model update

README.md

@@ -18,31 +18,31 @@ model-index:
     metrics:
     - name: F1
       type: f1
-      value: 0.6655896607431341
+      value: 0.6988727858293075
     - name: Precision
       type: precision
-      value: 0.6843853820598007
+      value: 0.7161716171617162
     - name: Recall
       type: recall
-      value: 0.6477987421383647
+      value: 0.6823899371069182
     - name: F1 (macro)
       type: f1_macro
-      value: 0.41006873243715347
+      value: 0.45636958249281745
     - name: Precision (macro)
       type: precision_macro
-      value: 0.41205729487653564
+      value: 0.4519134760270864
     - name: Recall (macro)
       type: recall_macro
-      value: 0.41564916564916565
+      value: 0.4705942205942206
     - name: F1 (entity span)
       type: f1_entity_span
-      value: 0.6883116883116883
+      value: 0.7087378640776698
     - name: Precision (entity span)
       type: precision_entity_span
-      value: 0.7043189368770764
+      value: 0.7227722772277227
     - name: Recall (entity span)
       type: recall_entity_span
-      value: 0.6730158730158731
+      value: 0.6952380952380952
 
 pipeline_tag: token-classification
 widget:
@@ -55,26 +55,26 @@ This model is a fine-tuned version of [roberta-large](https://huggingface.co/rob
 [tner/fin](https://huggingface.co/datasets/tner/fin) dataset.
 Model fine-tuning is done via [T-NER](https://github.com/asahi417/tner)'s hyper-parameter search (see the repository
 for more detail). It achieves the following results on the test set:
-- F1 (micro): 0.6655896607431341
-- Precision (micro): 0.6843853820598007
-- Recall (micro): 0.6477987421383647
-- F1 (macro): 0.41006873243715347
-- Precision (macro): 0.41205729487653564
-- Recall (macro): 0.41564916564916565
+- F1 (micro): 0.6988727858293075
+- Precision (micro): 0.7161716171617162
+- Recall (micro): 0.6823899371069182
+- F1 (macro): 0.45636958249281745
+- Precision (macro): 0.4519134760270864
+- Recall (macro): 0.4705942205942206
 
 The per-entity breakdown of the F1 score on the test set are below:
-- LOC: nan
-- MISC: nan
-- ORG: nan
-- PER: nan 
+- location: 0.5121951219512196
+- organization: 0.49624060150375937
+- other: 0.0
+- person: 0.8170426065162907 
 
 For F1 scores, the confidence interval is obtained by bootstrap as below:
 - F1 (micro): 
-    - 90%: [0.5896118118382531, 0.7350473550473551]
-    - 95%: [0.5793739107766132, 0.7500251004016066] 
+    - 90%: [0.6355508274231678, 0.7613829748047737]
+    - 95%: [0.624150263185174, 0.7724430709173716] 
 - F1 (macro): 
-    - 90%: [0.5896118118382531, 0.7350473550473551]
-    - 95%: [0.5793739107766132, 0.7500251004016066] 
+    - 90%: [0.6355508274231678, 0.7613829748047737]
+    - 95%: [0.624150263185174, 0.7724430709173716] 
 
 Full evaluation can be found at [metric file of NER](https://huggingface.co/tner/roberta-large-fin/raw/main/eval/metric.json) 
 and [metric file of entity span](https://huggingface.co/tner/roberta-large-fin/raw/main/eval/metric_span.json).

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "tner_ckpt/fin_roberta_large/best_model",
+  "_name_or_path": "tner_ckpt/fin_roberta_large/model_rcsnba/epoch_5",
   "architectures": [
     "RobertaForTokenClassification"
   ],

