##qwenstr
this repo contains tools to train qwen3-0.6b using nostr data. after training you'll have a model that generates responses in nostr json format. generated events need to be finalized before sending. model can learn multiple event kinds and tag usage formats.

#howto
fetch some data:
nosdump relays | head -n 100000 > data

generate dataset:
cat data | node filterdata.js | tee traindata2.jsonl; wc -l traindata2.jsonl

download model:
git lfs install
git clone https://huggingface.co/Qwen/Qwen3-0.6B

train model:
./train-qwen.sh traindata2.jsonl

run model:
ollama run qwen-vl-finetune

continue training later:
./train-qwen.sh traindata2.jsonl --continue