Add Galaxy wrapper for BiaPy tool #7168

danifranco · 2025-07-23T12:48:55Z

Hi all,

I'm submitting this initial version of a Galaxy wrapper for BiaPy, a deep learning framework for bioimage analysis.
This tool is designed to run a complete BiaPy workflow with two main usage modes:

Use a custom YAML configuration: Users can upload their own YAML config file and run BiaPy directly.
Create the YAML interactively: The tool generates a config file based on user inputs through a set of guided questions (conditional logic).

I’ve been working on this for a few days and, as this is my first Galaxy tool, I have some questions and doubts that I would really appreciate your feedback on:

Parameter help text. Is there a better way to provide detailed help for parameters, other than using long help="..." strings directly in the XML? The inputs have quite a bit of context, so I'm wondering if there's a cleaner way to manage this.
Conda packaging. I’m in the process of adding BiaPy to conda-forge (PR here). Once it’s accepted, can I declare it in the requirements section like this?
```
<requirement type="package" version="3.6.2">biapy</requirement>
```
using the conda-forge channel?
GPU support. BiaPy supports specifying a GPU index as a parameter (e.g., --gpu 0). Training without GPU is very slow, but inference is fine on CPU. What's the best practice in Galaxy for handling GPU support? Should I remove this parameter or keep it optional?
Clickable links in parameter help. Can we include URLs in help="" so they are rendered as clickable links in the Galaxy interface? I did not test how they look like, because there are still things to do in this PR so it passes the checks, and to try it locally, so maybe the are already displayed like links.
Conditional logic depending on two parameters. I’d like to show or hide certain inputs based on the combination of two user choices. From this issue, it seems that’s not yet supported. Is that still the case? If so, I’ve repeated some inputs unnecessarily to work around this, but it would be good to confirm.
GUI-side "check consistency" button. In our current GUI version of BiaPy (outside Galaxy), we allow users to click a button to validate dataset consistency after selecting input folders. Is there a Galaxy-native way to do this (e.g., a button, a validation script or pre-run check)?
EDAM terms. I'm not 100% sure the EDAM operations I selected are correct. Some aren’t even defined in EDAM, but are common DL tasks (like classification, image-to-image translation, etc.). Would it be better to refer to EDAM-bioimaging terms? Or just leave out undefined terms entirely?
Creating the YAML config programmatically. In the second usage mode, I need to generate the YAML file first, then pass it to the BiaPy command. Currently I use a Python script to generate it. Is that acceptable for Galaxy tool execution flow (i.e., intermediate file creation within the command section)?
Testing output folders. Since Galaxy doesn’t support a type="directory" output, I was advised (by ChatGPT) to compress the output folder into a .tar.gz file:
```
<data name="results_archive" format="tar.gz" ... />
```
Is this the correct way to handle directory outputs?

I had two test scenarios in mind:
1. Inference using a custom config (test_data/input_test_data/example.yaml) and pretrained model (just to check basic output).
2. Running with full interactive config creation, to verify YAML creation and complete execution.
I wanted to validate the test by asserting the presence of specific files in the tar archive (e.g. checkpoints, predictions). The full output structure is in test_data/my_experiment_example.tar.gz, and the test data is in test-data/input_test_data/, but I wasn't sure how best to structure this in the test block. In general, I have doubts on how to configure the output and the tests cases so maybe you can guide me a little bit.

Thanks a lot!

@bgruening @iuc-member

bernt-matthias

Few first comments.

tools/biapy/biapy.xml

bernt-matthias · 2025-07-23T13:45:13Z

tools/biapy/biapy.xml

+                        </param>
+                        <when value="biapy_pretrained">
+                            <!-- Q4 -->
+                            <param name="biapy_model_path" type="data" format="data" label="Select the model trained with BiaPy before" help="This question is about setting the path to a pre-trained model trained before using BiaPy. The model should be stored in the folder you previously selected for saving results. Specifically, in the 'checkpoints' folder, there will be a file with a '.pth' extension, which represents the checkpoint of the pre-trained model"/>


do we have a more specific format than data in Galaxy

Same here, I'm not sure how I need to handle this.

