bert-base-squadv1-block-pruning-hybrid-filled-lt-nncf-50.0sparse-qat-lt
/
XP_linear_layer_sparsity_20M_params_50.00_sparsity.md
| layer_id | layer_type | param_type | shape | nparam | nnz | sparsity | |
|---|---|---|---|---|---|---|---|
| 5 | nncf_module.bert.encoder.layer.0.attention.self.query | NNCFLinear | weight | [320, 768] | 245760 | 106337 | 0.567314 |
| 7 | nncf_module.bert.encoder.layer.0.attention.self.key | NNCFLinear | weight | [320, 768] | 245760 | 112375 | 0.542745 |
| 9 | nncf_module.bert.encoder.layer.0.attention.self.value | NNCFLinear | weight | [320, 768] | 245760 | 131965 | 0.463033 |
| 11 | nncf_module.bert.encoder.layer.0.attention.output.dense | NNCFLinear | weight | [768, 320] | 245760 | 137497 | 0.440523 |
| 15 | nncf_module.bert.encoder.layer.0.intermediate.dense | NNCFLinear | weight | [185, 768] | 142080 | 110056 | 0.225394 |
| 17 | nncf_module.bert.encoder.layer.0.output.dense | NNCFLinear | weight | [768, 185] | 142080 | 108631 | 0.235424 |
| 21 | nncf_module.bert.encoder.layer.1.attention.self.query | NNCFLinear | weight | [320, 768] | 245760 | 136222 | 0.445711 |
| 23 | nncf_module.bert.encoder.layer.1.attention.self.key | NNCFLinear | weight | [320, 768] | 245760 | 136267 | 0.445528 |
| 25 | nncf_module.bert.encoder.layer.1.attention.self.value | NNCFLinear | weight | [320, 768] | 245760 | 124892 | 0.491813 |
| 27 | nncf_module.bert.encoder.layer.1.attention.output.dense | NNCFLinear | weight | [768, 320] | 245760 | 130185 | 0.470276 |
| 31 | nncf_module.bert.encoder.layer.1.intermediate.dense | NNCFLinear | weight | [315, 768] | 241920 | 176203 | 0.271648 |
| 33 | nncf_module.bert.encoder.layer.1.output.dense | NNCFLinear | weight | [768, 315] | 241920 | 172407 | 0.287339 |
| 37 | nncf_module.bert.encoder.layer.2.attention.self.query | NNCFLinear | weight | [576, 768] | 442368 | 198811 | 0.550575 |
| 39 | nncf_module.bert.encoder.layer.2.attention.self.key | NNCFLinear | weight | [576, 768] | 442368 | 201526 | 0.544438 |
| 41 | nncf_module.bert.encoder.layer.2.attention.self.value | NNCFLinear | weight | [576, 768] | 442368 | 163541 | 0.630306 |
| 43 | nncf_module.bert.encoder.layer.2.attention.output.dense | NNCFLinear | weight | [768, 576] | 442368 | 167245 | 0.621932 |
| 47 | nncf_module.bert.encoder.layer.2.intermediate.dense | NNCFLinear | weight | [339, 768] | 260352 | 185491 | 0.287538 |
| 49 | nncf_module.bert.encoder.layer.2.output.dense | NNCFLinear | weight | [768, 339] | 260352 | 183083 | 0.296787 |
| 53 | nncf_module.bert.encoder.layer.3.attention.self.query | NNCFLinear | weight | [576, 768] | 442368 | 205898 | 0.534555 |
| 55 | nncf_module.bert.encoder.layer.3.attention.self.key | NNCFLinear | weight | [576, 768] | 442368 | 217621 | 0.508054 |
| 57 | nncf_module.bert.encoder.layer.3.attention.self.value | NNCFLinear | weight | [576, 768] | 442368 | 209726 | 0.525901 |
| 59 | nncf_module.bert.encoder.layer.3.attention.output.dense | NNCFLinear | weight | [768, 576] | 442368 | 208038 | 0.529717 |
| 63 | nncf_module.bert.encoder.layer.3.intermediate.dense | NNCFLinear | weight | [368, 768] | 282624 | 198154 | 0.298878 |
