nammy

A proof-of-concept trainer for Neural Amp Modeler's classic A1 WaveNet architecture, implemented with tinygrad instead of PyTorch.

What's implemented

• The standard A1 WaveNet: two layer arrays (16 and 8 channels), dilations 1–512, kernel size 3, Tanh, residual 1x1s, per-array head rechannel, head_scale = 0.02, receptive field 4093 — a faithful port of nam.models.wavenet.
• NAM-style data pipeline: WAV loading (PCM 16/24/32 and IEEE float), latency compensation, and (nx+ny-1, ny) window slicing matching nam.data.Dataset.
• Training matching NAM's standard learning config: Adam(lr=0.004), per-epoch exponential LR decay (gamma 0.993), MSE loss, ESR validation with best-checkpoint restore, JIT-compiled train step.
• Export to .nam (classic v0.5.4 schema) loadable by the NAM plugin.

Not implemented (yet): A2/packed training, gated/FiLM variants, MRSTFT loss, pre-emphasis, output loudness normalization, the standardized input-file splits/checks.

Usage

Train from an input/output pair:

$ uv run python -m nammy train input.wav output.wav --epochs 100 --out model.nam

input.wav is the DI/reamp source, output.wav the processed capture; both must share a sample rate and be time-aligned (use --latency <samples> to compensate reamp latency).

Run audio through a trained model (reamp):

$ uv run python -m nammy process model.nam input.wav output.wav

This also loads classic-schema (non-gated Tanh WaveNet) .nam files trained elsewhere.

Notes

• Runs on tinygrad's default device. Without a GPU and without clang installed, it falls back to the libLLVM CPU renderer automatically.
• Validation: uv run tests/test_poc.py checks receptive field, forward parity against an independent numpy implementation, .nam export round-trip, dataset alignment, and a training smoke test.