The model checkpoint expected here is in state_dict format (.pth extension). Do you have something for that or I can leave it as data as it is? I suppose that I would need to do the ln -s as it is done with images, no?

Can someone confirm this please? @bgruening

We do not have a datatype for these pytorch models. Seems that under the hood these are just pickled files. We try to avoid these files since pickle seems to unsafe: https://stackoverflow.com/questions/25353753/python-can-i-safely-unpickle-untrusted-data

Are there alternative formats that can be used?

My question was more related to how to handle the file. If I need to proceed as with the training/test data.

Can you be more specific?

As there is no specific format for this file extension I can just put a generic data format as the one I was using I asumme.

I need to stress that processing pth file (ie pickle files) is problematic. There is no problem if you work on private infrastructure and only have trustworthy users. But on public infrastructure it may become one because pickle files may allow arbitrary code execution.

It would be really cool if biapy would support a safer format for saving the models.

@danifranco if onnx is not possible, do you think you can support https://github.com/huggingface/safetensors ?

Can you be more specific?

In the same way you explained here: #7168 (comment) . Because this is the way Galaxy handles the input files, no?

Regarding the checkpoint we do not support any other extension but .pth. We load the weights as follows:

torch.load(checkpoint_path, weights_only=True, map_location="cpu")

This restricts deserialization to tensors only, blocking arbitrary Python object execution and greatly reducing pickle-related risks. We also load these weights into a predefined, trusted model architecture (never from user-supplied code).

In the same way you explained here: #7168 (comment) . Because this is the way Galaxy handles the input files, no?

If biapy requires a specific file extensions for any of the files you need to create a symlink before calling the main program, e.g. ln -s $mode_selection.selected_mode.cfg_config_process.pretrained_model.biapy_model_path biapy_model.pth` and then use the symlink.

I have just made a new release of BiaPy supporting safetensors. In the code here I mader changes to always load and generate .safetensors checkpoints. I'm pushing the changes now.

bgruening · 2025-07-23T14:00:31Z

Wuha! This is cool!

I'm submitting this initial version of a Galaxy wrapper for BiaPy, a deep learning framework for bioimage analysis. This tool is designed to run a complete BiaPy workflow with two main usage modes:
1. Use a custom YAML configuration: Users can upload their own YAML config file and run BiaPy directly.

There are no paths in such files, aren't they?

2. Create the YAML interactively: The tool generates a config file based on user inputs through a set of guided questions (conditional logic).

You can use https://docs.galaxyproject.org/en/latest/dev/schema.html#tool-configfiles-configfile for this, it might make the wrapper easier.

I’ve been working on this for a few days and, as this is my first Galaxy tool, I have some questions and doubts that I would really appreciate your feedback on:
1. **Parameter help text**. Is there a better way to provide detailed help for parameters, other than using long help="..." strings directly in the XML? The inputs have quite a bit of context, so I'm wondering if there's a cleaner way to manage this.

The tool help is the best place. You can write this tool help in markdown or rst.

2. **Conda packaging**. I’m in the process of adding BiaPy to `conda-forge` ([PR here](https://github.com/conda-forge/staged-recipes/pull/30597)). Once it’s accepted, can I declare it in the requirements section like this?
   ```
   <requirement type="package" version="3.6.2">biapy</requirement>
   ```
   using the `conda-forge` channel?

Yes! :)

3. **GPU support**. BiaPy supports specifying a GPU index as a parameter (e.g., `--gpu 0`). Training without GPU is very slow, but inference is fine on CPU. What's the best practice in Galaxy for handling GPU support? Should I remove this parameter or keep it optional?

There is no automatic detection in place?

If so use an environment variable. Something like \${GALAXY_GPU:-0} but I guess we also need to disable it dynamically. So maybe \${BIAPY_GPU:-} and then an admin could BIAPY_GPU='--gpu=0'.

4. **Clickable links in parameter help**. Can we include URLs in `help=""` so they are rendered as clickable links in the Galaxy interface? I did not test how they look like, because there are still things to do in this PR so it passes the checks, and to try it locally, so maybe the are already displayed like links.

