# SakanaAI/text-to-lora

Hypernetworks that adapt LLMs for specific benchmark tasks using only textual task description as the input

Repository: https://github.com/SakanaAI/text-to-lora
Canonical: https://ross.abutalabs.com/products/text-to-lora
Homepage: https://arxiv.org/abs/2506.06105
Language: Python
License: Apache-2.0
License Family: permissive
Topics: fine-tuning, hypernetworks, llm, lora, machine-learning
Last push: 2025-06-08T14:42:10+00:00

## Health v2 (maintenance only)
Score: 30/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 25, release rhythm 35, longevity 34
- inputs: {"age_days": 485, "days_push": 451, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1300, forks 89 (observed 2026-08-28T04:04:17.514092+00:00)

## What it is
Text-to-LoRA (T2L) is a hypernetwork that generates LoRA adapters for large language models in a single forward pass, using only a natural language task description as input. This repository is the reference implementation with CLI LoRA generation, evaluation scripts, and a web UI demo.

## Use cases
- generate a LoRA adapter from a text task description
- adapt an LLM to a new task without fine-tuning
- compress many task-specific LoRA adapters into one model
- zero-shot generalize adapters to unseen tasks
- evaluate generated LoRAs on benchmarks like GSM8K and Arc

## When to choose
- you want task-specific LLM adaptation without dataset curation or training
- you need to serve many specialized LoRAs cheaply
- you're researching hypernetworks or parameter-efficient fine-tuning

## When to avoid
- you need fully trained, high-accuracy adapters for production tasks
- you lack a >16GB GPU
- you need a battle-tested enterprise fine-tuning pipeline rather than research code

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, llm-training, llm-inference
- domain: large-language-models, machine-learning, artificial-intelligence
- platform: python
- tags: hypernetwork, lora, fine-tuning, peft, research-code, transformers, gpu, linux

## Member repositories
- SakanaAI/text-to-lora (main) score 30

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:17.514092+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T04:52:17.530947+00:00, confidence not recorded.
  - readme: https://github.com/SakanaAI/text-to-lora (fetched 2026-08-28T04:04:17.514092+00:00, sha 5c2a1babf006)
  - homepage: https://arxiv.org/abs/2506.06105 (fetched 2026-08-29T12:09:52.801842+00:00, sha 2983e04708db)
  - site_page: https://info.arxiv.org/about/donate.html (fetched 2026-08-29T12:09:52.811192+00:00, sha cca9c3a11c56)
  - site_page: https://info.arxiv.org/about/ourmembers.html (fetched 2026-08-29T12:09:52.814670+00:00, sha 47cbc55ff1de)
  - site_page: https://info.arxiv.org/about (fetched 2026-08-29T12:09:52.816319+00:00, sha a1f16f915a9a)
  - site_page: https://info.arxiv.org/labs/index.html (fetched 2026-08-29T12:09:52.813002+00:00, sha b14a8d05a0ec)
- Data as of 2026-08-30T08:39:29.467469+00:00.