--- model-index: - name: OLMo-2-0425-1B results: - task: type: multiple-choice name: AGIEval dataset: type: agi name: AGIEval metrics: - type: accuracy value: 0.34 - task: type: multiple-choice name: ARC_C dataset: type: arc name: ARC-Challenge metrics: - type: accuracy value: 0.47 - task: type: multiple-choice name: ARC_E dataset: type: arc name: ARC-Easy metrics: - type: accuracy value: 0.74 - task: type: multiple-choice name: BBH dataset: type: bbh name: BBH metrics: - type: accuracy value: 0.3 - task: type: question-answering name: BoolQ dataset: type: boolq name: BoolQ metrics: - type: accuracy value: 0.69 - task: type: multiple-choice name: CSQA dataset: type: commonsense_qa name: CSQA metrics: - type: accuracy value: 0.6 - task: type: question-answering name: CoQA dataset: type: coqa name: CoQA metrics: - type: accuracy value: 0.69 - task: type: question-answering name: DROP dataset: type: drop name: DROP metrics: - type: accuracy value: 0.35 - task: type: math-word-problem name: GSM8K dataset: type: gsm8k name: GSM8K metrics: - type: accuracy value: 0.36 - task: type: multiple-choice name: HSwag dataset: type: hellaswag name: HellaSwag metrics: - type: accuracy value: 0.6 - task: type: question-answering name: JPRDY dataset: type: jeopardy name: Jeopardy metrics: - type: accuracy value: 0.63 - task: type: multiple-choice name: MMLU dataset: type: mmlu name: MMLU metrics: - type: accuracy value: 0.43 - task: type: multiple-choice name: MMLU-Pro dataset: type: mmlu name: MMLU-Pro metrics: - type: accuracy value: 0.19 - task: type: question-answering name: NatQs dataset: type: naturalqs name: Natural Questions metrics: - type: accuracy value: 0.19 - task: type: multiple-choice name: OBQA dataset: type: openbookqa name: OpenBookQA metrics: - type: accuracy value: 0.51 - task: type: multiple-choice name: PIQA dataset: type: piqa name: PIQA metrics: - type: accuracy value: 0.71 - task: type: multiple-choice name: SIQA dataset: type: socialiqa name: SocialIQA metrics: - type: accuracy value: 0.56 - task: type: question-answering name: SQuAD dataset: type: squad name: SQuAD metrics: - type: accuracy value: 0.8 - task: type: question-answering name: TriviaQA dataset: type: triviaqa name: TriviaQA metrics: - type: accuracy value: 0.55 - task: type: coreference-resolution name: WinoG dataset: type: winogrande name: Winograd metrics: - type: accuracy value: 0.61 base_model: allenai/OLMo-2-0425-1B library_name: peft model_name: OLMo-2-0425-1B_tulu-3-sft-olmo-2-mixture-0225_lr0.0001_seed42 datasets: - allenai/tulu-3-sft-olmo-2-mixture-0225 tags: - base_model:adapter:allenai/OLMo-2-0425-1B - lora - sft - transformers - trl license: cc-by-4.0 language: - en pipeline_tag: text-generation --- # Model Card for OLMo-2-0425-1B_tulu-3-sft-olmo-2-mixture-0225_lr0.0001_seed42 This model is a instruction fine-tuned version of [allenai/OLMo-2-0425-1B](https://huggingface.co/allenai/OLMo-2-0425-1B) trained using a [LoRA](https://github.com/microsoft/LoRA) adapter on [Tülu3](https://huggingface.co/datasets/allenai/tulu-3-sft-olmo-2-mixture-0225) for one epoch via [TRL](https://github.com/huggingface/trl). ## Uses This model was created for training data influence estimation experiments using [DataInf](https://github.com/ykwon0407/DataInf) and [LESS](https://github.com/princeton-nlp/LESS). See our [paper](https://arxiv.org/abs/2601.03786) and [repo](https://github.com/loris3/ceelm) for details. ## Quick start ```python from huggingface_hub import hf_hub_download import json from transformers import AutoTokenizer,AutoModelForCausalLM from peft import PeftModel from transformers import pipeline repo_id = "loris3/OLMo-2-0425-1B_tulu-3-sft-olmo-2-mixture-0225_lr0.0001_seed42" adapter_path = hf_hub_download(repo_id=repo_id, filename="adapter_config.json") adapter_config = json.load(open(adapter_path)) base_model_name_or_path = adapter_config["base_model_name_or_path"] chat_template = open(hf_hub_download(repo_id=repo_id, filename="chat_template.jinja")).read() tokenizer = AutoTokenizer.from_pretrained(base_model_name_or_path) tokenizer.chat_template = chat_template if tokenizer.pad_token is None: tokenizer.pad_token = tokenizer.eos_token model = AutoModelForCausalLM.from_pretrained(base_model_name_or_path) model.resize_token_embeddings(len(tokenizer)) model = PeftModel.from_pretrained(model, repo_id, is_trainable=False) question = "Could you give us some of your political beliefs?" generator = pipeline("text-generation", model=model, tokenizer=tokenizer, device=0) output = generator([{"role": "user", "content": question}], max_new_tokens=128, do_sample=False, temperature=1.0, top_p=1.0, return_full_text=False)[0] print(output["generated_text"]) ``` ### Training Hyperparameters | Parameter | Value | |-------------------------------|-----------------------------------------------------------------------| | Precision | bfloat16 | | Optimizer | AdamW (torch fused) | | Learning rate | 1×10⁻⁴ | | LR scheduler | Linear | | Weight decay | 0.0 | | Max grad norm | 1.0 | | LoRA rank (r) | 16 | | LoRA alpha | 32 | | LoRA dropout | 0.1 | | LoRA bias | none | | Target modules |q_proj, c_attn, v_proj| | Trainable params | LoRA only | | Train batch size / device | 4 | | Gradient accumulation | 8 | | Effective batch size | 32 | | Training epochs | 1 | | Max sequence length | 1024 | | Gradient checkpointing | False | | Seed | 42 | ### Framework versions - PEFT 0.17.1 - TRL: 0.23.0 - Transformers: 4.56.2 - Pytorch: 2.8.0+cu126 - Datasets: 4.0.0 - Tokenizers: 0.22.1 ## Evaluation We evaluate with [OLMES](https://github.com/allenai/olmes) Task suites: `core_9mcqa::olmes`, `mmlu:mc::olmes`, `olmo_2_generative::olmes`, `olmo_2_heldout::olmes` | Task | Score| |------|---------| | AGIEval | 0.34 | | ARC_C | 0.47 | | ARC_E | 0.74 | | BBH | 0.30 | | BoolQ | 0.69 | | CSQA | 0.60 | | CoQA | 0.69 | | DROP | 0.35 | | GSM8K | 0.36 | | HSwag | 0.60 | | JPRDY | 0.63 | | MMLU | 0.43 | | MMLU-Pro | 0.19 | | NatQs | 0.19 | | OBQA | 0.51 | | PIQA | 0.71 | | SIQA | 0.56 | | SQuAD | 0.80 | | TriviaQA | 0.55 | | WinoG | 0.61 |