I think you can, I would not do that but rather use the tool-help.

5. **Conditional logic depending on two parameters**. I’d like to show or hide certain inputs based on the combination of two user choices. From [this issue](https://github.com/galaxyproject/galaxy/issues/13931), it seems that’s not yet supported. Is that still the case? If so, I’ve repeated some inputs unnecessarily to work around this, but it would be good to confirm.

You can have two nested conditionals.

6. **GUI-side "check consistency" button**. In our current GUI version of BiaPy (outside Galaxy), we allow users to click a button to validate dataset consistency after selecting input folders. Is there a Galaxy-native way to do this (e.g., a button, a validation script or pre-run check)?

On which data should this operate? The input data? If so you can add a separate tool that validates the input and then you create a workflow that runs 1. validation and 2. your tool.

7. **EDAM terms**. I'm not 100% sure the EDAM operations I selected are correct. Some aren’t even defined in EDAM, but are common DL tasks (like classification, image-to-image translation, etc.). Would it be better to refer to EDAM-bioimaging terms? Or just leave out undefined terms entirely?

Leave them out for now. @matuskalas has promised to merge those BIII stuff into EDAM at some point :(

8. **Creating the YAML config programmatically**. In the second usage mode, I need to generate the YAML file first, then pass it to the BiaPy command. Currently I use a Python script to generate it. Is that acceptable for Galaxy tool execution flow (i.e., intermediate file creation within the command section)?

It is acceptable, but I think its rather cumbersome for you, I can imagine. Please look at https://docs.galaxyproject.org/en/latest/dev/schema.html#tool-configfiles-configfile ... you can just "fill-out" a YAML file.

9. **Testing output folders**. Since Galaxy doesn’t support a `type="directory"` output, I was advised (by ChatGPT) to compress the output folder into a .tar.gz file:
   ```
   <data name="results_archive" format="tar.gz" ... />
   ```
   Is this the correct way to handle directory outputs?

What is your output? You can put N images for example in a collection. And have a second output with a single output txt file. Outputs can be very complex if you like. And what many people call directory, we call collection - and collections can be structured.

   I had two test scenarios in mind:
   
   1. Inference using a custom config (`test_data/input_test_data/example.yaml`) and pretrained model (just to check basic output).
   2. Running with full interactive config creation, to verify YAML creation and complete execution.
   
   I wanted to validate the test by asserting the presence of specific files in the tar archive (e.g. checkpoints, predictions). The full output structure is in `test_data/my_experiment_example.tar.gz`, and the test data is in` test-data/input_test_data/`, but I wasn't sure how best to structure this in the test block. In general, I have doubts on how to configure the output and the tests cases so maybe you can guide me a little bit.

Can you provide us with a listing of a typical output file and mark what is relevant for users?

Thanks for the tool, that looks very cool!
Bjoern

Thanks a lot!

@bgruening @iuc-member

…calling whenever they are used

danifranco · 2025-07-24T10:46:55Z

Here I go:

Parameter help text.

There are no paths in such files, aren't they?

Oh, true. The paths should be updated as they must match the ones locally. I will redo this part as a config file will always need to be stored.

You can use https://docs.galaxyproject.org/en/latest/dev/schema.html#tool-configfiles-configfile for this, it might make the wrapper easier.

Thank for the hint. Will take a look!

The tool help is the best place. You can write this tool help in markdown or rst.

Oh, perfect!

Conda packaging.

I’m in the process of adding BiaPy to conda-forge (PR here). Once it’s accepted, can I declare it in the requirements section like this?
<requirement type="package" version="3.6.2">biapy</requirement>
using the conda-forge channel?
Yes! :)

It's being accepted in conda so everything is ready to go! (https://anaconda.org/conda-forge/biapy)

GPU support.

There is no automatic detection in place?

If so use an environment variable. Something like \${GALAXY_GPU:-0} but I guess we also need to disable it dynamically. So maybe \${BIAPY_GPU:-} and then an admin could BIAPY_GPU='--gpu=0'.

