anira-project/guitar-lstm

GuitarLSTM for anira

GuitarML's GuitarLSTM guitar amplifier model (convolutions and an LSTM), for LibTorch, ONNX Runtime, LiteRT (256-sample input) and ExecuTorch. anira's hybrid network example.

Files

File	Engine	Size	SHA256
onnx/GuitarLSTM-libtorch-dynamic.onnx	ONNX Runtime	50,450	4789254a3bf75e0e012fbf248db431ecb71f57c7ccea10850ee85cf67bb6937f
litert/GuitarLSTM.tflite	LiteRT, TensorFlow Lite	205,632	8950b4d51b64263cdf72a4ed1b637acf569dd57d9695c54c72692d84fca9d945
libtorch/GuitarLSTM-dynamic.pt	LibTorch	60,766	7fc4deca50fa192c8b99e8688788e96cc0336656337fa46529925a18bc704f7a
executorch/GuitarLSTM-executorch.pte	ExecuTorch 1.3.1, CPU	62,724	f8172e5a88e93bce2149f296cbd07d1e21df7b395f0f88a022cfffb0bf200190
executorch/GuitarLSTM_coreml.pte	ExecuTorch 1.3.1, Core ML (Apple)	290,570	1c37e6b986c3cfafe717fbc0f4cff72bbf79ab582e91728421f97c8d9cf098fa
executorch/GuitarLSTM_vulkan.pte	ExecuTorch 1.3.1, Vulkan (GPUs)	59,268	96971ef5a9f02d3d10c0465914c2cb2fa6776aacba98d35bcbb178c28b394017
coreml/GuitarLSTM.mlpackage	Core ML (native)	89,490	per file in the catalog

Source

faressc/GuitarLSTM @ 08c7418 (a fork of GuitarML/GuitarLSTM), pytorch-version/models/model_0/ and tensorflow-version/models/model_0/; the export code is in that repository.

Formats and checks

Every model here runs on ONNX Runtime, LiteRT and TensorFlow Lite, LibTorch, ExecuTorch (XNNPACK or portable kernels on the CPU, Core ML on Apple, Vulkan on GPUs) and as a native Core ML ML program, with the ONNX models' tensor names. anira foundation models streams every format of every description over twelve consecutive calls with the state fed back and checks that they agree (tools/check_streaming.py). Deviations below are from ONNX Runtime, per model, in the order of the files.

Format	Files	Max deviation
LiteRT, TensorFlow Lite (re-exported)	litert/*.tflite	8.4e-08
ExecuTorch, Core ML (Apple, float32)	executorch/*_coreml.pte	7.1e-08
ExecuTorch, Vulkan (GPUs)	executorch/*_vulkan.pte	exported; not run here (no Vulkan runtime)
Core ML ML program (anira's native engine, coming with v3; macOS 13, iOS 16)	coreml/*.mlpackage	7.8e-08

uv run --no-project export/export_litert.py <model.onnx> <out.tflite> [<description>] [--guitar-lstm <pt>]
uv run --no-project export/export_executorch.py <model.onnx> <out.pte> [<description>] [--guitar-lstm <pt>] [--coreml | --vulkan]
uv run --no-project export/export_coreml.py <model.onnx> <out.mlpackage> [<description>] [--guitar-lstm <pt>]
uv run --no-project export/export_libtorch.py <model.onnx> <out.pt> [<description>] [--guitar-lstm <pt>]

<description> is the model's description in anira foundation models, for the input shapes where the ONNX model's time axis is dynamic. The PyTorch module comes from the ONNX model (export/onnx_to_torch.py: onnxsim to static shapes, onnx2torch with traceable converters), or for GuitarLSTM its rebuild from the TorchScript model (--guitar-lstm, export/sources.py), since onnx2torch does not convert an LSTM; each export is checked against ONNX Runtime over eight consecutive calls of random input with the state fed back, relative to the output's peak (export/check.py).

upstream/ keeps anira's original LiteRT file (upstream/litert/GuitarLSTM-256.tflite), which the descriptions no longer name: it computes something else than the other formats, 12 to 16 % off from the first call. litert/GuitarLSTM.tflite is exported from the same PyTorch module as the other formats.

This repository holds the weights, the export scripts and the licence. The anira model descriptions live with the code in anira foundation models (models/guitar-lstm), versioned with git. Pin a revision and check each file's SHA256 when you download:

https://huggingface.co/anira-project/guitar-lstm/resolve/<revision>/onnx/GuitarLSTM-libtorch-dynamic.onnx

Licence

GPL-3.0: LICENSE, copied from the source repository. Software that ships these weights may have to meet the GPL's terms; check before you do.