upload model

.gitattributes

@@ -32,3 +32,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+*.xml filter=lfs diff=lfs merge=lfs -text
+*.mapping filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,28 @@
+---
+license: apache-2.0
+tags:
+    - generated_from_trainer
+datasets:
+    - squad
+model-index:
+    - name: mobilebert-uncased-squadv1-14blocks-structured39.8-int8
+      results: []
+---
+
+<!-- This model card has been generated automatically according to the information the Trainer had access to. You
+should probably proofread and complete it, then remove this comment. -->
+
+# mobilebert-uncased-squadv1-14blocks-structured39.8-int8
+
+This model is a fine-tuned version of [google/mobilebert-uncased](https://huggingface.co/google/mobilebert-uncased) on the squad dataset.
+
+Notice that this model only has the first 14 transformer blocks. It is quantized and structually pruned by NNCF. The sparsity in remaining linear layers is 39.8%.
+
+-   Torch f1: 90.15
+
+### Framework versions
+
+-   Transformers 4.25.1
+-   Pytorch 1.13.1+cu116
+-   Datasets 2.8.0
+-   Tokenizers 0.13.2

config.json

@@ -0,0 +1,33 @@
+{
+  "_name_or_path": "google/mobilebert-uncased",
+  "architectures": [
+    "NNCFNetwork"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "classifier_activation": false,
+  "classifier_dropout": null,
+  "embedding_size": 128,
+  "hidden_act": "relu",
+  "hidden_dropout_prob": 0.0,
+  "hidden_size": 512,
+  "initializer_range": 0.02,
+  "intermediate_size": 512,
+  "intra_bottleneck_size": 128,
+  "key_query_shared_bottleneck": true,
+  "layer_norm_eps": 1e-12,
+  "max_position_embeddings": 512,
+  "model_type": "mobilebert",
+  "normalization_type": "no_norm",
+  "num_attention_heads": 4,
+  "num_feedforward_networks": 4,
+  "num_hidden_layers": 14,
+  "pad_token_id": 0,
+  "torch_dtype": "float32",
+  "transformers_version": "4.25.1",
+  "trigram_input": true,
+  "true_hidden_size": 128,
+  "type_vocab_size": 2,
+  "use_bottleneck": true,
+  "use_bottleneck_attention": false,
+  "vocab_size": 30522
+}

openvino_config.json

@@ -0,0 +1,104 @@
+{
+  "compression": [
+    {
+      "algorithm": "movement_sparsity",
+      "ignored_scopes": [
+        "{re}.*MobileBertEmbeddings.*",
+        "{re}.*Bottleneck.*",
+        "{re}.*OutputBottleneck.*",
+        "{re}.*qa_outputs.*"
+      ],
+      "params": {
+        "enable_structured_masking": true,
+        "importance_regularization_factor": 0.065,
+        "warmup_end_epoch": 10,
+        "warmup_start_epoch": 3
+      },
+      "sparse_structure_by_scopes": [
+        {
+          "mode": "block",
+          "sparse_factors": [
+            16,
+            16
+          ],
+          "target_scopes": "{re}.*MobileBertAttention.*"
+        },
+        {
+          "axis": 0,
+          "mode": "per_dim",
+          "target_scopes": "{re}.*MobileBertIntermediate.*"
+        },
+        {
+          "axis": 1,
+          "mode": "per_dim",
+          "target_scopes": "{re}.*MobileBertOutput.*"
+        },
+        {
+          "axis": 1,
+          "mode": "per_dim",
+          "target_scopes": "{re}.*FFNOutput.*"
+        }
+      ]
+    },
+    {
+      "algorithm": "quantization",
+      "export_to_onnx_standard_ops": false,
+      "ignored_scopes": [
+        "{re}.*__add___[0-1]",
+        "{re}.*__truediv__*"
+      ],
+      "initializer": {
+        "batchnorm_adaptation": {
+          "num_bn_adaptation_samples": 16
+        },
+        "range": {
+          "num_init_samples": 128,
+          "params": {
+            "max_percentile": 99.99,
+            "min_percentile": 0.01
+          },
+          "type": "percentile"
+        }
+      },
+      "overflow_fix": "disable",
+      "preset": "mixed",
+      "scope_overrides": {
+        "activations": {
+          "{re}.*matmul_0": {
+            "mode": "symmetric"
+          }
+        }
+      }
+    }
+  ],
+  "input_info": [
+    {
+      "keyword": "input_ids",
+      "sample_size": [
+        32,
+        384
+      ],
+      "type": "long"
+    },
+    {
+      "keyword": "token_type_ids",
+      "sample_size": [
+        32,
+        384
+      ],
+      "type": "long"
+    },
+    {
+      "keyword": "attention_mask",
+      "sample_size": [
+        32,
+        384
+      ],
+      "type": "long"
+    }
+  ],
+  "log_dir": "/nvme2/yujiepan/workspace/jpqd-test/LOGS/optimum-mobilebert-qa/0127_ttev_ftz8ih_epo18lr2e-06teacher0.95cosDecayRestart",
+  "optimum_version": "1.6.1",
+  "save_onnx_model": false,
+  "transformers_version": "4.25.1"
+}

openvino_model.mapping

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:acc4c474811935d0ef15b528efd6873ba34c4a5c0122b9d77c8a2e40156d1883
+size 699863

openvino_model.xml

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fec9d0f0a4beaf0ff1e8e72c489dcc90d24a2e4adf375d66986ae20a5dab8d43
+size 2281114

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:88f176b8eabfceb10310ab3474f79f8c340d38a88a23c6776062f82bff6bfe97
+size 102725011

special_tokens_map.json

@@ -0,0 +1,7 @@
+{
+  "cls_token": "[CLS]",
+  "mask_token": "[MASK]",
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "unk_token": "[UNK]"
+}

tokenizer_config.json

@@ -0,0 +1,14 @@
+{
+  "cls_token": "[CLS]",
+  "do_lower_case": true,
+  "mask_token": "[MASK]",
+  "model_max_length": 1000000000000000019884624838656,
+  "name_or_path": "google/mobilebert-uncased",
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "special_tokens_map_file": null,
+  "strip_accents": null,
+  "tokenize_chinese_chars": true,
+  "tokenizer_class": "MobileBertTokenizer",
+  "unk_token": "[UNK]"
+}