{ "base_model": "deepseek-ai/deepseek-math-7b-base", "shared_expert": "/scratch/ctoxtli/moexp/models/expert_shared/final", "domain_experts": { "algebraic_geometry": "/scratch/ctoxtli/moexp/models/expert_algebraic_geometry/final", "discrete_math": "/scratch/ctoxtli/moexp/models/expert_discrete_math/final", "number_theory": "/scratch/ctoxtli/moexp/models/expert_number_theory/final", "analysis": "/scratch/ctoxtli/moexp/models/expert_analysis/final", "algebra": "/scratch/ctoxtli/moexp/models/expert_algebra/final", "geometry_topology": "/scratch/ctoxtli/moexp/models/expert_geometry_topology/final", "probability_statistics": "/scratch/ctoxtli/moexp/models/expert_probability_statistics/final" }, "router": "/scratch/ctoxtli/moexp/models/moe_assembled/router.pt", "num_experts": 7, "top_k": 2, "domains": [ "algebraic_geometry", "discrete_math", "number_theory", "analysis", "algebra", "geometry_topology", "probability_statistics" ], "timestamp": "2026-02-25 20:59:34" }