tinyko - A Tiny LLM for korean Language

installation

# https://github.com/astral-sh/uv/issues/6437#issuecomment-2535324784
uv sync --no-install-package flash-attn
uv sync --no-build-isolation

uv run 00-tknz.py
uv run 01-preprocess.py
uv run accelerate launch 02-train.py --hf_model_id your-hf/model-id

Name		Name	Last commit message	Last commit date
Latest commit History 99 Commits
.gitignore		.gitignore
.python-version		.python-version
00-tknz.py		00-tknz.py
01-preprocess.py		01-preprocess.py
02-train.py		02-train.py
README.md		README.md
chat.py		chat.py
dataset.ipynb		dataset.ipynb
generate.py		generate.py
input.txt		input.txt
ko-dpo.yaml		ko-dpo.yaml
ko-inst-tool.yaml		ko-inst-tool.yaml
ko-inst.yaml		ko-inst.yaml
model-exp.ipynb		model-exp.ipynb
moonlight-muon-train.py		moonlight-muon-train.py
muon_optimizer.py		muon_optimizer.py
packed_dataset.py		packed_dataset.py
pose.yaml		pose.yaml
pretrain_llama_with_packing.py		pretrain_llama_with_packing.py
pyproject.toml		pyproject.toml
test-collator.py		test-collator.py
test-data.py		test-data.py
test-pack-collator.py		test-pack-collator.py
test-tknz.py		test-tknz.py
test.py		test.py
test_masking_utils.py		test_masking_utils.py
tknz-eval.py		tknz-eval.py
trl-pack.py		trl-pack.py
uv.lock		uv.lock