eval/metric.json

@@ -1 +1 @@
-{"micro/f1": 0.6655896607431341, "micro/f1_ci": {"90": [0.5896118118382531, 0.7350473550473551], "95": [0.5793739107766132, 0.7500251004016066]}, "micro/recall": 0.6477987421383647, "micro/precision": 0.6843853820598007, "macro/f1": 0.41006873243715347, "macro/f1_ci": {"90": [0.3622851809142188, 0.47011199188849295], "95": [0.3559337045826972, 0.48143136272532056]}, "macro/recall": 0.41564916564916565, "macro/precision": 0.41205729487653564, "per_entity_metric": {"LOC": {"f1": NaN, "f1_ci": {"90": [NaN, NaN], "95": [NaN, NaN]}, "precision": 0.0, "recall": 0.0}, "MISC": {"f1": NaN, "f1_ci": {"90": [NaN, NaN], "95": [NaN, NaN]}, "precision": 0.0, "recall": 0.0}, "ORG": {"f1": NaN, "f1_ci": {"90": [NaN, NaN], "95": [NaN, NaN]}, "precision": 0.0, "recall": 0.0}, "PER": {"f1": NaN, "f1_ci": {"90": [NaN, NaN], "95": [NaN, NaN]}, "precision": 0.0, "recall": 0.0}}}
+{"micro/f1": 0.6988727858293075, "micro/f1_ci": {"90": [0.6355508274231678, 0.7613829748047737], "95": [0.624150263185174, 0.7724430709173716]}, "micro/recall": 0.6823899371069182, "micro/precision": 0.7161716171617162, "macro/f1": 0.45636958249281745, "macro/f1_ci": {"90": [0.41305101617635914, 0.5074221171791465], "95": [0.4040123551318039, 0.5160178907804478]}, "macro/recall": 0.4705942205942206, "macro/precision": 0.4519134760270864, "per_entity_metric": {"location": {"f1": 0.5121951219512196, "f1_ci": {"90": [0.3933107216883362, 0.6522182786157941], "95": [0.36663461538461534, 0.6849957191780824]}, "precision": 0.4883720930232558, "recall": 0.5384615384615384}, "organization": {"f1": 0.49624060150375937, "f1_ci": {"90": [0.38706011730205275, 0.6047002947920078], "95": [0.3694267515923566, 0.6220274390243905]}, "precision": 0.42857142857142855, "recall": 0.5892857142857143}, "other": {"f1": 0.0, "f1_ci": {"90": [NaN, NaN], "95": [NaN, NaN]}, "precision": 0.0, "recall": 0.0}, "person": {"f1": 0.8170426065162907, "f1_ci": {"90": [0.7555181623931624, 0.8732394366197184], "95": [0.7435141509433961, 0.8834370718923105]}, "precision": 0.8907103825136612, "recall": 0.7546296296296297}}}

eval/metric_span.json

@@ -1 +1 @@
-{"micro/f1": 0.6883116883116883, "micro/f1_ci": {"90": [0.6137984272716044, 0.757765305655086], "95": [0.604156373368873, 0.7718631178707224]}, "micro/recall": 0.6730158730158731, "micro/precision": 0.7043189368770764, "macro/f1": 0.6883116883116883, "macro/f1_ci": {"90": [0.6137984272716044, 0.757765305655086], "95": [0.604156373368873, 0.7718631178707224]}, "macro/recall": 0.6730158730158731, "macro/precision": 0.7043189368770764}
+{"micro/f1": 0.7087378640776698, "micro/f1_ci": {"90": [0.6446955883397667, 0.7724148983200707], "95": [0.6329228885677585, 0.782443539886519]}, "micro/recall": 0.6952380952380952, "micro/precision": 0.7227722772277227, "macro/f1": 0.7087378640776698, "macro/f1_ci": {"90": [0.6446955883397667, 0.7724148983200707], "95": [0.6329228885677585, 0.782443539886519]}, "macro/recall": 0.6952380952380952, "macro/precision": 0.7227722772277227}

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:79c62875ca79122bbf02bfb9383a23c97f89fe594505b5e909184cdfb708f2eb
-size 1417408561
+oid sha256:38a49e6a47f4e1b56c57c12d907d7a710d67b179331f0490b6e6a0ec6b31d83e
+size 1417414001

tokenizer_config.json

@@ -6,7 +6,7 @@
   "errors": "replace",
   "mask_token": "<mask>",
   "model_max_length": 512,
-  "name_or_path": "tner_ckpt/fin_roberta_large/best_model",
+  "name_or_path": "tner_ckpt/fin_roberta_large/model_rcsnba/epoch_5",
   "pad_token": "<pad>",
   "sep_token": "</s>",
   "special_tokens_map_file": "tner_ckpt/fin_roberta_large/model_rcsnba/epoch_5/special_tokens_map.json",