qwen3.sh

# ditill
axolotl train examples/qwen3/qlora-fsdp-kd_qwen3-4b.yaml
# axolotl train examples/qwen3/qlora-fsdp-kd_qwen3-4b_distilabel.yaml

# SFT
axolotl train examples/qwen3/qlora-fsdp_qwen3-4b.yaml

# RL
# rm -rf /root/.cache/*
HIP_VISIBLE_DEVICES=2,3 axolotl vllm-serve examples/qwen3/rl_qwen3-4b.yaml
HIP_VISIBLE_DEVICES=0,1 axolotl train examples/qwen3/rl_qwen3-4b.yaml --num-processes 2