Upload 12 files
Browse files- .gitattributes +5 -0
- results/gqa_comparison/gqa_vs_mha_comparison.png +3 -0
- results/gqa_moe/best_model.pt +3 -0
- results/lb_comparison/expert_balance_comparison.png +3 -0
- results/lb_comparison/loss_comparison.png +3 -0
- results/lora_comparison/comparison_metrics.json +43 -0
- results/lora_moe/best_model.pt +3 -0
- results/moe_hash/best_model.pt +3 -0
- results/moe_hash/expert_usage_history.json +427 -0
- results/moe_hash/moe_hash_test_predictions.json +3 -0
- results/moe_hash/training_curves.png +3 -0
- results/moe_hash_without_lb/best_model.pt +3 -0
- results/moe_hash_without_lb/training_metrics.json +98 -0
.gitattributes
CHANGED
|
@@ -60,3 +60,8 @@ vizualisations/expert_usage_over_epochs_token_choice.png filter=lfs diff=lfs mer
|
|
| 60 |
vizualisations/routing_comparison.png filter=lfs diff=lfs merge=lfs -text
|
| 61 |
vizualisations/test_expert_usage_hash.png filter=lfs diff=lfs merge=lfs -text
|
| 62 |
vizualisations/test_expert_usage_token_choice.png filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 60 |
vizualisations/routing_comparison.png filter=lfs diff=lfs merge=lfs -text
|
| 61 |
vizualisations/test_expert_usage_hash.png filter=lfs diff=lfs merge=lfs -text
|
| 62 |
vizualisations/test_expert_usage_token_choice.png filter=lfs diff=lfs merge=lfs -text
|
| 63 |
+
results/gqa_comparison/gqa_vs_mha_comparison.png filter=lfs diff=lfs merge=lfs -text
|
| 64 |
+
results/lb_comparison/expert_balance_comparison.png filter=lfs diff=lfs merge=lfs -text
|
| 65 |
+
results/lb_comparison/loss_comparison.png filter=lfs diff=lfs merge=lfs -text
|
| 66 |
+
results/moe_hash/moe_hash_test_predictions.json filter=lfs diff=lfs merge=lfs -text
|
| 67 |
+
results/moe_hash/training_curves.png filter=lfs diff=lfs merge=lfs -text
|
results/gqa_comparison/gqa_vs_mha_comparison.png
ADDED
|
Git LFS Details
|
results/gqa_moe/best_model.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:94f48872136d77019b1955056e8dcba3500bf1def0e6baaabb33398673f5a17f
|
| 3 |
+
size 701671341
|
results/lb_comparison/expert_balance_comparison.png
ADDED
|
Git LFS Details
|
results/lb_comparison/loss_comparison.png
ADDED
|
Git LFS Details
|
results/lora_comparison/comparison_metrics.json
ADDED
|
@@ -0,0 +1,43 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"standard_moe": {
|
| 3 |
+
"total_params": 179944868,
|
| 4 |
+
"trainable_params": 179944868,
|
| 5 |
+
"train_losses": [
|
| 6 |
+
6.993803797149658,
|
| 7 |
+
5.954492663574219,
|
| 8 |
+
5.53921777420044,
|
| 9 |
+
5.24759983291626,
|
| 10 |
+
5.0148901168823246
|
| 11 |
+
],
|
| 12 |
+
"val_losses": [
|
| 13 |
+
6.212501205444336,
|
| 14 |
+
