Model files.

.gitattributes

@@ -25,3 +25,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zstandard filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+*.th filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,6 @@
+---
+tags:
+- allennlp
+---
+
+# TODO: Fill this model card

best.th

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:547d784aabec69b259f6f40ad469a2ccdf25a83673587a54cb01ee57da0c9ce5
+size 504059851

config.json

@@ -0,0 +1,177 @@
+{
+    "dataset_reader": {
+        "type": "multitask",
+        "readers": {
+            "ner": {
+                "type": "wikiann",
+                "token_indexers": {
+                    "transformer": {
+                        "type": "pretrained_transformer_mismatched",
+                        "max_length": 512,
+                        "model_name": "MLRS/BERTu"
+                    }
+                }
+            }
+        }
+    },
+    "model": {
+        "type": "multitask",
+        "arg_name_mapping": {
+            "backbone": {
+                "tokens": "text",
+                "words": "text"
+            }
+        },
+        "backbone": {
+            "type": "embedder_and_mask",
+            "text_field_embedder": {
+                "token_embedders": {
+                    "transformer": {
+                        "type": "pretrained_transformer_mismatched_with_dropout",
+                        "last_layer_only": false,
+                        "layer_dropout": 0.1,
+                        "max_length": 512,
+                        "model_name": "MLRS/BERTu",
+                        "tokenizer_kwargs": {},
+                        "train_parameters": true
+                    }
+                }
+            }
+        },
+        "heads": {
+            "ner": {
+                "type": "crf_tagger",
+                "calculate_span_f1": true,
+                "constrain_crf_decoding": true,
+                "dropout": 0.2,
+                "encoder": {
+                    "type": "pass_through",
+                    "input_dim": 768
+                },
+                "include_start_end_transitions": false,
+                "label_encoding": "BIO"
+            }
+        }
+    },
+    "train_data_path": {
+        "ner": "panx_dataset/mt/train"
+    },
+    "validation_data_path": {
+        "ner": "panx_dataset/mt/dev"
+    },
+    "trainer": {
+        "callbacks": [
+            {
+                "tensorboard_writer": {
+                    "should_log_learning_rate": true,
+                    "should_log_parameter_statistics": true
+                },
+                "type": "tensorboard"
+            }
+        ],
+        "cuda_device": 0,
+        "grad_norm": 5,
+        "learning_rate_scheduler": {
+            "type": "ulmfit_sqrt",
+            "affected_group_count": 2,
+            "decay_factor": 0.05,
+            "discriminative_fine_tuning": true,
+            "factor": 5,
+            "gradual_unfreezing": true,
+            "model_size": 1,
+            "start_step": 2,
+            "warmup_steps": 2
+        },
+        "num_epochs": 200,
+        "optimizer": {
+            "type": "huggingface_adamw",
+            "betas": [
+                0.9,
+                0.999
+            ],
+            "correct_bias": false,
+            "lr": 0.0005,
+            "parameter_groups": [
+                [
+                    [
+                        "text_field_embedder.*transformer_model.embeddings.*_embeddings.*",
+                        "text_field_embedder.*transformer_model.encoder.*.(key|query|value|dense).weight"
+                    ],
+                    {}
+                ],
+                [
+                    [
+                        "text_field_embedder.*transformer_model.embeddings.LayerNorm.*",
+                        "text_field_embedder.*transformer_model.encoder.*.output.LayerNorm.*",
+                        "text_field_embedder.*transformer_model.encoder.*.(key|query|value|dense).bias",
+                        "text_field_embedder.*transformer_model.pooler.dense.bias"
+                    ],
+                    {
+                        "weight_decay": 0
+                    }
+                ],
+                [
+                    [
+                        "text_field_embedder.*._scalar_mix.*",
+                        "text_field_embedder.*transformer_model.pooler.dense.weight",
+                        "_head_sentinel",
+                        "head_arc_feedforward._linear_layers.*.weight",
+                        "child_arc_feedforward._linear_layers.*.weight",
+                        "head_tag_feedforward._linear_layers.*.weight",
+                        "child_tag_feedforward._linear_layers.*.weight",
+                        "arc_attention._weight_matrix",
+                        "tag_bilinear.weight",
+                        "tag_projection_layer._module.weight",
+                        "crf",
+                        "linear.weight",
+                        "tagger_linear.weight"
+                    ],
+                    {}
+                ],
+                [
+                    [
+                        "head_arc_feedforward._linear_layers.*.bias",
+                        "child_arc_feedforward._linear_layers.*.bias",
+                        "head_tag_feedforward._linear_layers.*.bias",
+                        "child_tag_feedforward._linear_layers.*.bias",
+                        "arc_attention._bias",
+                        "tag_bilinear.bias",
+                        "tag_projection_layer._module.bias",
+                        "linear.bias",
+                        "tagger_linear.bias"
+                    ],
+                    {
+                        "weight_decay": 0
+                    }
+                ]
+            ],
+            "weight_decay": 0.01
+        },
+        "patience": 20,
+        "validation_metric": [
+            "+ner_f1-measure-overall"
+        ]
+    },
+    "data_loader": {
+        "type": "multitask",
+        "scheduler": {
+            "type": "unbalanced_homogeneous_roundrobin",
+            "batch_size": 64,
+            "dataset_sizes": {
+                "ner": 100
+            }
+        },
+        "shuffle": true
+    },
+    "numpy_seed": 1337,
+    "pytorch_seed": 133,
+    "random_seed": 13370,
+    "validation_data_loader": {
+        "type": "multitask",
+        "scheduler": {
+            "type": "homogeneous_roundrobin",
+            "batch_size": 64
+        },
+        "shuffle": true
+    }
+}

