| --- |
| library_name: peft |
| license: apache-2.0 |
| tags: |
| - evalexplorer |
| - document-classification |
| - structured-output |
| --- |
| |
| # evalexplorer-classify-adapters |
|
|
| Experimental LoRA adapters for the EvalExplorer document classifier, one subfolder per run. Every adapter here is |
| on an Apache-2.0 base (Qwen3.5, Gemma 4); each keeps its base model's terms. Adapters worth using |
| on their own get their own repo (`baobabtech/evalexplorer-classify-<model>-sft`); variants from GRPO and other |
| trials live here so they do not each need one. Each subfolder holds `adapter_config.json`, |
| `adapter_model.safetensors`, the tokenizer and `training_log.json`. |
|
|
| | Adapter | Base model | Mean field score | Exact match | Run | |
| |---|---|---:|---:|---| |
| | [`qwen3.5-2b-grpo-countries`](https://huggingface.co/baobabtech/evalexplorer-classify-adapters/tree/main/qwen3.5-2b-grpo-countries) | unsloth/Qwen3.5-2B | 0.847 | 0.246 | [report](https://huggingface.co/datasets/baobabtech/evalexplorer-classify-experiments/blob/main/runs/qwen3.5-2b-grpo-countries--test/README.md) | |
| | [`qwen3.5-2b-grpo-lr5e6`](https://huggingface.co/baobabtech/evalexplorer-classify-adapters/tree/main/qwen3.5-2b-grpo-lr5e6) | unsloth/Qwen3.5-2B | 0.843 | 0.254 | [report](https://huggingface.co/datasets/baobabtech/evalexplorer-classify-experiments/blob/main/runs/qwen3.5-2b-grpo-lr5e6--test/README.md) | |
| | [`gemma-4-e2b-grpo-lr5e6`](https://huggingface.co/baobabtech/evalexplorer-classify-adapters/tree/main/gemma-4-e2b-grpo-lr5e6) | unsloth/gemma-4-E2B-it | 0.827 | 0.276 | [report](https://huggingface.co/datasets/baobabtech/evalexplorer-classify-experiments/blob/main/runs/gemma-4-e2b-grpo-lr5e6--test/README.md) | |
| | [`qwen3.5-2b-grpo`](https://huggingface.co/baobabtech/evalexplorer-classify-adapters/tree/main/qwen3.5-2b-grpo) | unsloth/Qwen3.5-2B | 0.822 | 0.179 | [report](https://huggingface.co/datasets/baobabtech/evalexplorer-classify-experiments/blob/main/runs/qwen3.5-2b-grpo--test/README.md) | |
| | [`gemma-4-e2b-grpo`](https://huggingface.co/baobabtech/evalexplorer-classify-adapters/tree/main/gemma-4-e2b-grpo) | unsloth/gemma-4-E2B-it | 0.819 | 0.157 | [report](https://huggingface.co/datasets/baobabtech/evalexplorer-classify-experiments/blob/main/runs/gemma-4-e2b-grpo--test/README.md) | |
| | [`gliner2.5-base-passage`](https://huggingface.co/baobabtech/evalexplorer-classify-adapters/tree/main/gliner2.5-base-passage) | fastino/gliner2.5-base-v1 | 0.584 | 0.007 | [report](https://huggingface.co/datasets/baobabtech/evalexplorer-classify-experiments/blob/main/runs/gliner2.5-base-passage--test/README.md) | |
| | [`gliner2.5-small-passage`](https://huggingface.co/baobabtech/evalexplorer-classify-adapters/tree/main/gliner2.5-small-passage) | fastino/gliner2.5-small-v1 | 0.532 | 0.000 | [report](https://huggingface.co/datasets/baobabtech/evalexplorer-classify-experiments/blob/main/runs/gliner2.5-small-passage--test/README.md) | |
|
|
| Scores are on the 134-document test split of |
| [`baobabtech/evalexplorer-data`](https://huggingface.co/datasets/baobabtech/evalexplorer-data), config `classify_codes`. Context and every other |
| run: [`baobabtech/evalexplorer-classify-experiments`](https://huggingface.co/datasets/baobabtech/evalexplorer-classify-experiments). |
|
|
| ```python |
| from huggingface_hub import snapshot_download |
| from peft import PeftModel |
| |
| path = snapshot_download("baobabtech/evalexplorer-classify-adapters", allow_patterns=["gemma-4-e2b-grpo-lr5e6/*"]) |
| model = PeftModel.from_pretrained(base_model, f"{path}/gemma-4-e2b-grpo-lr5e6") |
| ``` |
|
|
| New runs can write here with `--output-repo baobabtech/evalexplorer-classify-adapters/<name>`. |
|
|
