Feature Extraction
Transformers
Safetensors
hetero_moe_projspace
mixture-of-experts
heterogeneous
dimension-projection-space
ffn-fusion
custom_code
Instructions to use redmingwei/HeteroMoEProjSpace-v2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use redmingwei/HeteroMoEProjSpace-v2 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("feature-extraction", model="redmingwei/HeteroMoEProjSpace-v2", trust_remote_code=True)# pip install -U transformers accelerate # Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("redmingwei/HeteroMoEProjSpace-v2", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
v62.4: report.json
Browse files- checkpoints/v62.4/report.json +133 -133
checkpoints/v62.4/report.json
CHANGED
|
@@ -2,186 +2,186 @@
|
|
| 2 |
"hist": [
|
| 3 |
{
|
| 4 |
"step": 0,
|
| 5 |
-
"ce": 1.
|
| 6 |
-
"train_acc": 0.
|
| 7 |
-
"val_acc": 0.
|
| 8 |
-
"bal": 1.
|
| 9 |
"load": [
|
| 10 |
-
0.
|
| 11 |
-
0.
|
| 12 |
-
0.
|
| 13 |
-
0.
|
| 14 |
],
|
| 15 |
-
"ent": 0.
|
| 16 |
-
"router_wnorm": 0.
|
| 17 |
-
"router_gnorm": 0.
|
| 18 |
-
"slot_ent": 0.
|
| 19 |
"slot_ema": {
|
| 20 |
-
"Q": 0.
|
| 21 |
-
"K": 0.
|
| 22 |
-
"V": 0.
|
| 23 |
-
"O": 0.
|
| 24 |
},
|
| 25 |
-
"loss_ent": -0.
|
| 26 |
},
|
| 27 |
{
|
| 28 |
"step": 250,
|
| 29 |
-
"ce": 0.
|
| 30 |
-
"train_acc": 0.
|
| 31 |
-
"val_acc": 0.
|
| 32 |
-
"bal": 1.
|
| 33 |
"load": [
|
| 34 |
-
0.
|
| 35 |
-
0.
|
| 36 |
-
0.
|
| 37 |
-
0.
|
| 38 |
],
|
| 39 |
-
"ent": 0.
|
| 40 |
-
"router_wnorm": 0.
|
| 41 |
-
"router_gnorm": 0.
|
| 42 |
-
"slot_ent": 0.
|
| 43 |
"slot_ema": {
|
| 44 |
-
"Q": 0.
|
| 45 |
-
"K": 0.
|
| 46 |
-
"V": 0.
|
| 47 |
-
"O": 0.
|
| 48 |
},
|
| 49 |
-
"loss_ent": -0.
|
| 50 |
},
|
| 51 |
{
|
| 52 |
"step": 500,
|
| 53 |
-
"ce": 0.
|
| 54 |
-
"train_acc": 0.
|
| 55 |
-
"val_acc": 0.
|
| 56 |
-
"bal": 1.
|
| 57 |
"load": [
|
| 58 |
-
0.
|
| 59 |
-
0.
|
| 60 |
-
0.
|
| 61 |
-
0.
|
| 62 |
],
|
| 63 |
-
"ent": 0.
|
| 64 |
-
"router_wnorm": 0.
|
| 65 |
-
"router_gnorm": 0.
|
| 66 |
-
"slot_ent": 0.
|
| 67 |
"slot_ema": {
|
| 68 |
-
"Q": 0.
|
| 69 |
-
"K": 0.
|
| 70 |
-
"V": 0.
|
| 71 |
-
"O": 0.
|
| 72 |
},
|
| 73 |
-
"loss_ent": -0.
|
| 74 |
},
|
| 75 |
{
|
| 76 |
"step": 750,
|
| 77 |
-
"ce": 0.
|
| 78 |
"train_acc": 0.90625,
|
| 79 |
-
"val_acc": 0.
|
| 80 |
-
"bal": 1.
