Updates the SRL model

CHANGELOG.md

@@ -9,6 +9,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Fixed
 
+- Updated the BERT SRL model to be compatible with the new huggingface tokenizers.
 - `CopyNetSeq2Seq` model now works with pretrained transformers.
 - A bug with `NextTokenLM` that caused simple gradient interpreters to fail.
 - A bug in `training_config` of `qanet` and `bimpm` that used the old version of `regularizer` and `initializer`.

allennlp_models/pretrained.py

@@ -29,7 +29,7 @@ def _load_predictor(archive_file: str, predictor_name: Optional[str] = None) ->
 
 def bert_srl_shi_2019() -> SemanticRoleLabelerPredictor:
  predictor = _load_predictor(
- "https://storage.googleapis.com/allennlp-public-models/bert-base-srl-2020.03.24.tar.gz"
+ "https://storage.googleapis.com/allennlp-public-models/bert-base-srl-2020.07.14.tar.gz"
  )
  return predictor
 

training_config/structured_prediction/bert_base_srl.jsonnet

@@ -11,6 +11,8 @@
  }
  },
 
+ // "train_data_path": "/net/nfs.corp/allennlp/data/ontonotes/conll-formatted-ontonotes-5.0/data/train",
+ // "validation_data_path": "/net/nfs.corp/allennlp/data/ontonotes/conll-formatted-ontonotes-5.0/data/development",
  "train_data_path": std.extVar("SRL_TRAIN_DATA_PATH"),
  "validation_data_path": std.extVar("SRL_VALIDATION_DATA_PATH"),
 
@@ -33,16 +35,12 @@
 
  "learning_rate_scheduler": {
  "type": "slanted_triangular",
- "num_epochs": 15,
- "num_steps_per_epoch": 8829,
  },
  "checkpointer": {
  "num_serialized_models_to_keep": 2,
  },
  "grad_norm": 1.0,
  "num_epochs": 15,
  "validation_metric": "+f1-measure-overall",
- "cuda_device": 0,
  },
-
 }