Could you upload a d20 checkpoint, even if janky?
#8
by
RonanMcGovern
- opened
There are other uploads on HuggingFace, incl. this one I did.
However, it would be good to have all of your model sizes on HF to get a sense for variation in final results.
Many other nanochat model repos don't include the full reports in the repo.
@RonanMcGovern
You can use below process to convert it to huggingface format. Even it has documentation for vllm also.
uv run \
--with "transformers @ git+https://github.com/huggingface/transformers.git@main" \
--with "tiktoken>=0.12.0" \
https://raw.githubusercontent.com/huggingface/transformers/main/src/transformers/models/nanochat/convert_nanochat_checkpoints.py \
--input_dir ./nanochat-d34 \
--output_dir ./nanochat-d3-hf