|
| 81 |
"load": [
|
| 82 |
-
0.
|
| 83 |
-
0.
|
| 84 |
-
0.
|
| 85 |
-
0.
|
| 86 |
],
|
| 87 |
-
"ent": 0.
|
| 88 |
-
"router_wnorm": 0.
|
| 89 |
-
"router_gnorm": 0.
|
| 90 |
-
"slot_ent": 0.
|
| 91 |
"slot_ema": {
|
| 92 |
-
"Q": 0.
|
| 93 |
-
"K": 0.
|
| 94 |
-
"V": 0.
|
| 95 |
-
"O": 0.
|
| 96 |
},
|
| 97 |
-
"loss_ent": -
|
| 98 |
},
|
| 99 |
{
|
| 100 |
"step": 1000,
|
| 101 |
-
"ce": 0.
|
| 102 |
-
"train_acc": 0.
|
| 103 |
"val_acc": 0.8626373626373627,
|
| 104 |
-
"bal": 1.
|
| 105 |
"load": [
|
| 106 |
-
0.
|
| 107 |
-
0.
|
| 108 |
-
0.
|
| 109 |
-
0.
|
| 110 |
],
|
| 111 |
-
"ent": 0.
|
| 112 |
-
"router_wnorm": 0.
|
| 113 |
-
"router_gnorm": 0.
|
| 114 |
-
"slot_ent": 0.
|
| 115 |
"slot_ema": {
|
| 116 |
-
"Q": 0.
|
| 117 |
-
"K": 0.
|
| 118 |
-
"V": 0.
|
| 119 |
-
"O": 0.
|
| 120 |
},
|
| 121 |
-
"loss_ent": -
|
| 122 |
},
|
| 123 |
{
|
| 124 |
"step": 1250,
|
| 125 |
-
"ce": 0.
|
| 126 |
-
"train_acc": 0.
|
| 127 |
-
"val_acc": 0.
|
| 128 |
-
"bal": 1.
|
| 129 |
"load": [
|
| 130 |
-
0.
|
| 131 |
-
0.
|
| 132 |
-
0.
|
| 133 |
-
0.
|
| 134 |
],
|
| 135 |
-
"ent": 0.
|
| 136 |
-
"router_wnorm": 0.
|
| 137 |
-
"router_gnorm": 0.
|
| 138 |
-
"slot_ent": 0.
|
| 139 |
"slot_ema": {
|
| 140 |
-
"Q": 0.
|
| 141 |
-
"K": 0.
|
| 142 |
-
"V": 0.
|
| 143 |
-
"O": 0.
|
| 144 |
},
|
| 145 |
-
"loss_ent": -
|
| 146 |
},
|
| 147 |
{
|
| 148 |
"step": 1500,
|
| 149 |
-
"ce": 0.
|
| 150 |
"train_acc": 1.0,
|
| 151 |
-
"val_acc": 0.
|
| 152 |
-
"bal":
|
| 153 |
"load": [
|
| 154 |
-
0.
|
| 155 |
-
0.
|
| 156 |
-
0.
|
| 157 |
-
0.
|
| 158 |
],
|
| 159 |
-
"ent": 0.
|
| 160 |
-
"router_wnorm": 0.
|
| 161 |
-
"router_gnorm": 0.
|
| 162 |
-
"slot_ent": 0.
|
| 163 |
"slot_ema": {
|
| 164 |
-
"Q": 0.
|
| 165 |
-
"K": 0.
|
| 166 |
-
"V": 0.
|
| 167 |
-
"O": 0.
|
| 168 |
},
|
| 169 |
-
"loss_ent": -
|
| 170 |
}
|
| 171 |
],
|
| 172 |
-
"best_val_acc": 0.
|
| 173 |
"split": "按类别分层 80/20",
|
| 174 |
"eval": {
|
| 175 |
-
"acc_heldout": 0.