It's not automatic in BiaPy, it is up to the user. I don't know if I'm understanding correctly how the GPU is selected here, but maybe we can add some check before calling BiaPy to detect if there is a GPU available and if so call BiaPy with the '--gpu 0' . Something like this (assuming the shell is bash):

if nvidia-smi > /dev/null 2>&1; then
    echo "GPU is available"
    gpu=1
else
    echo "No GPU found"
    gpu=0
fi

What do you think?

Clickable links in parameter help.

I think you can, I would not do that but rather use the tool-help.

I made it with markdown so it will be better now I think.

Conditional logic depending on two parameters.

You can have two nested conditionals.

I think I can not use them (I just checked again in ChaGPT). I'm referring to having nested conditionals not inside the <command> section but in the rest of the file. ChaGPT pointed me to the same issue I found galaxyproject/galaxy#13931 . Anyway, I have move those param definitions to the macros.xml so the main xml is more clean.

GUI-side "check consistency" button.

On which data should this operate? The input data? If so you can add a separate tool that validates the input and then you create a workflow that runs 1. validation and 2. your tool.

Yeah, the validation would be done in the inputs paths provided. I mean, that check is a thing that is done inside BiaPy always but just to let the user check it before running anything else. I'd like to have something after the path is given so if it is not possible don't know if creating a tool to check the data will add more complexity. If there is no option to do that but creating another tool maybe we can leave this and see if it will be a good thing to add in the future.

EDAM terms.

Leave them out for now. @matuskalas has promised to merge those BIII stuff into EDAM at some point :(

ok, no problem!

Creating the YAML config programmatically.

It is acceptable, but I think its rather cumbersome for you, I can imagine. Please look at https://docs.galaxyproject.org/en/latest/dev/schema.html#tool-configfiles-configfile ... you can just "fill-out" a YAML file.

Mmmm, I'm not understanding very well how the things are filled there. My approach should work too, no?

Testing output folders.

What is your output? You can put N images for example in a collection. And have a second output with a single output txt file. Outputs can be very complex if you like. And what many people call directory, we call collection - and collections can be structured.

The output is a folder with different files depending on the workflow. For example, for a semantic segmentation workflow the output is something like this:

my_2d_semantic_segmentation/
├── config_files
│   └── my_2d_semantic_segmentation_1.yaml
├── checkpoints
│   └── my_2d_semantic_segmentation_1-checkpoint-best.pth
└── results
    └── my_2d_semantic_segmentation_1
        ├── aug
        │   └── .tif files
        ├── charts
        │   ├── my_2d_semantic_segmentation_1_*.png
        │   └── my_2d_semantic_segmentation_1_loss.png
        ├── per_image
        │   ├── .tif files
        │   └── .zarr files (or.h5)
        ├── per_image_binarized
        │   └── .tif files
        ├── full_image
        │   └── .tif files
        ├── full_image_binarized
        │   └── .tif files
        ├── full_post_processing
        │   └── .tif files
        ├── tensorboard
        └── train_logs

Some of the files are optional so I was thinking on create a test to just check the main files to be generated such as the predictions from the deep learning model that are placed in my_2d_semantic_segmentation/results/my_2d_semantic_segmentation_1/per_image .

Thanks a lot for the support and the quick response!

bgruening · 2025-08-05T14:36:06Z

@danifranco please let us know if you need any help.

The linter complains that you are missing a .shed.yml file. This will also trigger than additional tests.

If you have any questions for the config-file handling let me know.

danifranco · 2025-08-05T14:41:25Z

Hi,

I still need help with the point 3 and 9. The third one maybe is quick but for the 9 I would need a bit more guidance on how to test the output.
Thanks a lot!

bgruening · 2025-08-06T09:39:54Z

For 3) just assume there is a variable (maybe call if BIAPY_GPU_ENABLED) and its set to "true". If this is there just enable GPU. The Galaxy admin can then set this for this tool if needed.

For 9) I would do the following:

Add a user-input option (type=select multiple=true) and let the user specify which output they want. e.g. charts, binarized images, post-processed images ....