5.760108711242676,
|
| 15 |
+
5.491593559265136,
|
| 16 |
+
5.284299610137939,
|
| 17 |
+
5.145402297973633
|
| 18 |
+
],
|
| 19 |
+
"best_val_loss": 5.145402297973633,
|
| 20 |
+
"training_time": 1290.9204235076904
|
| 21 |
+
},
|
| 22 |
+
"lora_moe": {
|
| 23 |
+
"total_params": 182828452,
|
| 24 |
+
"trainable_params": 2621440,
|
| 25 |
+
"lora_percent": 1.43382497161875,
|
| 26 |
+
"train_losses": [
|
| 27 |
+
7.029633666229248,
|
| 28 |
+
5.872530696105957,
|
| 29 |
+
5.428599870300293,
|
| 30 |
+
5.109180227661133,
|
| 31 |
+
4.84611487121582
|
| 32 |
+
],
|
| 33 |
+
"val_losses": [
|
| 34 |
+
6.160612037658692,
|
| 35 |
+
5.704808525085449,
|
| 36 |
+
5.452941730499267,
|
| 37 |
+
5.289849403381347,
|
| 38 |
+
5.177924713134765
|
| 39 |
+
],
|
| 40 |
+
"best_val_loss": 5.177924713134765,
|
| 41 |
+
"training_time": 18967.75845336914
|
| 42 |
+
}
|
| 43 |
+
}
|
results/lora_moe/best_model.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fe830c30e4365e4cc873e86af27dc33b40e32e7967070df6276f1184957d27c7
|
| 3 |
+
size 731567101
|
results/moe_hash/best_model.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:611bd48bcdf6619e04c9b8aa64375ba6d7e45b4f45855c82419c743de6c78213
|
| 3 |
+
size 2159269635
|
results/moe_hash/expert_usage_history.json
ADDED
|
@@ -0,0 +1,427 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"epoch": 1,
|
| 4 |
+
"usage": {
|
| 5 |
+
"encoder_layer_0": {
|
| 6 |
+
"0": 261455302,
|
| 7 |
+
"1": 262893842,
|
| 8 |
+
"2": 3077719,
|
| 9 |
+
"3": 3319592,
|
| 10 |
+
"4": 3253708,
|
| 11 |
+
"5": 3070038,
|
| 12 |
+
"6": 283558135,
|
| 13 |
+
"7": 282061392
|
| 14 |
+
},
|
| 15 |
+
"encoder_layer_1": {
|
| 16 |
+
"0": 240464181,
|
| 17 |
+
"1": 249133057,
|
| 18 |
+
"2": 19122883,
|
| 19 |
+
"3": 22247029,
|
| 20 |
+
"4": 24482469,
|
| 21 |
+
"5": 25912739,
|
| 22 |
+
"6": 267275331,
|
| 23 |
+
"7": 254052039
|
| 24 |
+
},
|
| 25 |
+
"encoder_layer_2": {
|
| 26 |
+
"0": 207284432,
|
| 27 |
+
"1": 235091343,
|
| 28 |
+
"2": 64242756,
|
| 29 |
+
"3": 74434438,
|
| 30 |
+
"4": 72444627,
|
| 31 |
+
"5": 62671279,
|
| 32 |
+
"6": 207373049,
|
| 33 |
+
"7": 179147804
|
| 34 |
+
},
|
| 35 |
+
"encoder_layer_3": {
|
| 36 |
+
"0": 135826351,
|
| 37 |
+
"1": 181241685,
|
| 38 |
+
"2": 103904554,
|
| 39 |
+
"3": 118421274,
|
| 40 |
+
"4": 122321377,
|
| 41 |
+
"5": 112346323,
|
| 42 |
+
"6": 189292582,
|
| 43 |
+
"7": 139335582
|
| 44 |
+
},
|
| 45 |
+
"decoder_layer_0": {
|
| 46 |
+
"0": 9569968,
|
| 47 |
+
"1": 15139042,
|
| 48 |
+
"2": 10527019,
|
| 49 |
+
"3": 11571099,
|
| 50 |
+
"4": 14076991,
|
| 51 |
+
"5": 12499317,
|
| 52 |
+
"6": 15760636,
|
| 53 |
+
"7": 10725156
|
| 54 |
+
},
|
| 55 |
