mirror of
https://github.com/deepseek-ai/ESFT
synced 2024-11-22 11:37:57 +00:00
18d23501ab
update readme update readme update readme Update benchmarks.py Update download_adapters.sh Update esft.py
1 line
851 B
JSON
1 line
851 B
JSON
{"experts": {"1": [12, 40, 7, 13, 37, 16, 1, 17, 9], "10": [5, 57, 21, 14, 47, 15], "11": [44, 60, 14, 31, 61, 23], "12": [19, 14, 54, 46, 28, 12], "13": [25, 51, 56, 8, 34], "14": [50, 56, 25, 41, 52, 49], "15": [3, 29, 30, 58], "16": [34, 52, 14, 5, 54, 21], "17": [14, 25, 3, 5, 0, 62], "18": [52, 24, 31, 22, 11, 18], "19": [54, 50, 51], "2": [35, 58, 21, 16, 59, 22, 20, 31, 9], "20": [8, 42, 26, 6, 3], "21": [0, 5, 13, 46, 11, 30], "22": [46, 37, 47, 55, 39, 15], "23": [19, 2, 54, 35, 56], "24": [0, 29, 56], "25": [4, 14, 11, 25, 60], "26": [37, 22, 55, 49, 3, 20], "3": [27, 34, 42, 29, 38, 13, 1, 22], "4": [10, 37, 58, 45, 43, 40, 6], "5": [19, 16, 25, 40, 5, 42, 33], "6": [62, 21, 24, 15, 51, 4], "7": [60, 59, 23, 5, 63, 22], "8": [61, 37, 7, 32, 16], "9": [6, 56, 30, 59, 57, 50]}, "shared_experts": false, "non_expert_modules": false} |