djsull/aha_class_v1

README.md

@@ -1,5 +1,5 @@
 ---
-base_model: klue/roberta-base
+base_model: monologg/kobigbird-bert-base
 tags:
 - generated_from_trainer
 metrics:
@@ -15,12 +15,12 @@ should probably proofread and complete it, then remove this comment. -->
 
 # aha_class_v1
 
-This model is a fine-tuned version of [klue/roberta-base](https://huggingface.co/klue/roberta-base) on an unknown dataset.
+This model is a fine-tuned version of [monologg/kobigbird-bert-base](https://huggingface.co/monologg/kobigbird-bert-base) on an unknown dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.0955
-- F1: 0.9580
-- Roc Auc: 0.9679
-- Accuracy: 0.9304
+- Loss: 0.0796
+- F1: 0.9623
+- Roc Auc: 0.9721
+- Accuracy: 0.9478
 
 ## Model description
 
@@ -51,16 +51,19 @@ The following hyperparameters were used during training:
 
 | Training Loss | Epoch | Step | Validation Loss | F1     | Roc Auc | Accuracy |
 |:-------------:|:-----:|:----:|:---------------:|:------:|:-------:|:--------:|
-| No log        | 1.0   | 42   | 0.1536          | 0.9496 | 0.9615  | 0.9130   |
-| No log        | 2.0   | 84   | 0.1121          | 0.9540 | 0.9657  | 0.9217   |
-| No log        | 3.0   | 126  | 0.1088          | 0.9617 | 0.9682  | 0.9478   |
-| No log        | 4.0   | 168  | 0.1023          | 0.9580 | 0.9679  | 0.9391   |
-| No log        | 5.0   | 210  | 0.0955          | 0.9580 | 0.9679  | 0.9304   |
-| No log        | 6.0   | 252  | 0.0979          | 0.9576 | 0.9659  | 0.9391   |
-| No log        | 7.0   | 294  | 0.1009          | 0.9627 | 0.9741  | 0.9391   |
-| No log        | 8.0   | 336  | 0.0972          | 0.9580 | 0.9679  | 0.9304   |
-| No log        | 9.0   | 378  | 0.1148          | 0.95   | 0.9635  | 0.9217   |
-| No log        | 10.0  | 420  | 0.1093          | 0.9544 | 0.9677  | 0.9217   |
+| No log        | 1.0   | 42   | 0.3116          | 0.9256 | 0.9463  | 0.8870   |
+| No log        | 2.0   | 84   | 0.1718          | 0.9746 | 0.9788  | 0.9565   |
+| No log        | 3.0   | 126  | 0.1375          | 0.9576 | 0.9659  | 0.9391   |
+| No log        | 4.0   | 168  | 0.1007          | 0.9702 | 0.9746  | 0.9565   |
+| No log        | 5.0   | 210  | 0.0927          | 0.9617 | 0.9682  | 0.9478   |
+| No log        | 6.0   | 252  | 0.0927          | 0.9617 | 0.9682  | 0.9478   |
+| No log        | 7.0   | 294  | 0.0910          | 0.9617 | 0.9682  | 0.9478   |
+| No log        | 8.0   | 336  | 0.0796          | 0.9623 | 0.9721  | 0.9478   |
+| No log        | 9.0   | 378  | 0.0953          | 0.9580 | 0.9679  | 0.9391   |
+| No log        | 10.0  | 420  | 0.0796          | 0.9661 | 0.9724  | 0.9565   |
+| No log        | 11.0  | 462  | 0.0890          | 0.9580 | 0.9679  | 0.9391   |
+| 0.1161        | 12.0  | 504  | 0.0910          | 0.9580 | 0.9679  | 0.9391   |
+| 0.1161        | 13.0  | 546  | 0.0916          | 0.9580 | 0.9679  | 0.9391   |
 
