Fynn Beuttenmüller

turns out this is not yet implemented, but we planned this here: https://github.com/bioimage-io/spec-bioimage-io/issues/252

> Also, for the models with run_mode deepimagej, the test validator says that is not being recognized, do I comment that? Or how can I fix the issue? there is...

> https://drive.google.com/drive/folders/14qLzIV2qMx-Yy5xD4dp2dfAwO5YSV6BB?usp=sharing this model is missing the tensorflow weights and for the keras weights I get: ``` 'error': "Shape (720, 576, 1, 1, 0) of test input 0 'input' does...

How did you install spec and core? Those versions are not compatible

hmm.. really good to know that the pin we set [here](https://github.com/bioimage-io/core-bioimage-io-python/blob/7f26c9b1dac17f48fb5006e4c239ca2ecac3ef10/setup.py#L29) does not make it into the conda-recipe...

currently yes, the new core release is coming very soon though, so maybe wait just a few days if you can

looks like you're trying to run a model traced on GPU on CPU (or vice versa??) (`libtorch_cpu.so`) makes sure inputs and model are on the same device and try GPU...

input/output layer names are important (https://github.com/bioimage-io/collection-bioimage-io/issues/571#issuecomment-1486480847) , but not always very informative or dependent on the architecture "conv2d_19". I wonder if we shouldn't add them as fields to the corresponding...

To illustrate the separation of the fields: ```yaml inputs: [{name: raw_input, ...}] outputs: [{name: prediction, ...}] weights: tensorflow_saved_model_bundle: input_layer_names: [input_1] output_layer_names: [conv2d_19] ``` **edit:** changed input_layer_name to input_layer_name**s** list developers...

I always thought of the input and output **tensor** names as content descriptions, `raw_input`, `em_volume`, or even something like `left_camera` and `right_camera` for a two camera setup... i.e. identifiers you...