| pretty_name: OPD Method Comparison — 27 Completed LoRA Adapters | |
| tags: | |
| - lora | |
| - peft | |
| - distillation | |
| - on-policy | |
| - qwen2.5 | |
| # OPD Method Comparison — all 27 training-complete adapters | |
| This public dataset contains all 27 training-complete LoRA adapters from the OPD | |
| method-comparison experiment. | |
| **Status:** training is complete for all 27 conditions; final evaluation is still in | |
| progress. These artifacts should not yet be interpreted as final benchmark results. | |
| Base model: 'Qwen/Qwen2.5-7B-Instruct' at revision | |
| 'a09a35458c702b33eeacc393d103063234e8bc28'. | |
| Each 'adapters/<condition>/' directory contains the PEFT adapter, tokenizer files, the | |
| resolved training configuration, and a signed training-completion record. Every adapter | |
| completed 861 optimizer steps. Optimizer state is excluded. | |
| To load an adapter, download its directory with | |
| 'huggingface_hub.snapshot_download(..., repo_type="dataset")', then pass the local | |
| adapter directory to 'peft.PeftModel.from_pretrained'. | |