 
 ### Framework versions

config.json

@@ -1,14 +1,16 @@
 {
-  "_name_or_path": "klue/roberta-base",
+  "_name_or_path": "monologg/kobigbird-bert-base",
   "architectures": [
-    "RobertaForSequenceClassification"
+    "BigBirdForSequenceClassification"
   ],
   "attention_probs_dropout_prob": 0.1,
-  "bos_token_id": 0,
+  "attention_type": "block_sparse",
+  "block_size": 64,
+  "bos_token_id": 5,
   "classifier_dropout": null,
-  "eos_token_id": 2,
+  "eos_token_id": 6,
   "gradient_checkpointing": false,
-  "hidden_act": "gelu",
+  "hidden_act": "gelu_new",
   "hidden_dropout_prob": 0.1,
   "hidden_size": 768,
   "id2label": {
@@ -23,18 +25,22 @@
     "Service": 1,
     "Token": 2
   },
-  "layer_norm_eps": 1e-05,
-  "max_position_embeddings": 514,
-  "model_type": "roberta",
+  "layer_norm_eps": 1e-12,
+  "max_position_embeddings": 4096,
+  "model_type": "big_bird",
   "num_attention_heads": 12,
   "num_hidden_layers": 12,
-  "pad_token_id": 1,
+  "num_random_blocks": 3,
+  "pad_token_id": 0,
   "position_embedding_type": "absolute",
   "problem_type": "multi_label_classification",
+  "rescale_embeddings": false,
+  "sep_token_id": 3,
   "tokenizer_class": "BertTokenizer",
   "torch_dtype": "float32",
   "transformers_version": "4.42.3",
-  "type_vocab_size": 1,
+  "type_vocab_size": 2,
+  "use_bias": true,
   "use_cache": true,
-  "vocab_size": 32000
+  "vocab_size": 32500
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5df0a87f84d5e7742762474048aa5bcdde3509b6bbeb0bedeb79ff982d7b9339
-size 442505820
+oid sha256:3f7544ab951561caa653b97aa5d56430ae1f2c4b02f9fdb24fba7b62542ff90f
+size 457410764

runs/Sep27_08-46-43_129-146-79-76/events.out.tfevents.1727426804.129-146-79-76.1524859.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:54e2374ae50fc2a5c58a6043cd885fee13c05722aa1ab9da2dc8ed8b0d57ff77
+size 11308

runs/Sep27_08-46-43_129-146-79-76/events.out.tfevents.1727427293.129-146-79-76.1524859.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1b9dd88b902c2d1cae14fb17faf22c851c0e2d6e7c6bf659fed291fb31185870
+size 508

special_tokens_map.json

@@ -1,6 +1,6 @@
 {
   "bos_token": {
-    "content": "[CLS]",
+    "content": "<s>",
     "lstrip": false,
     "normalized": false,
     "rstrip": false,
@@ -14,7 +14,7 @@
     "single_word": false
   },
   "eos_token": {
-    "content": "[SEP]",
+    "content": "</s>",
     "lstrip": false,
     "normalized": false,
     "rstrip": false,

tokenizer_config.json

@@ -1,7 +1,7 @@
 {
   "added_tokens_decoder": {
     "0": {
-      "content": "[CLS]",
+      "content": "[PAD]",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
@@ -9,7 +9,7 @@
       "special": true
     },
     "1": {
-      "content": "[PAD]",
+      "content": "[UNK]",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
@@ -17,7 +17,7 @@
       "special": true
     },
     "2": {
-      "content": "[SEP]",
+      "content": "[CLS]",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
@@ -25,7 +25,7 @@
       "special": true
     },
     "3": {
-      "content": "[UNK]",
+      "content": "[SEP]",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
@@ -39,16 +39,32 @@
       "rstrip": false,
       "single_word": false,
       "special": true
+    },
+    "5": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "6": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
     }
   },
-  "bos_token": "[CLS]",
+  "bos_token": "<s>",
   "clean_up_tokenization_spaces": true,
   "cls_token": "[CLS]",
   "do_basic_tokenize": true,
   "do_lower_case": false,
-  "eos_token": "[SEP]",
+  "eos_token": "</s>",
   "mask_token": "[MASK]",
-  "model_max_length": 512,
+  "model_max_length": 4096,
   "never_split": null,
   "pad_token": "[PAD]",
   "sep_token": "[SEP]",

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c66cc7763e4f98b80a1d0db799d0f5b04c0415d70aa4048211748aec2c06f3da
+oid sha256:925348b829f5ec55e7d35fd3744a5b98a5af4861be2064125dfdc907294131ba
 size 5112