log/train/events.out.tfevents

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:74d9f10dae08413337cb5f4642d4a33d011931ebc1d467fc252d1c5a962379ee
+size 651225

log/validation/events.out.tfevents

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5c75764f815d7056d7e8b0983fae22d57a1811b2ae2643db7f6fd72a70b1db68
+size 16894

metrics.json

@@ -0,0 +1,29 @@
+{
+  "best_epoch": 33,
+  "peak_worker_0_memory_MB": 4608.48828125,
+  "peak_gpu_0_memory_MB": 4295.302734375,
+  "training_duration": "0:02:50.854943",
+  "training_start_epoch": 0,
+  "training_epochs": 52,
+  "epoch": 52,
+  "training_ner_accuracy": 1.0,
+  "training_ner_accuracy3": 1.0,
+  "training_ner_precision-overall": 0.9999999999999991,
+  "training_ner_recall-overall": 0.9999999999999991,
+  "training_ner_f1-measure-overall": 0.9999999999999492,
+  "training_loss": 7.62939453125e-06,
+  "training_worker_0_memory_MB": 4608.48828125,
+  "training_gpu_0_memory_MB": 4295.302734375,
+  "validation_ner_accuracy": 0.9371293001186239,
+  "validation_ner_accuracy3": 0.9572953736654805,
+  "validation_ner_precision-overall": 0.7948717948717942,
+  "validation_ner_recall-overall": 0.7815126050420161,
+  "validation_ner_f1-measure-overall": 0.7881355932202884,
+  "validation_loss": 234.9528350830078,
+  "best_validation_ner_accuracy": 0.9442467378410438,
+  "best_validation_ner_accuracy3": 0.9584816132858838,
+  "best_validation_ner_precision-overall": 0.822033898305084,
+  "best_validation_ner_recall-overall": 0.8151260504201674,
+  "best_validation_ner_f1-measure-overall": 0.8185654008438312,
+  "best_validation_loss": 242.68551635742188
+}

vocabulary/labels.txt

@@ -0,0 +1,7 @@
+O
+I-LOC
+I-ORG
+I-PER
+B-ORG
+B-LOC
+B-PER

vocabulary/non_padded_namespaces.txt

@@ -0,0 +1,2 @@
+*tags
+*labels