And then you define the output collections like this:

tools-iuc/tools/hybpiper/hybpiper.xml

Line 220 in bb37578

Every collection can contain as many files as you want.

I hope that makes sense. Let me know if we can help.

tools/biapy/create_yaml.py

bernt-matthias · 2025-08-18T08:22:02Z

tools/biapy/biapy.xml

-                ln -s '$mode_selection["biapy_model_path"]' ${checkpoint_file} 2>/dev/null || true &&
+            #set $mpath = $mode_selection.get('biapy_model_path')
+            #if $mpath and str($mpath) not in ['None', '']
+                ln -s '$mpath' ${checkpoint_file} 2>/dev/null || true &&


Why are you hiding the errors? Would the following be an option for all the symlink creations:

Suggested change

ln -s '$mpath' ${checkpoint_file} 2>/dev/null || true &&

ln -fs '$mpath' ${checkpoint_file} &&

You are right! I have just added some changes.

…puts, only those that the user selected

…he files exists previously in bash. This will prevent crashing when the users check some outputs even when they are not created

danifranco · 2025-08-19T10:43:48Z

I think I have finally corrected everything. What do you think?

bernt-matthias

Some more comments. Biggest problem I still have with the pickle files.

tools/biapy/.shed.yml

tools/biapy/macros.xml

tools/biapy/biapy.xml

bernt-matthias · 2025-08-21T07:42:35Z

tools/biapy/biapy.xml

+                                <option value="torchvision">TorchVision</option>
+                            </param>
+                            <when value="bmz">
+                                <param name="bmz_model_name" type="text" format="data" value="sensible-cat" label="BioImage Model Zoo model name" help="Enter the name of a pre-trained model from the BioImage Model Zoo (https://bioimage.io/#/models); filter by the BiaPy icon and ensure the model matches your dimensionality (2D/3D) and task (e.g. semantic segmentation)."/>


type="data"?

Would be cool to have tests for more of the possibilities how this tool can be used.

If you use name="model_name" here and in the torchvision case you can simplify the command section (because the parameter can be accessed in the same way in the parameter dict).

I guess you want to add a no_value validator in order to require users to enter something here.

Do we know anything about the names, could we add a regexp validator to avoid users entering malicious text.

Thanks for the comments, I thing I have addressed everything I can in the commits that I will push soon.

Regarding the tests: I have added one more test, although it is quite similar to the second one. At the moment, I am somewhat limited, as I cannot train models on CPU, this would take too much time. Because of this, the tests are restricted to inference only.

In the first test, I demonstrated the branch that allows loading a custom configuration file. The results are not good, since no pretrained weights are being loaded.

In the second test, I showcased the branch for creating a custom configuration file and loading a pretrained model from the BioImage Model Zoo. In this case, the results are good, as the pretrained model I used was originally trained on data similar to my input.

Finally, I added a third test using a different workflow in BiaPy to denoise images. The results here are also not good, again because I am not loading a pretrained model.

The main challenge is that I cannot include a local .pth (model weights) file in the test-data folder, as these files exceed 5MB and would cause the Galaxy tool linting to fail. The only workaround is to configure the tests to load pretrained models directly from the BioImage Model Zoo (BMZ). However, I would prefer not to rely exclusively on this setup, since it depends on the BMZ website, which is not always consistently available and BiaPy has much more to offer than loading always pretrained models from the zoo.

