Weights for ControlUTR. Sorted by training sequence:
- Qwen2.5-0.5b-mrna-pretrained-ch145000
- Qwen2.5-0.5b-mrna-stage2-1ed-ch110000
- Qwen2.5-0.5b-mrna-stage2-2ed-ch84000
- Qwen2.5-0.5b-mrna-stage2-3ed-ch50000
- Qwen2.5-0.5b-mrna-stage2-4-2ed-ch70000
- Qwen2.5-0.5b-mrna-stage2-5-2ed-ch45000
- Qwen2.5-0.5b-mrna-stage3-dpo-1-ch1000
We recommend prioritizing the last two weights.
Code: https://github.com/sherlockma11/ControlUTR
Dataset: https://huggingface.co/datasets/SherlockMa/ControlUTR_training_data