following these steps:
     - replace ClassName with the name of your model,
@@ -56,7 +56,7 @@ class ClassName(TorchABC):"""
             template += f"""
     @cached_property
     def {name}(self):"""
-            if not args.minimal:
+            if not args.min:
                 template += f"""
         \"\"\"{doc}\"\"\""""
             template += f"""
@@ -66,14 +66,14 @@ class ClassName(TorchABC):"""
             sig, doc = methods[name]
             template += f"""
     def {name}{sig}:"""
-            if not args.minimal:
+            if not args.min:
                 template += f"""
         \"\"\"{doc}\"\"\""""
             template += f"""
         {defaults[name] if name in defaults else 'raise NotImplementedError'}
     """
     
-    if not args.minimal:
+    if not args.min:
         template += f"""
 
 if __name__ == "__main__":