stevensu/YuE2-3B-instrumental-kto-lora
AI model by stevensu
From the publisher
Model card & documentation
Source previewRead the publisher’s intended use, setup instructions, evaluations and limitations. The original model card is the source of truth.
A 70 MB LoRA adapter for YuE2-3B that makes instrumental generation actually instrumental. With the official model, a "no vocals" prompt still produces singing, humming or chanting in roughly half of the songs in our tests; with this adapter the semantic stage is steered away from vocal content while the score, the NAR and the VAE stay the official ones. 中文摘要：给 YuE2-3B 用的 70 MB LoRA，目标是让"纯音乐"真的没有人声。原模型即使在 prompt 里写了 no vocals，我们测试中仍有约一半的曲子出现演唱、哼唱或吟唱；打开这个 adapter 后，在评测集配对盲听里人声率从 56% 降到 0%（95% 区间 0–16%）。它只在语义 token 生成阶段生效，乐谱规划、NAR、VAE 仍用官方权重；关闭时输出与官方模型逐位一致。只限非商业使用（CC BY-NC 4.0）。 Same prompt, same planned score, same seed — the only difference is whether the adapter is on: More, generated with the adapter on (prompts in samples/samples.json): Samples are AAC 192 kbps; the model outputs 48 kHz stereo. Blind listening (the main result). The evaluation set has 24 game/film-BGM prompts that were never used in training, rendered twice with identical scores and seeds: adapter off and on (40 pairs).…
Read the full model card ↗ · Preview checked 2026-11-07T09:46:30.071Z
Inside the original model card — Document outline
- YuE2-3B Instrumental KTO LoRA (experimental)
- Listen
- Results
- Usage
- Training
- Limitations
- License and attribution
Headings are captured from the source. Links open the publisher’s document, not a locally hosted copy.
Links from the model card
References supplied by the publisher, not independently verified endorsements. Check the destination before downloading files or entering credentials.
Documentation belongs to its respective authors. Reported project/model license: cc-by-nc-4.0. A listing is not a grant of reuse or training rights. Confirm the document’s own terms at the source.
Model overview
stevensu/YuE2-3B-instrumental-kto-lora is a task-unspecified model repository published by stevensu on Hugging Face. A library was not reported.
This page summarizes Hub metadata. For intended use, training data, evaluation results and limitations, consult the original model card.
Model facts
- Task
- Not reported
- Library
- Not reported
- Recent downloads (30 days)
- 0
- Cumulative likes
- 0
- Hugging Face trending score
- 0
- Reported safetensors parameters
- Not reported
- Architecture
- Not reported
- License
- cc-by-nc-4.0
- Access
- Not gated by Hugging Face
- Created
- 2026-11-07T06:42:12.000Z
- Last modified
- 2026-11-07T06:43:00.000Z
Compatibility and lineage
Language coverage was not reported.
Reported base models: m-a-p/YuE2-3B
Parameter count is not a RAM/VRAM requirement. Check precision, quantization, context length and runtime compatibility in the model card; no hardware or API-cost claim is inferred here.
Use and evaluate
Check license terms and access requirements first. Review the model card’s documented loading instructions, evaluations and safety limitations. Benchmark scores are not imported by this directory, and popularity is not an accuracy ranking.
No model weights are downloaded or executed by AltAPIs. Never enable remote model code without reviewing it.
Source and freshness
Source: Hugging Face Hub. Metadata observed 2026-11-07T06:54:27.277Z. Daily imports are snapshots, not real-time monitoring.
Popularity and source listings do not establish security, suitability, licensing rights or benchmark performance.