|
| 176 |
"chance": 0.25,
|
| 177 |
"load": [
|
| 178 |
-
0.
|
| 179 |
-
0.
|
| 180 |
-
0.
|
| 181 |
-
0.
|
| 182 |
],
|
| 183 |
-
"entropy": 0.
|
| 184 |
-
"balance": 0.
|
| 185 |
"proj_space_slot_gate": {
|
| 186 |
"load": [
|
| 187 |
1.0,
|
|
@@ -189,18 +189,18 @@
|
|
| 189 |
1.0,
|
| 190 |
1.0
|
| 191 |
],
|
| 192 |
-
"entropy": 0.
|
| 193 |
"balance": 0.9999999989999999
|
| 194 |
},
|
| 195 |
"fusion_expert": {
|
| 196 |
"load": [
|
| 197 |
-
0.
|
| 198 |
-
0.
|
| 199 |
-
0.
|
| 200 |
-
0.
|
| 201 |
],
|
| 202 |
-
"entropy": 0.
|
| 203 |
-
"balance": 0.
|
| 204 |
}
|
| 205 |
},
|
| 206 |
"version": "62.4",
|
|
@@ -212,8 +212,8 @@
|
|
| 212 |
"bal_w": 0.01
|
| 213 |
},
|
| 214 |
"params": {
|
| 215 |
-
"proj_space":
|
| 216 |
-
"fusion":
|
| 217 |
},
|
| 218 |
"note": "融合层吃1024维,专家解冻(偏离项目原设定)"
|
| 219 |
}
|
|
|
|
| 2 |
"hist": [
|
| 3 |
{
|
| 4 |
"step": 0,
|
| 5 |
+
"ce": 1.3385847806930542,
|
| 6 |
+
"train_acc": 0.4375,
|
| 7 |
+
"val_acc": 0.36213786213786214,
|
| 8 |
+
"bal": 1.9764597415924072,
|
| 9 |
"load": [
|
| 10 |
+
0.086669921875,
|
| 11 |
+
0.8828125,
|
| 12 |
+
0.66259765625,
|
| 13 |
+
0.367919921875
|
| 14 |
],
|
| 15 |
+
"ent": 0.5117544566167753,
|
| 16 |
+
"router_wnorm": 0.7798083424568176,
|
| 17 |
+
"router_gnorm": 0.0,
|
| 18 |
+
"slot_ent": 0.5473473072052002,
|
| 19 |
"slot_ema": {
|
| 20 |
+
"Q": 0.24939,
|
| 21 |
+
"K": 0.24965,
|
| 22 |
+
"V": 0.25011,
|
| 23 |
+
"O": 0.25085
|
| 24 |
},
|
| 25 |
+
"loss_ent": -0.7587845325469971
|
| 26 |
},
|
| 27 |
{
|
| 28 |
"step": 250,
|
| 29 |
+
"ce": 0.5585795044898987,
|
| 30 |
+
"train_acc": 0.75,
|
| 31 |
+
"val_acc": 0.8601398601398601,
|
| 32 |
+
"bal": 1.9076179265975952,
|
| 33 |
"load": [
|
| 34 |
+
0.30029296875,
|
| 35 |
+
0.61279296875,
|
| 36 |
+
0.313232421875,
|
| 37 |
+
0.773681640625
|
| 38 |
],
|
| 39 |
+
"ent": 0.5042119424848733,
|
| 40 |
+
"router_wnorm": 0.7798083424568176,
|
| 41 |
+
"router_gnorm": 0.0,
|
| 42 |
+
"slot_ent": 0.06296975165605545,
|
| 43 |
"slot_ema": {
|
| 44 |
+
"Q": 0.2026,
|
| 45 |
+
"K": 0.02662,
|
| 46 |
+
"V": 0.12207,
|
| 47 |
+
"O": 0.64871
|
| 48 |
},
|
| 49 |
+
"loss_ent": -0.08729460835456848
|
| 50 |
},
|
| 51 |
{
|
| 52 |
"step": 500,
|
| 53 |
+
"ce": 0.33162036538124084,
|
| 54 |
