bert-base-squad-jpqd-r0.030

BERT-base compressed by JPQD with Regularization Factor 0.03

F1: 87.66
EM: 80.23 

Description of important files

├── r0.030-squad-bert-b-mvmt-8bit
│   ├── 8bit_ref_bert_squad_nncf_mvmt.json (nncf config used with ssbs-feb branch)
│   ├── checkpoint-110000 (trained checkpoint for generation)
│   ├── ir
│   │   ├── sparsity_structures.csv
│   │   ├── sparsity_structures.md  (layer wise sparsity reporting, for linear layer in transformer block only)
│   │   ├── sparsity_structures.pkl (containing pruned structure id, e.g. particular head in MHSA or dimension in FFN, useful for debug)
│   │   └── squad-BertForQuestionAnswering.cropped.8bit.xml (custom discard of pruned dimension and onnx export, followed by ir translation
│   ├── ir_uncropped
│   │   ├── mo-pruned-ir
│   │   │   ├── mo.log (see Model Optimizer version here)
│   │   │   └── squad-BertForQuestionAnswering.8bit.xml (pruned structures are removed using Model Optimier --transform=Pruning)
│   │   └── squad-BertForQuestionAnswering.8bit.xml (pruned structures are sparsified/zero-ed only)