Upload model

config.json

@@ -7,7 +7,7 @@
   "attention_probs_dropout_prob": 0.1,
   "augmentation": "unk",
   "auto_map": {
-    "AutoConfig": "modeling_bionexttagger.BioNextTaggerConfig",
+    "AutoConfig": "configuration_bionexttager.BioNextTaggerConfig",
     "AutoModel": "modeling_bionexttagger.BioNextTaggerModel"
   },
   "classifier_dropout": null,
@@ -64,6 +64,6 @@
   "transformers_version": "4.37.2",
   "type_vocab_size": 2,
   "use_cache": true,
-  "version": "0.1.1",
+  "version": "0.1.2",
   "vocab_size": 28895
 }

configuration_bionexttager.py

@@ -13,7 +13,7 @@ class BioNextTaggerConfig(PretrainedConfig):
         percentage_tags = 0.2,
         p_augmentation = 0.5,
         crf_reduction = "mean",
-        version="0.1.1",
+        version="0.1.2",
         **kwargs,
     ):
         self.version = version

modeling_bionexttagger.py

@@ -1,7 +1,7 @@
 
 import os
 from typing import Optional, Union
-from transformers import AutoModel, PreTrainedModel, AutoConfig, BertModel, PretrainedConfig
+from transformers import AutoModel, PreTrainedModel, AutoConfig, BertModel
 from transformers.modeling_outputs import  TokenClassifierOutput
 from torch import nn
 from torch.nn import CrossEntropyLoss
@@ -10,35 +10,8 @@ from typing import List, Optional
 
 import torch
 from itertools import islice
+from .configuration_bionexttager import BioNextTaggerConfig
 
-class BioNextTaggerConfig(PretrainedConfig):
-    model_type = "crf-tagger"
-
-    def __init__(
-        self,
-        augmentation = "unk",
-        context_size = 64,
-        percentage_tags = 0.2,
-        p_augmentation = 0.5,
-        crf_reduction = "mean",
-        version="0.1.1",
-        **kwargs,
-    ):
-        self.version = version
-        self.augmentation = augmentation
-        self.context_size = context_size
-        self.percentage_tags = percentage_tags
-        self.p_augmentation = p_augmentation
-        self.crf_reduction = crf_reduction
-        super().__init__(**kwargs)
-        
-    def get_backbonemodel_config(self):
-        backbonemodel_cfg = AutoConfig.from_pretrained(self._name_or_path)#.to_dict()
-        for k in backbonemodel_cfg.to_dict():
-            if hasattr(self, k):
-                setattr(backbonemodel_cfg,k, getattr(self,k))
-                
-        return backbonemodel_cfg
 
 NUM_PER_LAYER = 16
 
@@ -49,13 +22,14 @@ class BioNextTaggerModel(PreTrainedModel):
     def __init__(self, config):
         super().__init__(config)
         self.num_labels = config.num_labels
-        #print("LOAD BERT?")
+        print("LOAD BERT?", flush=True)
         self.bert = BertModel(config.get_backbonemodel_config(), add_pooling_layer=False)
                     #AutoModel.from_pretrained(config._name_or_path, 
                     #                          config=config.get_backbonemodel_config(), 
                     #                          add_pooling_layer=False)
+        print("LOADED BERT", flush=True)
         # self.vocab_size = config.vocab_size
-        classifier_dropout = (config.classifier_dropout if config.classifier_dropout is not None else config.hidden_dropout_prob)
+        #classifier_dropout = (config.classifier_dropout if config.classifier_dropout is not None else config.hidden_dropout_prob)
         self.dropout = nn.Dropout(config.hidden_dropout_prob)
         self.dense = nn.Linear(config.hidden_size, config.hidden_size)
         self.dense_activation = nn.GELU(approximate='none')