+
"decoder_layer_1": {
|
| 56 |
+
"0": 12376231,
|
| 57 |
+
"1": 15921848,
|
| 58 |
+
"2": 9631421,
|
| 59 |
+
"3": 14336388,
|
| 60 |
+
"4": 14508117,
|
| 61 |
+
"5": 9245838,
|
| 62 |
+
"6": 13418845,
|
| 63 |
+
"7": 10430540
|
| 64 |
+
},
|
| 65 |
+
"decoder_layer_2": {
|
| 66 |
+
"0": 24368930,
|
| 67 |
+
"1": 24373976,
|
| 68 |
+
"2": 9665,
|
| 69 |
+
"3": 8794,
|
| 70 |
+
"4": 7449,
|
| 71 |
+
"5": 5545,
|
| 72 |
+
"6": 25548570,
|
| 73 |
+
"7": 25546299
|
| 74 |
+
},
|
| 75 |
+
"decoder_layer_3": {
|
| 76 |
+
"0": 13153999,
|
| 77 |
+
"1": 17946588,
|
| 78 |
+
"2": 10940872,
|
| 79 |
+
"3": 11284613,
|
| 80 |
+
"4": 12384130,
|
| 81 |
+
"5": 10910761,
|
| 82 |
+
"6": 13455613,
|
| 83 |
+
"7": 9792652
|
| 84 |
+
}
|
| 85 |
+
}
|
| 86 |
+
},
|
| 87 |
+
{
|
| 88 |
+
"epoch": 2,
|
| 89 |
+
"usage": {
|
| 90 |
+
"encoder_layer_0": {
|
| 91 |
+
"0": 261455302,
|
| 92 |
+
"1": 262893842,
|
| 93 |
+
"2": 3077719,
|
| 94 |
+
"3": 3319592,
|
| 95 |
+
"4": 3253708,
|
| 96 |
+
"5": 3070038,
|
| 97 |
+
"6": 283558135,
|
| 98 |
+
"7": 282061392
|
| 99 |
+
},
|
| 100 |
+
"encoder_layer_1": {
|
| 101 |
+
"0": 240464181,
|
| 102 |
+
"1": 249133057,
|
| 103 |
+
"2": 19122883,
|
| 104 |
+
"3": 22247029,
|
| 105 |
+
"4": 24482469,
|
| 106 |
+
"5": 25912739,
|
| 107 |
+
"6": 267275331,
|
| 108 |
+
"7": 254052039
|
| 109 |
+
},
|
| 110 |
+
"encoder_layer_2": {
|
| 111 |
+
"0": 207284432,
|
| 112 |
+
"1": 235091343,
|
| 113 |
+
"2": 64242756,
|
| 114 |
+
"3": 74434438,
|
| 115 |
+
"4": 72444627,
|
| 116 |
+
"5": 62671279,
|
| 117 |
+
"6": 207373049,
|
| 118 |
+
"7": 179147804
|
| 119 |
+
},
|
| 120 |
+
"encoder_layer_3": {
|
| 121 |
+
"0": 135826351,
|
| 122 |
+
"1": 181241685,
|
| 123 |
+
"2": 103904554,
|
| 124 |
+
"3": 118421274,
|
| 125 |
+
"4": 122321377,
|
| 126 |
+
"5": 112346323,
|
| 127 |
+
"6": 189292582,
|
| 128 |
+
"7": 139335582
|
| 129 |
+
},
|
| 130 |
+
"decoder_layer_0": {
|
| 131 |
+
"0": 9569968,
|
| 132 |
+
"1": 15139042,
|
| 133 |
+
"2": 10527019,
|
| 134 |
+
"3": 11571099,
|
| 135 |
+
"4": 14076991,
|
| 136 |
+
"5": 12499317,
|
| 137 |
+
"6": 15760636,
|
| 138 |
+
"7": 10725156
|
| 139 |
+
},
|
| 140 |
+
"decoder_layer_1": {
|
| 141 |
+
"0": 12376231,
|
| 142 |
+
"1": 15921848,
|
| 143 |
+
"2": 9631421,
|
| 144 |
+
"3": 14336388,
|
| 145 |
+
"4": 14508117,
|
| 146 |
+
"5": 9245838,
|
| 147 |
+
"6": 13418845,
|
| 148 |
+
"7": 10430540
|
| 149 |
+
},
|
| 150 |
+
"decoder_layer_2": {
|
| 151 |
+
"0": 24368930,
|
| 152 |
+
"1": 24373976,
|
| 153 |
+
"2": 9665,
|
| 154 |
+
"3": 8794,
|
| 155 |
+
"4": 7449,
|
| 156 |
+
"5": 5545,
|
| 157 |
+
"6": 25548570,
|
| 158 |
+
"7": 25546299
|
| 159 |
+
},
|
| 160 |
+
