| base_model: Qwen/Qwen3.5-4B | |
| library_name: peft | |
| tags: | |
| - tinker | |
| - peft | |
| - lora | |
| - opus-magnum | |
| # opus-4b-dsl-mixed-step150-2026-04-29 | |
| LoRA adapter (rank 32) trained with RL on a custom Opus-Magnum-style motion-planning task using the **dsl** answer representation. Snapshot at training step 150 / 300. | |
| ## Source training run | |
| - wandb: [opus-4b-dsl-mixed-2026-04-29 (mqhz79iy)](https://wandb.ai/websim/opus-task/runs/mqhz79iy) | |
| - tinker checkpoint: `tinker://addea7a1-fbe5-59b8-a467-5de3736c3404:train:0/sampler_weights/000150` | |
| - distances: 1, 2, 3, 4 | |
| - task types: move, transmute, bond | |
| - hard_task_fraction: 0.15 (cap on bond + d=4 share of train pool) | |
| - learning rate: 1e-5 | |
| - group size: 8, groups per batch: 16 | |
| - renderer: qwen3_5_disable_thinking | |
| ## Usage | |
| ```python | |
| from peft import PeftModel | |
| from transformers import AutoModelForCausalLM, AutoTokenizer | |
| base = "Qwen/Qwen3.5-4B" | |
| adapter = "maxbittker/opus-4b-dsl-mixed-step150-2026-04-29" | |
| tok = AutoTokenizer.from_pretrained(base) | |
| model = AutoModelForCausalLM.from_pretrained(base, device_map="auto") | |
| model = PeftModel.from_pretrained(model, adapter) | |
| ``` | |