15 lines
427 B
JSON
15 lines
427 B
JSON
{
|
|
"base_model": "Qwen/Qwen3-1.7B",
|
|
"specialists": {
|
|
"math": "cs-552-2026-databand/math_model",
|
|
"knowledge": "cs-552-2026-databand/general_knowledge_model",
|
|
"safety": "cs-552-2026-databand/safety_model",
|
|
"multilingual": "cs-552-2026-databand/multilingual_model"
|
|
},
|
|
"density": 0.9,
|
|
"weight": 1.0,
|
|
"normalize": true,
|
|
"int8_mask": true,
|
|
"output_dir": "/scratch/runs/group_final",
|
|
"hf_repo": null
|
|
} |