End of training

README.md

@@ -12,6 +12,11 @@ should probably proofread and complete it, then remove this comment. -->
 # sinhala-roman-transformer
 
 This model is a fine-tuned version of [](https://huggingface.co/) on an unknown dataset.
+It achieves the following results on the evaluation set:
+- Loss: 0.0029
+- Rouge2 Precision: 0.0
+- Rouge2 Recall: 0.0
+- Rouge2 Fmeasure: 0.0
 
 ## Model description
 
@@ -37,16 +42,23 @@ The following hyperparameters were used during training:
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: linear
 - lr_scheduler_warmup_steps: 2000
-- num_epochs: 3.0
+- training_steps: 40000
 - mixed_precision_training: Native AMP
 
 ### Training results
 
+| Training Loss | Epoch   | Step  | Validation Loss | Rouge2 Precision | Rouge2 Recall | Rouge2 Fmeasure |
+|:-------------:|:-------:|:-----:|:---------------:|:----------------:|:-------------:|:---------------:|
+| 0.006         | 5.1440  | 7500  | 0.0052          | 0.0              | 0.0           | 0.0             |
+| 0.0002        | 10.2881 | 15000 | 0.0034          | 0.0              | 0.0           | 0.0             |
+| 0.0           | 15.4321 | 22500 | 0.0030          | 0.0              | 0.0           | 0.0             |
+| 0.0           | 20.5761 | 30000 | 0.0030          | 0.0              | 0.0           | 0.0             |
+| 0.0           | 25.7202 | 37500 | 0.0029          | 0.0              | 0.0           | 0.0             |
 
 
 ### Framework versions
 
 - Transformers 4.41.2
-- Pytorch 2.3.0+cu121
-- Datasets 2.20.0
+- Pytorch 2.1.2
+- Datasets 2.19.2
 - Tokenizers 0.19.1

generation_config.json

@@ -3,8 +3,6 @@
   "early_stopping": true,
   "eos_token_id": 3,
   "length_penalty": 2.0,
-  "max_length": 142,
-  "min_length": 56,
   "no_repeat_ngram_size": 3,
   "num_beams": 4,
   "pad_token_id": 0,