tools/biapy/biapy.xml

bernt-matthias · 2025-08-22T10:31:39Z

tools/biapy/biapy.xml

+            #if 'raw' in $outs
+                ######## 
+                ## RAW #
+                ########
+                && mkdir -p raw && chmod -R 777 raw && {
+                    ## Instance segmentation
+                    if [ -d "output/my_experiment/results/my_experiment_1/per_image_instances" ]; then
+                        cp output/my_experiment/results/my_experiment_1/per_image_instances/* raw/;
+
+                    ## Instance segmentation
+                    elif [ -d "output/my_experiment/results/my_experiment_1/full_image_instances" ]; then
+                        cp output/my_experiment/results/my_experiment_1/full_image_instances/* raw/;
+
+                    ## Semantic segmentation
+                    elif [ -d "output/my_experiment/results/my_experiment_1/per_image_binarized" ]; then
+                        cp output/my_experiment/results/my_experiment_1/per_image_binarized/* raw/;
+
+                    ## Semantic segmentation
+                    elif [ -d "output/my_experiment/results/my_experiment_1/full_image_binarized" ]; then
+                        cp output/my_experiment/results/my_experiment_1/full_image_binarized/* raw/;
+
+                    ## I2I
+                    elif [ -d "output/my_experiment/results/my_experiment_1/full_image" ]; then
+                        cp output/my_experiment/results/my_experiment_1/full_image/* raw/;
+
+                    ## Detection
+                    elif [ -d "output/my_experiment/results/my_experiment_1/per_image_local_max_check" ]; then
+                        cp output/my_experiment/results/my_experiment_1/per_image_local_max_check/* raw/;
+
+                    ## Detection, Denoising, I2I, SSL, SR
+                    elif [ -d "output/my_experiment/results/my_experiment_1/per_image" ]; then
+                        cp output/my_experiment/results/my_experiment_1/per_image/* raw/;


Would it be an option to discover_datasets from the output/my_experiment/results/my_experiment_1/ and use recursive? Then we can save all this copying. Otherwise, since you only use h5 and tiff files we might want to copy only those.

We cannot really do it that way. Each experiment generates additional tiff/h5 files inside subfolders, which is why I originally proposed using the entire results folder as the output. That approach avoids having to reproduce all this logic and also keeps the configuration much simpler. It is also important to note that, in this case with the current if-else commands, the order in which the data is copied must be preserved.

tools/biapy/biapy.xml

bernt-matthias · 2025-08-22T10:33:05Z

tools/biapy/biapy.xml

+                && mkdir -p post_proc && chmod -R 777 post_proc && {
+                    ## Instance segmentation
+                    if [ -d "output/my_experiment/results/my_experiment_1/per_image_post_processing" ]; then
+                        cp output/my_experiment/results/my_experiment_1/per_image_post_processing/* post_proc/;
+
+                    ## Instance segmentation
+                    elif [ -d "output/my_experiment/results/my_experiment_1/full_image_post_processing" ]; then
+                        cp output/my_experiment/results/my_experiment_1/full_image_post_processing/* post_proc/;
+
+                    ## Detection
+                    elif [ -d "output/my_experiment/results/my_experiment_1/per_image_local_max_check_post_proc" ]; then
+                        cp output/my_experiment/results/my_experiment_1/per_image_local_max_check_post_proc/* post_proc/;
+                    fi;
+                }
+            #end if


Same question: discover form the longest common prefix using recursive?

same answer as before

bernt-matthias · 2025-08-22T10:34:12Z

tools/biapy/biapy.xml

+        #if $phase in ['train_test', 'train']:
+            ## Error redirection was necessary because chart creation does not always occur at the default interval of every 5 epochs
+            #if 'tcharts' in $outs
+                && mkdir -p train_charts && chmod -R 777 train_charts && 
+                cp output/my_experiment/results/my_experiment_1/charts/*.png train_charts/ 2>/dev/null || true
+            #end if
+            #if 'tlogs' in $outs
+                && mkdir -p train_logs && chmod -R 777 train_logs && 
+                cp output/my_experiment/train_logs/*.txt train_logs/ 2>/dev/null || true
+            #end if
+        #end if
+
+        #if 'checkpoint' in $outs
+            && mkdir -p checkpoints && chmod -R 777 checkpoints && 
+            cp output/my_experiment/checkpoints/*.pth checkpoints/ 2>/dev/null || true
+        #end if


Can these copies be saved? Just discover from these paths?

For these ones I think we can apply discover_datasets, yes. I'm pushing all the changes soon.