"decoder_layer_3": {
|
| 161 |
+
"0": 13153999,
|
| 162 |
+
"1": 17946588,
|
| 163 |
+
"2": 10940872,
|
| 164 |
+
"3": 11284613,
|
| 165 |
+
"4": 12384130,
|
| 166 |
+
"5": 10910761,
|
| 167 |
+
"6": 13455613,
|
| 168 |
+
"7": 9792652
|
| 169 |
+
}
|
| 170 |
+
}
|
| 171 |
+
},
|
| 172 |
+
{
|
| 173 |
+
"epoch": 3,
|
| 174 |
+
"usage": {
|
| 175 |
+
"encoder_layer_0": {
|
| 176 |
+
"0": 261455302,
|
| 177 |
+
"1": 262893842,
|
| 178 |
+
"2": 3077719,
|
| 179 |
+
"3": 3319592,
|
| 180 |
+
"4": 3253708,
|
| 181 |
+
"5": 3070038,
|
| 182 |
+
"6": 283558135,
|
| 183 |
+
"7": 282061392
|
| 184 |
+
},
|
| 185 |
+
"encoder_layer_1": {
|
| 186 |
+
"0": 240464181,
|
| 187 |
+
"1": 249133057,
|
| 188 |
+
"2": 19122883,
|
| 189 |
+
"3": 22247029,
|
| 190 |
+
"4": 24482469,
|
| 191 |
+
"5": 25912739,
|
| 192 |
+
"6": 267275331,
|
| 193 |
+
"7": 254052039
|
| 194 |
+
},
|
| 195 |
+
"encoder_layer_2": {
|
| 196 |
+
"0": 207284432,
|
| 197 |
+
"1": 235091343,
|
| 198 |
+
"2": 64242756,
|
| 199 |
+
"3": 74434438,
|
| 200 |
+
"4": 72444627,
|
| 201 |
+
"5": 62671279,
|
| 202 |
+
"6": 207373049,
|
| 203 |
+
"7": 179147804
|
| 204 |
+
},
|
| 205 |
+
"encoder_layer_3": {
|
| 206 |
+
"0": 135826351,
|
| 207 |
+
"1": 181241685,
|
| 208 |
+
"2": 103904554,
|
| 209 |
+
"3": 118421274,
|
| 210 |
+
"4": 122321377,
|
| 211 |
+
"5": 112346323,
|
| 212 |
+
"6": 189292582,
|
| 213 |
+
"7": 139335582
|
| 214 |
+
},
|
| 215 |
+
"decoder_layer_0": {
|
| 216 |
+
"0": 9569968,
|
| 217 |
+
"1": 15139042,
|
| 218 |
+
"2": 10527019,
|
| 219 |
+
"3": 11571099,
|
| 220 |
+
"4": 14076991,
|
| 221 |
+
"5": 12499317,
|
| 222 |
+
"6": 15760636,
|
| 223 |
+
"7": 10725156
|
| 224 |
+
},
|
| 225 |
+
"decoder_layer_1": {
|
| 226 |
+
"0": 12376231,
|
| 227 |
+
"1": 15921848,
|
| 228 |
+
"2": 9631421,
|
| 229 |
+
"3": 14336388,
|
| 230 |
+
"4": 14508117,
|
| 231 |
+
"5": 9245838,
|
| 232 |
+
"6": 13418845,
|
| 233 |
+
"7": 10430540
|
| 234 |
+
},
|
| 235 |
+
"decoder_layer_2": {
|
| 236 |
+
"0": 24368930,
|
| 237 |
+
"1": 24373976,
|
| 238 |
+
"2": 9665,
|
| 239 |
+
"3": 8794,
|
| 240 |
+
"4": 7449,
|
| 241 |
+
"5": 5545,
|
| 242 |
+
"6": 25548570,
|
| 243 |
+
"7": 25546299
|
| 244 |
+
},
|
| 245 |
+
"decoder_layer_3": {
|
| 246 |
+
"0": 13153999,
|
| 247 |
+
"1": 17946588,
|
| 248 |
+
"2": 10940872,
|
| 249 |
+
"3": 11284613,
|
| 250 |
+
"4": 12384130,
|
| 251 |
+
"5": 10910761,
|
| 252 |
+
"6": 13455613,
|
| 253 |
+
"7": 9792652
|
| 254 |
+
}
|
| 255 |
+
}
|
| 256 |
+
},
|
| 257 |
+
{
|
| 258 |
+
"epoch": 4,
|
| 259 |
+
"usage": {
|
| 260 |
+
"encoder_layer_0": {
|
| 261 |
+
"0": 261455302,
|
| 262 |
+
"1": 262893842,
|
| 263 |
+
"2": 3077719,
|
| 264 |
+
"3": 3319592,
