mirror of
https://github.com/deepseek-ai/ESFT
synced 2024-11-22 03:27:38 +00:00
18d23501ab
update readme update readme update readme Update benchmarks.py Update download_adapters.sh Update esft.py
1 line
740 B
JSON
1 line
740 B
JSON
{"experts": {"1": [8, 26, 55, 6, 20], "10": [33, 45, 59, 50, 41], "11": [28, 54, 16, 63, 58], "12": [41, 44, 61, 57, 28], "13": [6, 37, 29, 63], "14": [59, 43, 5, 11], "15": [0, 6, 15, 9, 16], "16": [20, 10, 24, 15], "17": [25, 61, 57, 36, 2], "18": [18, 6, 21, 0], "19": [32, 12, 14, 47, 1], "2": [57, 19, 47, 60, 35, 42], "20": [15, 4, 16, 25, 5], "21": [48, 25, 0, 54, 41], "22": [8, 61, 50, 2], "23": [46, 57, 25, 38, 23], "24": [36, 24, 28, 51], "25": [10, 44, 6, 54, 18], "26": [26, 48, 56, 36, 62], "3": [9, 13, 42, 15, 20, 22], "4": [59, 29, 13, 26, 48, 41], "5": [10, 41, 35, 56, 42], "6": [6, 35, 14, 33, 29], "7": [44, 8, 47, 50], "8": [6, 8, 17, 11], "9": [46, 53, 26, 61]}, "shared_experts": false, "non_expert_modules": false} |