| 65 | nncf_module.bert.encoder.layer.3.output.dense | NNCFLinear | weight | [768, 368] | 282624 | 194127 | 0.313126 |
| 69 | nncf_module.bert.encoder.layer.4.attention.self.query | NNCFLinear | weight | [576, 768] | 442368 | 212567 | 0.519479 |
| 71 | nncf_module.bert.encoder.layer.4.attention.self.key | NNCFLinear | weight | [576, 768] | 442368 | 214788 | 0.514459 |
| 73 | nncf_module.bert.encoder.layer.4.attention.self.value | NNCFLinear | weight | [576, 768] | 442368 | 197159 | 0.55431 |
| 75 | nncf_module.bert.encoder.layer.4.attention.output.dense | NNCFLinear | weight | [768, 576] | 442368 | 192495 | 0.564853 |
| 79 | nncf_module.bert.encoder.layer.4.intermediate.dense | NNCFLinear | weight | [386, 768] | 296448 | 205830 | 0.305679 |
| 81 | nncf_module.bert.encoder.layer.4.output.dense | NNCFLinear | weight | [768, 386] | 296448 | 199567 | 0.326806 |
| 85 | nncf_module.bert.encoder.layer.5.attention.self.query | NNCFLinear | weight | [384, 768] | 294912 | 131257 | 0.554928 |
| 87 | nncf_module.bert.encoder.layer.5.attention.self.key | NNCFLinear | weight | [384, 768] | 294912 | 154458 | 0.476257 |
| 89 | nncf_module.bert.encoder.layer.5.attention.self.value | NNCFLinear | weight | [384, 768] | 294912 | 159646 | 0.458666 |
| 91 | nncf_module.bert.encoder.layer.5.attention.output.dense | NNCFLinear | weight | [768, 384] | 294912 | 156889 | 0.468014 |
| 95 | nncf_module.bert.encoder.layer.5.intermediate.dense | NNCFLinear | weight | [336, 768] | 258048 | 184774 | 0.283955 |
| 97 | nncf_module.bert.encoder.layer.5.output.dense | NNCFLinear | weight | [768, 336] | 258048 | 178574 | 0.307981 |
| 101 | nncf_module.bert.encoder.layer.6.attention.self.query | NNCFLinear | weight | [448, 768] | 344064 | 153126 | 0.554949 |
| 103 | nncf_module.bert.encoder.layer.6.attention.self.key | NNCFLinear | weight | [448, 768] | 344064 | 169706 | 0.50676 |
| 105 | nncf_module.bert.encoder.layer.6.attention.self.value | NNCFLinear | weight | [448, 768] | 344064 | 154213 | 0.55179 |
| 107 | nncf_module.bert.encoder.layer.6.attention.output.dense | NNCFLinear | weight | [768, 448] | 344064 | 148815 | 0.567479 |
| 111 | nncf_module.bert.encoder.layer.6.intermediate.dense | NNCFLinear | weight | [280, 768] | 215040 | 158800 | 0.261533 |
| 113 | nncf_module.bert.encoder.layer.6.output.dense | NNCFLinear | weight | [768, 280] | 215040 | 156397 | 0.272707 |
| 117 | nncf_module.bert.encoder.layer.7.attention.self.query | NNCFLinear | weight | [448, 768] | 344064 | 153876 | 0.552769 |
| 119 | nncf_module.bert.encoder.layer.7.attention.self.key | NNCFLinear | weight | [448, 768] | 344064 | 180384 | 0.475725 |
| 121 | nncf_module.bert.encoder.layer.7.attention.self.value | NNCFLinear | weight | [448, 768] | 344064 | 166552 | 0.515927 |
| 123 | nncf_module.bert.encoder.layer.7.attention.output.dense | NNCFLinear | weight | [768, 448] | 344064 | 160346 | 0.533965 |
| 127 | nncf_module.bert.encoder.layer.7.intermediate.dense | NNCFLinear | weight | [211, 768] | 162048 | 125197 | 0.227408 |