|
| 265 |
+
"4": 3253708,
|
| 266 |
+
"5": 3070038,
|
| 267 |
+
"6": 283558135,
|
| 268 |
+
"7": 282061392
|
| 269 |
+
},
|
| 270 |
+
"encoder_layer_1": {
|
| 271 |
+
"0": 240464181,
|
| 272 |
+
"1": 249133057,
|
| 273 |
+
"2": 19122883,
|
| 274 |
+
"3": 22247029,
|
| 275 |
+
"4": 24482469,
|
| 276 |
+
"5": 25912739,
|
| 277 |
+
"6": 267275331,
|
| 278 |
+
"7": 254052039
|
| 279 |
+
},
|
| 280 |
+
"encoder_layer_2": {
|
| 281 |
+
"0": 207284432,
|
| 282 |
+
"1": 235091343,
|
| 283 |
+
"2": 64242756,
|
| 284 |
+
"3": 74434438,
|
| 285 |
+
"4": 72444627,
|
| 286 |
+
"5": 62671279,
|
| 287 |
+
"6": 207373049,
|
| 288 |
+
"7": 179147804
|
| 289 |
+
},
|
| 290 |
+
"encoder_layer_3": {
|
| 291 |
+
"0": 135826351,
|
| 292 |
+
"1": 181241685,
|
| 293 |
+
"2": 103904554,
|
| 294 |
+
"3": 118421274,
|
| 295 |
+
"4": 122321377,
|
| 296 |
+
"5": 112346323,
|
| 297 |
+
"6": 189292582,
|
| 298 |
+
"7": 139335582
|
| 299 |
+
},
|
| 300 |
+
"decoder_layer_0": {
|
| 301 |
+
"0": 9569968,
|
| 302 |
+
"1": 15139042,
|
| 303 |
+
"2": 10527019,
|
| 304 |
+
"3": 11571099,
|
| 305 |
+
"4": 14076991,
|
| 306 |
+
"5": 12499317,
|
| 307 |
+
"6": 15760636,
|
| 308 |
+
"7": 10725156
|
| 309 |
+
},
|
| 310 |
+
"decoder_layer_1": {
|
| 311 |
+
"0": 12376231,
|
| 312 |
+
"1": 15921848,
|
| 313 |
+
"2": 9631421,
|
| 314 |
+
"3": 14336388,
|
| 315 |
+
"4": 14508117,
|
| 316 |
+
"5": 9245838,
|
| 317 |
+
"6": 13418845,
|
| 318 |
+
"7": 10430540
|
| 319 |
+
},
|
| 320 |
+
"decoder_layer_2": {
|
| 321 |
+
"0": 24368930,
|
| 322 |
+
"1": 24373976,
|
| 323 |
+
"2": 9665,
|
| 324 |
+
"3": 8794,
|
| 325 |
+
"4": 7449,
|
| 326 |
+
"5": 5545,
|
| 327 |
+
"6": 25548570,
|
| 328 |
+
"7": 25546299
|
| 329 |
+
},
|
| 330 |
+
"decoder_layer_3": {
|
| 331 |
+
"0": 13153999,
|
| 332 |
+
"1": 17946588,
|
| 333 |
+
"2": 10940872,
|
| 334 |
+
"3": 11284613,
|
| 335 |
+
"4": 12384130,
|
| 336 |
+
"5": 10910761,
|
| 337 |
+
"6": 13455613,
|
| 338 |
+
"7": 9792652
|
| 339 |
+
}
|
| 340 |
+
}
|
| 341 |
+
},
|
| 342 |
+
{
|
| 343 |
+
"epoch": 5,
|
| 344 |
+
"usage": {
|
| 345 |
+
"encoder_layer_0": {
|
| 346 |
+
"0": 261455302,
|
| 347 |
+
"1": 262893842,
|
| 348 |
+
"2": 3077719,
|
| 349 |
+
"3": 3319592,
|
| 350 |
+
"4": 3253708,
|
| 351 |
+
"5": 3070038,
|
| 352 |
+
"6": 283558135,
|
| 353 |
+
"7": 282061392
|
| 354 |
+
},
|
| 355 |
+
"encoder_layer_1": {
|
| 356 |
+
"0": 240464181,
|
| 357 |
+
"1": 249133057,
|
| 358 |
+
"2": 19122883,
|
| 359 |
+
"3": 22247029,
|
| 360 |
+
"4": 24482469,
|
| 361 |
+
"5": 25912739,
|
| 362 |
+
"6": 267275331,
|
| 363 |
+
"7": 254052039
|
| 364 |
+
},
|
| 365 |
+
"encoder_layer_2": {
|
| 366 |
+
"0": 207284432,
|
| 367 |
+
"1": 235091343,
