anonymousgiraffe/Decomposer-Qwen3-8B
Decomposer-Qwen3-8B by anonymousgiraffe is an 8.2 billion parameter causal language model, post-trained from Qwen3-8B, specifically designed to decompile symbolic music (MIDI) into executable Strudel programs. This model excels at translating MIDI note events, including tempo and meter information, into structured Strudel code. Its primary application is in music generation and analysis, enabling the conversion of musical performances into programmatic representations.
Loading preview...
Decomposer-Qwen3-8B: MIDI to Strudel Program Decompiler
Decomposer-Qwen3-8B is an 8.2 billion parameter causal language model, post-trained from Qwen3-8B, engineered for the unique task of decompiling symbolic music (MIDI) into executable Strudel programs. This model translates MIDI data, including instrument-wise note events and cycle coordinates, into a structured programming language for music synthesis.
Key Capabilities
- Symbolic Music Decompilation: Converts MIDI files into Strudel programs, representing musical structures programmatically.
- MIDI Input Processing: Handles MIDI serialized as instrument-wise note events, including pitch, onset (in cycle coordinates), BPM, and meter.
- Specialized Training: Post-trained on the synthetic Strudel-Synth dataset, comprising 21,174 (MIDI, Strudel) pairs, to optimize for this specific decompilation task.
- Base Model: Built upon the robust Qwen3-8B architecture, leveraging its language understanding capabilities for code generation.
Use Cases
- Music Analysis: Programmatic analysis of musical compositions by converting them into a structured code format.
- Algorithmic Composition: Facilitating the creation and manipulation of music through code.
- Educational Tools: Assisting in understanding musical structure and programming concepts through a musical context.
Recommended inference settings include a temperature of 1.0 and max_new_tokens of 4096, with enable_thinking set to false.