Steven Atkinson

ptxas error : Entry function 'fusion_##' uses too much shared data

6

I'm running into an error when I attempt to JIT a rather involved gradient when using JAX with x64 enabled. The error I get is: ```bash 2020-05-17 13:01:34.512390: E external/org_tensorflow/tensorflow/compiler/xla/service/gpu/nvptx_compiler.cc:414]...

Include `__version__` in package

1

Currently, ```python from auraloss import __version__ ``` isn't supported. It'd be great to have `__version__` defined so that depending code can detect which version is installed and react accordingly (e.g....

When training on GPU

Follow advice to use ```python class C @property @abc.abstractmethod def foo(self): pass ```

Should be able to manage with some package resource files.

- [ ] Check what's the latest version when the GUI trainer starts. - [ ] If the local version is out of date, alert the user. - [ ]...

[DOCUMENTATION] How to set the latency compensation manually

6

Document for all 3 trainers in a page.

https://github.com/sdatkinson/neural-amp-modeler/blob/1649595220c36ca4c7f5153d8db534829728c850/nam/train/core.py#L324 This is probably too low. Given that it's got a safety factor, I can probably make it a bit less sensitive.

Just like #364, but for [AIDA-X's `input.wav`](https://colab.research.google.com/corgiredirector?site=https%3A%2F%2Fdrive.google.com%2Ffile%2Fd%2F1TNpaPPc9tdCu6OA1VETWvufc7wG2nQTJ%2Fview%3Fusp%3Dsharing) from [their training notebook](https://colab.research.google.com/github/AidaDSP/Automated-GuitarAmpModelling/blob/aidadsp_devel/AIDA_X_Model_Trainer.ipynb#scrollTo=5CQleTk7GJV8). It seems that it's identical to v1_1_1.wav, plus an additional ~2 minutes of audio appended. If that's the case,...

[FEATURE] GUI trainer: Validate all files before starting

1

Warnings are good! Currently they are checked as they are trained, which can result in a problem in the middle of the night. Not good!