pmahdavi/Olmo-3-7B-Think-Math-Code

1---2base_model:3- allenai/Olmo-3-1025-7B4- allenai/Olmo-3-7B-RL-Zero-Math5- allenai/Olmo-3-7B-RL-Zero-Code6- allenai/Olmo-3-7B-Think-SFT7library_name: transformers8tags:9- mergekit10- merge11 12---13# merged-model14 15This is a merge of pre-trained language models created using [mergekit](https://github.com/cg123/mergekit).16 17## Merge Details18### Merge Method19 20This model was merged using the [Task Arithmetic](https://arxiv.org/abs/2212.04089) merge method using [allenai/Olmo-3-1025-7B](https://huggingface.co/allenai/Olmo-3-1025-7B) as a base.21 22### Models Merged23 24The following models were included in the merge:25* [allenai/Olmo-3-7B-RL-Zero-Math](https://huggingface.co/allenai/Olmo-3-7B-RL-Zero-Math)26* [allenai/Olmo-3-7B-RL-Zero-Code](https://huggingface.co/allenai/Olmo-3-7B-RL-Zero-Code)27* [allenai/Olmo-3-7B-Think-SFT](https://huggingface.co/allenai/Olmo-3-7B-Think-SFT)28 29### Configuration30 31The following YAML configuration was used to produce this model:32 33```yaml34# Task arithmetic merge: Apply Math+Code task vectors to Think-SFT35#36# Mathematical formulation:37#   output = Think-SFT + 0.5*(RL-Zero-Math - base) + 0.5*(RL-Zero-Code - base)38#39# This is achieved by treating Think-SFT as a model with weight=1.0:40#   output = base + 1.0*(Think-SFT - base) + 0.5*(Math - base) + 0.5*(Code - base)41#42# Usage:43#   modal run modal_merge.py --config examples/olmo-think-math-code.yaml --hf-repo pmahdavi/Olmo-3-7B-Think-Math-Code44 45merge_method: task_arithmetic46base_model: allenai/Olmo-3-1025-7B47models:48  - model: allenai/Olmo-3-7B-Think-SFT49    parameters:50      weight: 1.051  - model: allenai/Olmo-3-7B-RL-Zero-Math52    parameters:53      weight: 0.554  - model: allenai/Olmo-3-7B-RL-Zero-Code55    parameters:56      weight: 0.557dtype: bfloat1658 59```60