andre930/persona-basis-run21-fkl
text-generation model by andre930
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.
Four LoRA adapters ("heads") for Qwen/Qwen3-4B-Instruct-2507. Each one gives the model one answer style, on two axes: voice (warm or impersonal) and depth (concise or thorough). This is the step-60 checkpoint of training run "run21", out of 100 steps; it was selected on validation data. The heads need no system prompt. The style is in the weights, so send the user's message as the only turn. To serve with vLLM, download a head folder and pass it as a LoRA adapter, with enablelora=True and maxlorarank of at least 32. A pair such as warm + concise can be served as one adapter whose weight update is a weighted sum of the heads': W = W₀ + Σₖ wₖ·ΔWₖ, where ΔWₖ = (α/r)·BₖAₖ. It can be built exactly by concatenation: stack the four heads' A matrices, scale each head's B by wₖ and stack those. The result is one rank-128, α-256…
Read the full model card ↗ · Preview checked 2026-09-26T07:11:44.709Z
Inside the original model card — Document outline
- persona-basis run21: persona LoRA heads (forward KL, step 60)
- Usage
- Combining two personas
- Training
- Limitations
Headings are captured from the source. Links open the publisher’s document, not a locally hosted copy.
Documentation belongs to its respective authors. A listing is not a grant of reuse or training rights. Confirm the document’s own terms at the source.
Model overview
andre930/persona-basis-run21-fkl is a text-generation model repository published by andre930 on Hugging Face. The source reports the peft library.
This page summarizes Hub metadata. For intended use, training data, evaluation results and limitations, consult the original model card.
Model facts
- Task
- text-generation
- Library
- peft
- Recent downloads (30 days)
- 0
- Cumulative likes
- 0
- Hugging Face trending score
- 0
- Reported safetensors parameters
- Not reported
- Architecture
- Not reported
- License
- Not reported — check the model card
- Access
- Not gated by Hugging Face
- Created
- 2026-09-26T06:36:10.000Z
- Last modified
- 2026-09-26T06:37:27.000Z
Compatibility and lineage
Language coverage was not reported.
Reported base models: Qwen/Qwen3-4B-Instruct-2507
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 2026-09-26T06:37:45.751Z. Daily imports are snapshots, not real-time monitoring.
Popularity and source listings do not establish security, suitability, licensing rights or benchmark performance.
Related models
trl-internal-testing/tiny-Qwen2ForCausalLM-2.5