NAM Trainer Reloaded v1.0.1

NAM Trainer Reloaded v1.0.1

NAM Trainer Reloaded v1.0.1 is a focused maintenance release for A2 two-stage training and scheduler visibility.

Changes

• A2 Full+Lite and other packed A2 two-stage runs now hand the best full and best lite sub-WaveNet checkpoints from the core stage to the refinement stage separately.
• The final A2 export still combines the refined packed submodels into a normal SlimmableContainer capture, preserving the regular full/lite behavior and slim option.
• Scheduler rows now keep completed jobs visible and mark them as [done], with the final ESR/result shown when available.
• The trainer version now reports 1.0.1 in the package metadata and NAM Trainer - V1.0.1 in the GUI title.
• INSTALL.md has been expanded into a clearer installation manual with dependencies, Windows/macOS setup steps, verification, and repair notes.

Validation

• Confirmed the package version reports 1.0.1.
• Confirmed the V1.0.1 GUI title string is present.
• Validated a real A2 Full+Lite core/refinement run: stage 2 loaded the separate stage 1 packed-best checkpoints, and the final .nam exported as a two-submodel SlimmableContainer with exact stage-2 packed-best weights.

Installation Summary

Windows:

1. Download and extract the release zip.
2. Run INSTALL_WINDOWS.bat.
3. Run run_trainer.bat.

Windows CPU-only:

INSTALL_WINDOWS.bat -CPU

macOS:

1. Download and extract the release zip.
2. Run INSTALL_MACOS.command.
3. Run run_trainer.command.

If macOS blocks the .command files after download, run:

chmod +x INSTALL_MACOS.command run_trainer.command INSTALL_MACOS.sh run_trainer.sh scripts/install_unix.sh

Then double-click the .command files again.

See INSTALL.md for full CPU/GPU dependency details.

NAM Trainer Reloaded v1.0.0

NAM Trainer Reloaded v1.0.0

NAM Trainer Reloaded is a compact desktop trainer package based on Neural Amp Modeler training code, with A2 architectures, paired-checkpoint resume support, additional WaveNet architectures, scheduler tools, and current TTS input-clip recognition updates.

Highlights

• Local desktop GUI for NAM training.
• A2 architecture support, including current packed A2 variants.
• Resume-from-checkpoint support, including A2 checkpoint pairs.
• Custom architecture list including RevYLo, Nano variants, Double/xDouble/yDouble, and A2 hybrids.
• Analyze Lightning folders for reviewing previous training runs, saved metrics, ESR values, and timing data.
• Core and refinement dual-stage training with separate epoch, learning-rate, decay, scheduler, and refinement-focus controls.
• Windows local install/run scripts.
• macOS install/run shell scripts and .command launchers.

Installation Summary

Windows:

1. Download and extract the release zip.
2. Run INSTALL_WINDOWS.bat.
3. Run run_trainer.bat.

Windows CPU-only:

INSTALL_WINDOWS.bat -CPU

macOS:

1. Download and extract the release zip.
2. Run INSTALL_MACOS.command.
3. Run run_trainer.command.

If macOS blocks the .command files after download, run:

chmod +x INSTALL_MACOS.command run_trainer.command INSTALL_MACOS.sh run_trainer.sh scripts/install_unix.sh

Then double-click the .command files again.

See INSTALL.md for full CPU/GPU dependency details.