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

1# Task arithmetic merge: Apply Math+Code task vectors to Think-SFT2#3# Mathematical formulation:4#   output = Think-SFT + 0.5*(RL-Zero-Math - base) + 0.5*(RL-Zero-Code - base)5#6# This is achieved by treating Think-SFT as a model with weight=1.0:7#   output = base + 1.0*(Think-SFT - base) + 0.5*(Math - base) + 0.5*(Code - base)8#9# Usage:10#   modal run modal_merge.py --config examples/olmo-think-math-code.yaml --hf-repo pmahdavi/Olmo-3-7B-Think-Math-Code11 12merge_method: task_arithmetic13base_model: allenai/Olmo-3-1025-7B14models:15  - model: allenai/Olmo-3-7B-Think-SFT16    parameters:17      weight: 1.018  - model: allenai/Olmo-3-7B-RL-Zero-Math19    parameters:20      weight: 0.521  - model: allenai/Olmo-3-7B-RL-Zero-Code22    parameters:23      weight: 0.524dtype: bfloat1625