一键导入
text-to-lora
Generate task-specific LoRA adapters from natural language descriptions using a trained T2L model for instant transformer adaptation.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
菜单
Generate task-specific LoRA adapters from natural language descriptions using a trained T2L model for instant transformer adaptation.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Onboard and manage Paperclip AI for research-paper knowledge and agent orchestration
Generate a structured scientific post and publish it to Infinite. Runs a focused single-agent investigation (PubMed search → LLM analysis → hypothesis/method/findings/conclusion) and posts the result. Faster than scienceclaw-investigate — best for targeted, single-topic posts.
Infinite platform integration for AI agent collaboration
Read a CSV or XLSX file and return columns, shape, dtypes, and first N rows as JSON.
Execute arbitrary Python code and return stdout. NumPy, pandas, scipy, matplotlib, and other scientific libraries are available.
Generate a structured scientific PDF report from a JSON description. Accepts a JSON file specifying title, authors, abstract, sections (headings, text, tables, figures), and inline data panels (heatmap, bar, scatter, line). Produces a publication-style A4 PDF using reportlab with no LaTeX dependency. All figures are either loaded from PNG paths or generated on-the-fly from inline data.
| name | text_to_lora |
| description | Generate task-specific LoRA adapters from natural language descriptions using a trained T2L model for instant transformer adaptation. |
| source_type | github |
| auth_required | true |
| repository_url | https://github.com/SakanaAI/text-to-lora |
| reference_url | https://openreview.net/forum?id=zWskCdu3QA |
Generate task-specific LoRA adapters from natural language descriptions using a trained T2L model for instant transformer adaptation.
https://github.com/SakanaAI/text-to-lora
Use this as the implementation source: clone the repo and follow its README for install, dependencies, and how to run code or experiments. The generated client prints JSON with a suggested git clone command.
https://openreview.net/forum?id=zWskCdu3QA
This is the paper or artifact home from DOI/registry metadata — not a JSON API. If this URL is arXiv, the generated client can still fetch live Atom metadata (title, abstract, authors) without a BASE_URL. For other hosts, the client uses stub mode until you set a real BASE_URL for a REST service.
The *_client.py script prints JSON that combines a GitHub repository (clone URL + suggested git clone) with optional paper context from arXiv (live Atom metadata when reference_url is arXiv). Run the real code by cloning the repo and following its README — the skill is your agent-facing entrypoint, not a substitute for the repo’s install steps.
To call a REST API instead, set BASE_URL in scripts/text_to_lora_client.py or wrap the upstream CLI with subprocess after clone.
Extracted for operators and agents. Confirm against the upstream repository or paper before relying on it in production.
16GB GPU memory (to run demo with both models simultaneously)
uv package manager (see https://docs.astral.sh/uv/getting-started/installation/)git clone https://github.com/SakanaAI/text-to-lora.git
cd text-to-lora
uv self update
uv venv --python 3.10 --seed
uv sync
uv pip install https://github.com/Dao-AILab/flash-attention/releases/download/v2.6.3/flash_attn-2.6.3+cu123torch2.3cxx11abiFALSE-cp310-cp310-linux_x86_64.whl
uv pip install src/fishfarm
Required before running any demo:
uv run huggingface-cli login
uv run huggingface-cli download SakanaAI/text-to-lora --local-dir . --include "trained_t2l/*"
uv run python scripts/generate_lora.py trained_t2l/llama_8b_t2l "This task challenges your problem-solving abilities through mathematical reasoning. You must carefully read each scenario and systematically work through the data to compute the final outcome."
For smaller GPUs, use Gemma-2-2B:
uv run python scripts/generate_lora.py trained_t2l/gemma_2b_t2l "This task challenges your problem-solving abilities through mathematical reasoning. You must carefully read each scenario and systematically work through the data to compute the final outcome."
uv run python scripts/run_eval.py --model-dir meta-llama/Llama-3.1-8B-Instruct --lora-dirs {PATH_TO_GENERATED_LORA} --save-results --tasks gsm8k
Runs Mistral-7B-Instruct-v0.2 locally alongside T2L:
uv run python webui/app.py
WANDB_MODE=disabled uv run python scripts/eval_hypermod_checkpoint.py --checkpoint_path trained_t2l/gemma_2b_t2l/hypermod.pt --full_eval --use-icl
Start async evaluator in separate process:
uv run watcher.py
Then run training script (each ~5 days on H100):
./scripts/train_t2l_mistral.sh
./scripts/train_t2l_llama.sh
./scripts/train_t2l_gemma.sh
First train oracle LoRA baselines (takes many hours):
./scripts/train_lora_baselines.sh
Then train T2L to reconstruct:
WANDB_MODE=disabled uv run python scripts/train_hyper_recon.py configs/hyper_lora_decontam_lol_tasks.yaml --model_dir=mistralai/Mistral-7B-Instruct-v0.2/ --emb_model=Alibaba-NLP/gte-large-en-v1.5 --warmup_frac=0.1 --lr=1e-3 --epochs=10000 --n_train_ds=479 --exp_setup=hyper_lora --encoder_type=linear --pred_z_score=True --n_descs_per_ds=128 --n_embs_per_sampled_task=1 --n_tasks_per_batch=4 --factorized=False --delta_w_scaling=10000 --shared_AB_head=True
uv run huggingface-cli login before downloading models/datasetsWANDB_MODE=disabled to disable Weights & Biases loggingThe same text lives in scripts/USAGE.md for tools that prefer reading files under scripts/.
--api-key (str) [required] API key for authentication --t2l-directory (str) [required] Path to the trained T2L model directory (e.g., trained_t2l/llama_8b_t2l) --task-description (str) [required] Natural language description of the task for which to generate a LoRA adapter --model-dir (str) [optional, default=None] Base model directory for evaluation (e.g., meta-llama/Llama-3.1-8B-Instruct) --lora-dirs (str) [optional, default=None] Path to generated LoRA directory for evaluation --tasks (str) [optional, default=None] Comma-separated list of evaluation tasks (e.g., gsm8k) --save-results (bool) [optional, default=False] Save evaluation results to disk --use-icl (bool) [optional, default=False] Include 3-shot in-context examples in evaluation queries --checkpoint-path (str) [optional, default=None] Path to T2L hypermod checkpoint for evaluation
python3 scripts/text_to_lora_client.py uv run python scripts/generate_lora.py trained_t2l/llama_8b_t2l "This task challenges your problem-solving abilities through mathematical reasoning."
{"lora_path": "/path/to/generated/lora", "model": "base_model_name"}