DeepSeek/ESFT
1
0
Fork 0
mirror of https://github.com/deepseek-ai/ESFT synced 2025-06-26 18:15:50 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

first commit

Browse Source 
update readme

update readme

update readme

Update benchmarks.py

Update download_adapters.sh

Update esft.py
This commit is contained in:
Zihan Wang
2024-07-04 21:37:15 +08:00
committed by ZihanWang314
commit 18d23501ab
141 changed files with 48220 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
results/expert_configs/intent.json Normal file
View File

@@ -0,0 +1 @@
{"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}

1
results/expert_configs/law.json Normal file
View File

@@ -0,0 +1 @@
{"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}

1
results/expert_configs/summary.json Normal file
View File

@@ -0,0 +1 @@
{"experts": {"1": [62, 25, 32, 30, 27, 55, 48, 33], "10": [0, 43, 37, 51, 1], "11": [62, 24, 57, 55], "12": [30, 17, 25, 28, 48, 42], "13": [19, 55, 51, 44, 21, 3], "14": [40, 5, 12, 36], "15": [8, 29, 30, 23, 60], "16": [24, 34, 3, 62, 38], "17": [55, 56, 42, 13, 57, 46, 52, 8], "18": [51, 35, 30, 25, 0], "19": [17, 53, 56, 47], "2": [7, 51, 33, 19, 42, 20, 46], "20": [57, 5, 26, 17], "21": [31, 53, 54], "22": [6, 25, 61, 57, 19], "23": [4, 36, 2, 60], "24": [20, 14, 16, 21, 53], "25": [57, 49, 8, 37], "26": [48, 36], "3": [58, 45, 44, 7, 22, 2], "4": [59, 9, 52, 13, 61, 26], "5": [31, 42, 15, 55, 35], "6": [6, 14, 19, 37, 12, 40], "7": [33, 20, 17, 16], "8": [46, 25, 2, 17], "9": [25, 46, 0]}, "shared_experts": false, "non_expert_modules": false}

1
results/expert_configs/translation.json Normal file
View File

@@ -0,0 +1 @@
{"experts": {"1": [35, 22, 13, 43], "10": [25, 9, 13], "11": [4, 17], "12": [28, 61, 56], "13": [15, 14, 45], "14": [5, 56, 27, 46], "15": [23, 30, 55], "16": [16, 9, 20], "17": [26, 62, 29], "18": [30, 59, 52], "19": [24, 22, 58], "2": [23, 33, 1], "20": [38, 43, 4], "21": [17, 15, 53], "22": [11, 55, 5], "23": [58, 54, 46, 4], "24": [47, 23, 57, 1], "25": [63, 29, 52, 46], "26": [56, 17, 21, 52], "3": [15, 5, 42, 14], "4": [51, 2, 59], "5": [55, 53, 63], "6": [23, 55, 41], "7": [16, 46, 32], "8": [63, 25], "9": [0, 44, 12]}, "shared_experts": false, "non_expert_modules": false}