+
"train_acc": 0.875,
|
| 55 |
+
"val_acc": 0.8741258741258742,
|
| 56 |
+
"bal": 1.7573927640914917,
|
| 57 |
"load": [
|
| 58 |
+
0.525390625,
|
| 59 |
+
0.4853515625,
|
| 60 |
+
0.396484375,
|
| 61 |
+
0.5927734375
|
| 62 |
],
|
| 63 |
+
"ent": 0.4636787703598875,
|
| 64 |
+
"router_wnorm": 0.7798083424568176,
|
| 65 |
+
"router_gnorm": 0.0,
|
| 66 |
+
"slot_ent": 0.06385663896799088,
|
| 67 |
"slot_ema": {
|
| 68 |
+
"Q": 0.16089,
|
| 69 |
+
"K": 0.01963,
|
| 70 |
+
"V": 0.01623,
|
| 71 |
+
"O": 0.80324
|
| 72 |
},
|
| 73 |
+
"loss_ent": -0.08852409571409225
|
| 74 |
},
|
| 75 |
{
|
| 76 |
"step": 750,
|
| 77 |
+
"ce": 0.24221858382225037,
|
| 78 |
"train_acc": 0.90625,
|
| 79 |
+
"val_acc": 0.8836163836163836,
|
| 80 |
+
"bal": 1.8517080545425415,
|
| 81 |
"load": [
|
| 82 |
+
0.648193359375,
|
| 83 |
+
0.416015625,
|
| 84 |
+
0.351806640625,
|
| 85 |
+
0.583984375
|
| 86 |
],
|
| 87 |
+
"ent": 0.4775710989399581,
|
| 88 |
+
"router_wnorm": 0.7798083424568176,
|
| 89 |
+
"router_gnorm": 0.0,
|
| 90 |
+
"slot_ent": 0.08177006244659424,
|
| 91 |
"slot_ema": {
|
| 92 |
+
"Q": 0.12117,
|
| 93 |
+
"K": 0.02607,
|
| 94 |
+
"V": 0.01805,
|
| 95 |
+
"O": 0.83471
|
| 96 |
},
|
| 97 |
+
"loss_ent": -0.11335738003253937
|
| 98 |
},
|
| 99 |
{
|
| 100 |
"step": 1000,
|
| 101 |
+
"ce": 0.0795382484793663,
|
| 102 |
+
"train_acc": 0.96875,
|
| 103 |
"val_acc": 0.8626373626373627,
|
| 104 |
+
"bal": 1.802364706993103,
|
| 105 |
"load": [
|
| 106 |
+
0.617431640625,
|
| 107 |
+
0.36962890625,
|
| 108 |
+
0.444580078125,
|
| 109 |
+
0.568359375
|
| 110 |
],
|
| 111 |
+
"ent": 0.47082520844685255,
|
| 112 |
+
"router_wnorm": 0.7798083424568176,
|
| 113 |
+
"router_gnorm": 0.0,
|
| 114 |
+
"slot_ent": 0.11181776225566864,
|
| 115 |
"slot_ema": {
|
| 116 |
+
"Q": 0.12738,
|
| 117 |
+
"K": 0.02665,
|
| 118 |
+
"V": 0.03122,
|
| 119 |
+
"O": 0.81475
|
| 120 |
},
|
| 121 |
+
"loss_ent": -0.1550123393535614
|
| 122 |
},
|
| 123 |
{
|
| 124 |
"step": 1250,
|
| 125 |
+
"ce": 0.17035461962223053,
|
| 126 |
+
"train_acc": 0.96875,
|
| 127 |
+
"val_acc": 0.8936063936063936,
|
| 128 |
+
"bal": 1.7554279565811157,
|
| 129 |
"load": [
|
| 130 |
+
0.42724609375,
|
| 131 |
+
0.368896484375,
|
| 132 |
+
0.58740234375,
|
| 133 |
+
0.616455078125
