johannawawi/v5_balanced_dataset_fine-tuning-java-indo-sentiment-analysist-3-class

README.md

@@ -18,14 +18,14 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model is a fine-tuned version of [w11wo/indonesian-roberta-base-sentiment-classifier](https://huggingface.co/w11wo/indonesian-roberta-base-sentiment-classifier) on the None dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.6295
-- Accuracy: 0.8273
-- F1 Macro: 0.8264
-- F1 Weighted: 0.8266
-- Precision Macro: 0.8267
-- Recall Macro: 0.8270
-- Precision Weighted: 0.8268
-- Recall Weighted: 0.8273
+- Loss: 1.0692
+- Accuracy: 0.8436
+- F1 Macro: 0.8431
+- F1 Weighted: 0.8433
+- Precision Macro: 0.8432
+- Recall Macro: 0.8434
+- Precision Weighted: 0.8433
+- Recall Weighted: 0.8436
 
 ## Model description
 
@@ -44,20 +44,24 @@ More information needed
 ### Training hyperparameters
 
 The following hyperparameters were used during training:
-- learning_rate: 8.772683881250156e-06
-- train_batch_size: 16
+- learning_rate: 8.879626978799419e-06
+- train_batch_size: 8
 - eval_batch_size: 8
 - seed: 42
 - optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
 - lr_scheduler_type: cosine
 - lr_scheduler_warmup_ratio: 0.1
-- num_epochs: 7
+- num_epochs: 10
 
 ### Training results
 
 | Training Loss | Epoch  | Step | Validation Loss | Accuracy | F1 Macro | F1 Weighted | Precision Macro | Recall Macro | Precision Weighted | Recall Weighted |
 |:-------------:|:------:|:----:|:---------------:|:--------:|:--------:|:-----------:|:---------------:|:------------:|:------------------:|:---------------:|
-| 0.2188        | 3.6232 | 500  | 0.6295          | 0.8273   | 0.8264   | 0.8266      | 0.8267          | 0.8270       | 0.8268             | 0.8273          |
+| 0.035         | 1.8182 | 500  | 1.0247          | 0.8327   | 0.8321   | 0.8323      | 0.8325          | 0.8325       | 0.8325             | 0.8327          |
+| 0.0829        | 3.6364 | 1000 | 1.0134          | 0.8273   | 0.8262   | 0.8263      | 0.8275          | 0.8270       | 0.8275             | 0.8273          |
+| 0.1858        | 5.4545 | 1500 | 1.0692          | 0.8436   | 0.8431   | 0.8433      | 0.8432          | 0.8434       | 0.8433             | 0.8436          |
+| 0.2844        | 7.2727 | 2000 | 0.9823          | 0.8255   | 0.8250   | 0.8251      | 0.8253          | 0.8253       | 0.8254             | 0.8255          |
+| 0.3299        | 9.0909 | 2500 | 0.9626          | 0.8255   | 0.8251   | 0.8252      | 0.8253          | 0.8253       | 0.8254             | 0.8255          |
 
 
 ### Framework versions

config.json

@@ -4,7 +4,7 @@
   ],
   "attention_probs_dropout_prob": 0.1,
   "bos_token_id": 0,
-  "classifier_dropout": 0.2,
+  "classifier_dropout": 0.3,
   "eos_token_id": 2,
   "gradient_checkpointing": false,
   "hidden_act": "gelu",

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cfe01fe7a6497d193659b143f7ca2375c4526cfc482707467bf2ac909cdf8dd0
+oid sha256:8200e6c7302793057b606fc7b96f8cd2a42cfd3a795aeb1bd5b9acaa6d5ad655
 size 498615900

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3beb66c94dd28553d61565b003034c7450f6cc32e2e5d50c6d5a024e713572b9
+oid sha256:639b0a4216fe618d1906bafbc0798346ad392a872b5061f36c1cdccc20506ff7
 size 5304