| 129 | nncf_module.bert.encoder.layer.7.output.dense | NNCFLinear | weight | [768, 211] | 162048 | 123451 | 0.238182 |
| 133 | nncf_module.bert.encoder.layer.8.attention.self.query | NNCFLinear | weight | [448, 768] | 344064 | 149419 | 0.565723 |
| 135 | nncf_module.bert.encoder.layer.8.attention.self.key | NNCFLinear | weight | [448, 768] | 344064 | 150699 | 0.562003 |
| 137 | nncf_module.bert.encoder.layer.8.attention.self.value | NNCFLinear | weight | [448, 768] | 344064 | 124770 | 0.637364 |
| 139 | nncf_module.bert.encoder.layer.8.attention.output.dense | NNCFLinear | weight | [768, 448] | 344064 | 120648 | 0.649344 |
| 143 | nncf_module.bert.encoder.layer.8.intermediate.dense | NNCFLinear | weight | [108, 768] | 82944 | 68989 | 0.168246 |
| 145 | nncf_module.bert.encoder.layer.8.output.dense | NNCFLinear | weight | [768, 108] | 82944 | 68556 | 0.173466 |
| 149 | nncf_module.bert.encoder.layer.9.attention.self.query | NNCFLinear | weight | [320, 768] | 245760 | 121225 | 0.506734 |
| 151 | nncf_module.bert.encoder.layer.9.attention.self.key | NNCFLinear | weight | [320, 768] | 245760 | 114789 | 0.532922 |
| 153 | nncf_module.bert.encoder.layer.9.attention.self.value | NNCFLinear | weight | [320, 768] | 245760 | 69260 | 0.71818 |
| 155 | nncf_module.bert.encoder.layer.9.attention.output.dense | NNCFLinear | weight | [768, 320] | 245760 | 73575 | 0.700623 |
| 159 | nncf_module.bert.encoder.layer.9.intermediate.dense | NNCFLinear | weight | [53, 768] | 40704 | 35399 | 0.130331 |
| 161 | nncf_module.bert.encoder.layer.9.output.dense | NNCFLinear | weight | [768, 53] | 40704 | 34918 | 0.142148 |
| 165 | nncf_module.bert.encoder.layer.10.attention.self.query | NNCFLinear | weight | [384, 768] | 294912 | 128113 | 0.565589 |
| 167 | nncf_module.bert.encoder.layer.10.attention.self.key | NNCFLinear | weight | [384, 768] | 294912 | 124553 | 0.57766 |
| 169 | nncf_module.bert.encoder.layer.10.attention.self.value | NNCFLinear | weight | [384, 768] | 294912 | 79608 | 0.730062 |
| 171 | nncf_module.bert.encoder.layer.10.attention.output.dense | NNCFLinear | weight | [768, 384] | 294912 | 85158 | 0.711243 |
| 175 | nncf_module.bert.encoder.layer.10.intermediate.dense | NNCFLinear | weight | [86, 768] | 66048 | 55025 | 0.166894 |
| 177 | nncf_module.bert.encoder.layer.10.output.dense | NNCFLinear | weight | [768, 86] | 66048 | 54584 | 0.173571 |
| 181 | nncf_module.bert.encoder.layer.11.attention.self.query | NNCFLinear | weight | [384, 768] | 294912 | 104737 | 0.644853 |
| 183 | nncf_module.bert.encoder.layer.11.attention.self.key | NNCFLinear | weight | [384, 768] | 294912 | 102558 | 0.652242 |
| 185 | nncf_module.bert.encoder.layer.11.attention.self.value | NNCFLinear | weight | [384, 768] | 294912 | 64855 | 0.780087 |
| 187 | nncf_module.bert.encoder.layer.11.attention.output.dense | NNCFLinear | weight | [768, 384] | 294912 | 69674 | 0.763747 |
| 191 | nncf_module.bert.encoder.layer.11.intermediate.dense | NNCFLinear | weight | [105, 768] | 80640 | 67724 | 0.160169 |
| 193 | nncf_module.bert.encoder.layer.11.output.dense | NNCFLinear | weight | [768, 105] | 80640 | 67519 | 0.162711 |