|
| 134 |
],
|
| 135 |
+
"ent": 0.4748968117147041,
|
| 136 |
+
"router_wnorm": 0.7798083424568176,
|
| 137 |
+
"router_gnorm": 0.0,
|
| 138 |
+
"slot_ent": 0.2617185711860657,
|
| 139 |
"slot_ema": {
|
| 140 |
+
"Q": 0.11925,
|
| 141 |
+
"K": 0.0456,
|
| 142 |
+
"V": 0.20374,
|
| 143 |
+
"O": 0.63142
|
| 144 |
},
|
| 145 |
+
"loss_ent": -0.36281898617744446
|
| 146 |
},
|
| 147 |
{
|
| 148 |
"step": 1500,
|
| 149 |
+
"ce": 0.03322920203208923,
|
| 150 |
"train_acc": 1.0,
|
| 151 |
+
"val_acc": 0.8916083916083916,
|
| 152 |
+
"bal": 1.6832056045532227,
|
| 153 |
"load": [
|
| 154 |
+
0.348388671875,
|
| 155 |
+
0.4345703125,
|
| 156 |
+
0.60107421875,
|
| 157 |
+
0.615966796875
|
| 158 |
],
|
| 159 |
+
"ent": 0.47328980582504204,
|
| 160 |
+
"router_wnorm": 0.7798083424568176,
|
| 161 |
+
"router_gnorm": 0.0,
|
| 162 |
+
"slot_ent": 0.3395337462425232,
|
| 163 |
"slot_ema": {
|
| 164 |
+
"Q": 0.13509,
|
| 165 |
+
"K": 0.1162,
|
| 166 |
+
"V": 0.22578,
|
| 167 |
+
"O": 0.52292
|
| 168 |
},
|
| 169 |
+
"loss_ent": -0.4706937372684479
|
| 170 |
}
|
| 171 |
],
|
| 172 |
+
"best_val_acc": 0.8936063936063936,
|
| 173 |
"split": "按类别分层 80/20",
|
| 174 |
"eval": {
|
| 175 |
+
"acc_heldout": 0.8936063936063936,
|
| 176 |
"chance": 0.25,
|
| 177 |
"load": [
|
| 178 |
+
0.4851203484015984,
|
| 179 |
+
0.47372549325674324,
|
| 180 |
+
0.4729645354645355,
|
| 181 |
+
0.5681896228771228
|
| 182 |
],
|
| 183 |
+
"entropy": 0.4658721940933291,
|
| 184 |
+
"balance": 0.8324061468021796,
|
| 185 |
"proj_space_slot_gate": {
|
| 186 |
"load": [
|
| 187 |
1.0,
|
|
|
|
| 189 |
1.0,
|
| 190 |
1.0
|
| 191 |
],
|
| 192 |
+
"entropy": 0.2820233207624333,
|
| 193 |
"balance": 0.9999999989999999
|
| 194 |
},
|
| 195 |
"fusion_expert": {
|
| 196 |
"load": [
|
| 197 |
+
0.4851203484015984,
|
| 198 |
+
0.47372549325674324,
|
| 199 |
+
0.4729645354645355,
|
| 200 |
+
0.5681896228771228
|
| 201 |
],
|
| 202 |
+
"entropy": 0.4658721940933291,
|
| 203 |
+
"balance": 0.8324061468021796
|
| 204 |
}
|
| 205 |
},
|
| 206 |
"version": "62.4",
|
|
|
|
| 212 |
"bal_w": 0.01
|
| 213 |
},
|
| 214 |
"params": {
|
| 215 |
+
"proj_space": 2867298,
|
| 216 |
+
"fusion": 50343940
|
| 217 |
},
|
| 218 |
"note": "融合层吃1024维,专家解冻(偏离项目原设定)"
|
| 219 |
}
|