Doc-to-LoRA (D2L): Learning to Instantly Internalize Contexts

:sparkles:Interactive Web | :newspaper:X | :scroll:Paper | :hugs:Hugging Face | :octocat:GitHub
A reference implementation of Doc-to-LoRA (D2L).

🛠️ Installation

curl -LsSf https://astral.sh/uv/install.sh | sh
./install.sh

🤗 Pre-Trained Models

uv run huggingface-cli login
uv run huggingface-cli download SakanaAI/doc-to-lora --local-dir trained_d2l --include "*/"

🚀 Python API Usage

# caveat: this interface only supports non-batched inputs
# for batched inference please see `src/ctx_to_lora/modeling/hypernet.py`
import torch

from ctx_to_lora.model_loading import get_tokenizer
from ctx_to_lora.modeling.hypernet import ModulatedPretrainedModel

# model loading
checkpoint_path = "trained_d2l/gemma_demo/checkpoint-80000/pytorch_model.bin"
state_dict = torch.load(checkpoint_path, weights_only=False)
model = ModulatedPretrainedModel.from_state_dict(
    state_dict, train=False, use_sequence_packing=False
)
model.reset()
tokenizer = get_tokenizer(model.base_model.name_or_path)

# prepare data
doc = open("data/sakana_wiki.txt", "r").read()
chat = [{"role": "user", "content": "Tell me about Sakana AI."}]
chat_ids = tokenizer.apply_chat_template(
    chat,
    add_special_tokens=False,
    return_attention_mask=False,
    add_generation_prompt=True,
    return_tensors="pt",
).to(model.device)


# calls after internalization will be influenced by internalized info
model.internalize(doc)

outputs = model.generate(input_ids=chat_ids, max_new_tokens=512)
print(tokenizer.decode(outputs[0]))


# remove internalized info
# model.reset()

# without internalized info, the model will halucinate
# outputs = model.generate(input_ids=chat_ids, max_new_tokens=512)
# print(tokenizer.decode(outputs[0]))

🎮 Interactive Demo

uv run demo/app.py

Video Demo

🧪 Experimental Scripts

To run any of the following scripts, use uv run $PATH_TO_SCRIPT from the root of this project.

Experiment	Data prep	Training	Evaluation	Notes
Main experiment	`scripts/main_exp/0-download_data.sh`	`scripts/main_exp/1-train.sh`	`scripts/main_exp/eval/*.sh`	Downloading data is fastest; regenerate only if you need fresh synthetic data. Evaluation scripts reproduce the main paper metrics.
NIAH	`scripts/niah/0-gen_data.sh`	`scripts/niah/1-train.sh`	`scripts/niah/2-eval.sh`	Run the scripts in order; data generation only needs to happen once

🔬 Self-Generated Data Viewer

After downloading/generating the data, we can see samples of the data using this script.

uv run webui/self_gen_viewer.py

See more info at webui/SELF_GEN_VIEWER.md.

📚 Citation

@techreport{sakana2025doc-to-lora,
  title       = {{Doc-to-LoRA: Learning to Instantly Internalize Contexts}},
  author      = {Rujikorn Charakorn and Edoardo Cetin and Shinnosuke Uesaka and Robert Tjarko Lange},
  institution = {Sakana AI},
  year        = {2026},
  month       = {Febuary},
  note        = {Technical Report}
}