Co-authored-by: M Bernt <[email protected]>

…laxy

Add Galaxy wrapper for BiaPy tool

569c66e

bernt-matthias reviewed Jul 23, 2025

View reviewed changes

danifranco added 3 commits July 24, 2025 11:41

Modify the provided input cfg file with paths to the local data

189f237

Reformat links in help tags using markdown

10d3461

Clean the main XML by moving repeated parameters into macros.xml and …

2db71a6

…calling whenever they are used

danifranco added 2 commits July 30, 2025 10:49

Modify how train/tets data is handled according to Bernt's suggestions

10b3431

Allow multiple inputs in test

e5a1685

Add .shed.yml file

a76ae16

Control GPU with an admin variable

fd71262

bernt-matthias reviewed Aug 8, 2025

View reviewed changes

tools/biapy/create_yaml.py Outdated Show resolved Hide resolved

danifranco added 10 commits August 11, 2025 12:18

Improve create_yaml.py by adding the version of the template to download

d815b46

First working version of the tool. Still need to be improved

9ab4c39

Update the tool correcting the errors found by testing it locally

5946b91

Solve flake8 errors in create_yaml.py

3c05116

Correct test cases

eaf1153

Update to BiaPy 3.6.4

98e0fd0

Correct tool type and other minor errors

4634370

Solve discover_datasets error

5847ee2

Correct tests

be4e387

Move data copy process after yaml creation and solve None param check

bdf88d4

bernt-matthias reviewed Aug 18, 2025

View reviewed changes

danifranco added 4 commits August 18, 2025 16:01

Do not hide the errors in some commands and avoid copying all the out…

cbd945f

…puts, only those that the user selected

Change outputs to selected_outputs

50af07b

Add some redirections again in those places where I do not check if t…

bb91abf

…he files exists previously in bash. This will prevent crashing when the users check some outputs even when they are not created

Try to solve opencv problem

65d2aee

danifranco added 4 commits August 19, 2025 11:35

Make Conda provide OpenCV (cv2) and OpenEXR from the same env

345f5c7

Add imath package version

d83e92a

Add assert_command blocks in tests

aec81e5

Correct assertions in test

e7838e8

Reduce help messages

1008e27

bernt-matthias reviewed Aug 22, 2025

View reviewed changes

danifranco and others added 22 commits August 23, 2025 09:58

Update tools/biapy/.shed.yml

8946e2e

Co-authored-by: M Bernt <[email protected]>

Specify versions of the dependencies

8e717e1

Delete the exit cases as they are covered by default

817fea1

Make the output selection mandatory

444c5f8

Remove format for bmz_model_name and torchvision_model_name params

12d4ae9

Add validators and regex for BMZ and Torchvision params

0ab10d1

Add a new test

1fb9159

Make test_gt_param optional

5f8e84d

Rename phase to selected_phase to avoid confusions

4338bbb

Remove chmods

9f73009

Add common_yaml_args

a5734a1

Add tiff and csv files as possible output too

b5c3d03

Add discover_datasets for charts, train_logs and checkpoints

e4584c9

Solve errors

bd88015

Rename gt_test to test_gt

7dcb206

Solve minor errors

a4d5279

Move to BiaPy 3.6.5 supporting safetesors and force their usage in Ga…

28ee108

…laxy

Remove optional tag from selected_outputs

c3327cf

Remove jpeg and hdf5 file extensions

847ac31

Solve again opencv importation error

83df263

Solve again opencv importation error

115bf78

Correct the number of outputs in the test3

6e002a3

danifranco requested a review from bernt-matthias September 6, 2025 16:52

	ln -s '$mpath' ${checkpoint_file} 2>/dev/null \|\| true &&
	ln -fs '$mpath' ${checkpoint_file} &&

Add Galaxy wrapper for BiaPy tool #7168

Are you sure you want to change the base?

Add Galaxy wrapper for BiaPy tool #7168

Uh oh!

Conversation

danifranco commented Jul 23, 2025

Uh oh!

bernt-matthias left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

bgruening commented Jul 23, 2025

Uh oh!

danifranco commented Jul 24, 2025

Uh oh!

bgruening commented Aug 5, 2025

Uh oh!

danifranco commented Aug 5, 2025

Uh oh!

bgruening commented Aug 6, 2025

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

danifranco commented Aug 19, 2025

Uh oh!

bernt-matthias left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!