|
| 368 |
+
"2": 64242756,
|
| 369 |
+
"3": 74434438,
|
| 370 |
+
"4": 72444627,
|
| 371 |
+
"5": 62671279,
|
| 372 |
+
"6": 207373049,
|
| 373 |
+
"7": 179147804
|
| 374 |
+
},
|
| 375 |
+
"encoder_layer_3": {
|
| 376 |
+
"0": 135826351,
|
| 377 |
+
"1": 181241685,
|
| 378 |
+
"2": 103904554,
|
| 379 |
+
"3": 118421274,
|
| 380 |
+
"4": 122321377,
|
| 381 |
+
"5": 112346323,
|
| 382 |
+
"6": 189292582,
|
| 383 |
+
"7": 139335582
|
| 384 |
+
},
|
| 385 |
+
"decoder_layer_0": {
|
| 386 |
+
"0": 9569968,
|
| 387 |
+
"1": 15139042,
|
| 388 |
+
"2": 10527019,
|
| 389 |
+
"3": 11571099,
|
| 390 |
+
"4": 14076991,
|
| 391 |
+
"5": 12499317,
|
| 392 |
+
"6": 15760636,
|
| 393 |
+
"7": 10725156
|
| 394 |
+
},
|
| 395 |
+
"decoder_layer_1": {
|
| 396 |
+
"0": 12376231,
|
| 397 |
+
"1": 15921848,
|
| 398 |
+
"2": 9631421,
|
| 399 |
+
"3": 14336388,
|
| 400 |
+
"4": 14508117,
|
| 401 |
+
"5": 9245838,
|
| 402 |
+
"6": 13418845,
|
| 403 |
+
"7": 10430540
|
| 404 |
+
},
|
| 405 |
+
"decoder_layer_2": {
|
| 406 |
+
"0": 24368930,
|
| 407 |
+
"1": 24373976,
|
| 408 |
+
"2": 9665,
|
| 409 |
+
"3": 8794,
|
| 410 |
+
"4": 7449,
|
| 411 |
+
"5": 5545,
|
| 412 |
+
"6": 25548570,
|
| 413 |
+
"7": 25546299
|
| 414 |
+
},
|
| 415 |
+
"decoder_layer_3": {
|
| 416 |
+
"0": 13153999,
|
| 417 |
+
"1": 17946588,
|
| 418 |
+
"2": 10940872,
|
| 419 |
+
"3": 11284613,
|
| 420 |
+
"4": 12384130,
|
| 421 |
+
"5": 10910761,
|
| 422 |
+
"6": 13455613,
|
| 423 |
+
"7": 9792652
|
| 424 |
+
}
|
| 425 |
+
}
|
| 426 |
+
}
|
| 427 |
+
]
|
results/moe_hash/moe_hash_test_predictions.json
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:aebe85737a13dc11530e579868d68fa7fd58f6f8d6cebea9fde02914242db6f4
|
| 3 |
+
size 27675266
|
results/moe_hash/training_curves.png
ADDED
|
Git LFS Details
|
results/moe_hash_without_lb/best_model.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f0bb4d7bf522808334d4eb22cc1ef24c434056ebc54f4c4e8f64cf5f16a2ceec
|
| 3 |
+
size 2159271619
|
results/moe_hash_without_lb/training_metrics.json
ADDED
|
@@ -0,0 +1,98 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"train_losses": [
|
| 3 |
+
7.003178263092041,
|
| 4 |
+
5.8915835350036625,
|
| 5 |
+
5.497518934631348,
|
| 6 |
+
5.223627810668945,
|
| 7 |
+
4.998379559326172
|
| 8 |
+
],
|
| 9 |
+
"train_nll_losses": [
|
| 10 |
+
7.003178263092041,
|
| 11 |
+
5.8915835350036625,
|
| 12 |
+
5.497518934631348,
|
| 13 |
+
5.223627810668945,
|
| 14 |
+
4.998379559326172
|
| 15 |
+
],
|
| 16 |
+
"train_lb_losses": [
|
| 17 |
+
0.0,
|
| 18 |
+
0.0,
|
| 19 |
+
0.0,
|
| 20 |
+
0.0,
|
| 21 |
+
0.0
|
| 22 |
+
],
|
| 23 |
+
"val_losses": [
|
| 24 |
+
6.157535878438798,
|
| 25 |
+
5.826862653096517,
|
| 26 |
+
5.508088460044255,
|
| 27 |
+
5.584775901976085,
|
| 28 |
+
5.546320551917667
|
| 29 |
+
],
|
| 30 |
+
"use_load_balancer": false,
|
| 31 |
+
"expert_balance_history": [
|
| 32 |
+
{
|
| 33 |
+
"epoch": 1,
|
| 34 |
+
"cv_scores": {
|
| 35 |
+
"encoder_layer_0": 0.9752403928100551,
|
| 36 |
+
"encoder_layer_1": 1.1604373739140559,
|
| 37 |
+
"encoder_layer_2": 1.1906502000710109,
|
| 38 |
+
"encoder_layer_3": 1.5189951633226693,
|
| 39 |
+
"decoder_layer_0": 1.6397420562599254,
|
| 40 |
+
"decoder_layer_1": 1.6088734749302642,
|
| 41 |
+
"decoder_layer_2": 1.1485815213057098,
|
| 42 |
+
"decoder_layer_3": 1.031611726454343
|
| 43 |
+
}
|
| 44 |
+
},
|
| 45 |
+
{
|
| 46 |
+
"epoch": 2,
|
| 47 |
+
"cv_scores": {
|
| 48 |
+
"encoder_layer_0": 1.2149520736371027,
|
| 49 |
+
"encoder_layer_1": 0.9475947854990925,
|
| 50 |
+
"encoder_layer_2": 0.973835742113084,
|
| 51 |
+
"encoder_layer_3": 1.5428490392307177,
|
| 52 |
+
"decoder_layer_0": 1.6692692285462098,
|
| 53 |
+
"decoder_layer_1": 1.6462829796890173,
|
| 54 |
+
"decoder_layer_2": 1.26960726561445,
|
| 55 |
+
"decoder_layer_3": 1.2212484174287248
|
| 56 |
+
}
|
| 57 |
+
},
|
| 58 |
+
{
|
| 59 |
+
"epoch": 3,
|
| 60 |
+
"cv_scores": {
|
| 61 |
+
"encoder_layer_0": 1.3859582784712385,
|
| 62 |
+
"encoder_layer_1": 0.774218401617478,
|
| 63 |
+
"encoder_layer_2": 0.7892838172863104,
|
| 64 |
+
"encoder_layer_3": 1.471245050784382,
|
| 65 |
+
"decoder_layer_0": 1.6428237035357693,
|
| 66 |
+
"decoder_layer_1": 1.6575447747063468,
|
| 67 |
+
"decoder_layer_2": 1.3360894947088227,
|
| 68 |
+
"decoder_layer_3": 1.2850289355113476
|
| 69 |
+
}
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 4,
|
| 73 |
+
"cv_scores": {
|
| 74 |
+
"encoder_layer_0": 1.470608977224471,
|
| 75 |
+
"encoder_layer_1": 0.8325088809517526,
|
| 76 |
+
"encoder_layer_2": 0.9110868199883019,
|
| 77 |
+
"encoder_layer_3": 1.4852236088974513,
|
| 78 |
+
"decoder_layer_0": 1.6453405313179363,
|
| 79 |
+
"decoder_layer_1": 1.6646367529819495,
|
| 80 |
+
"decoder_layer_2": 1.368985176021636,
|
| 81 |
+
"decoder_layer_3": 1.3377243523199316
|
| 82 |
+
}
|
| 83 |
+
},
|
| 84 |
+
{
|
| 85 |
+
"epoch": 5,
|
| 86 |
+
"cv_scores": {
|
| 87 |
+
"encoder_layer_0": 1.521552219913538,
|
| 88 |
+
"encoder_layer_1": 0.8826608997652614,
|
| 89 |
+
"encoder_layer_2": 1.0384956360485216,
|
| 90 |
+
"encoder_layer_3": 1.4437078323981245,
|
| 91 |
+
"decoder_layer_0": 1.6101629549482483,
|
| 92 |
+
"decoder_layer_1": 1.6548466841970308,
|
| 93 |
+
"decoder_layer_2": 1.3714362330355638,
|
| 94 |
+
"decoder_layer_3": 1.3546301554016826
|
| 95 |
+
}
|
| 96 |
+
}
|
| 97 |
+
]
|
| 98 |
+
}
|