Model save

README.md

@@ -20,21 +20,21 @@ model-index:
       name: cnec
       type: cnec
       config: default
-      split: test
+      split: validation
       args: default
     metrics:
     - name: Precision
       type: precision
-      value: 0.8282633808240277
+      value: 0.8366164542294322
     - name: Recall
       type: recall
-      value: 0.8837304847986853
+      value: 0.8946716232961586
     - name: F1
       type: f1
-      value: 0.8550983899821109
+      value: 0.8646706586826347
     - name: Accuracy
       type: accuracy
-      value: 0.9664021317268146
+      value: 0.9697812545270172
 ---
 
 <!-- This model card has been generated automatically according to the information the Trainer had access to. You
@@ -44,11 +44,11 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model is a fine-tuned version of [FacebookAI/xlm-roberta-large](https://huggingface.co/FacebookAI/xlm-roberta-large) on the cnec dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.1865
-- Precision: 0.8283
-- Recall: 0.8837
-- F1: 0.8551
-- Accuracy: 0.9664
+- Loss: 0.1777
+- Precision: 0.8366
+- Recall: 0.8947
+- F1: 0.8647
+- Accuracy: 0.9698
 
 ## Model description
 
@@ -68,26 +68,20 @@ More information needed
 
 The following hyperparameters were used during training:
 - learning_rate: 2e-05
-- train_batch_size: 32
-- eval_batch_size: 32
+- train_batch_size: 1
+- eval_batch_size: 1
 - seed: 42
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: linear
-- lr_scheduler_warmup_ratio: 0.1
-- lr_scheduler_warmup_steps: 500
-- num_epochs: 16
+- num_epochs: 3
 
 ### Training results
 
-| Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1     | Accuracy |
-|:-------------:|:-----:|:----:|:---------------:|:---------:|:------:|:------:|:--------:|
-| 0.6852        | 2.22  | 500  | 0.1614          | 0.7278    | 0.8250 | 0.7733 | 0.9574   |
-| 0.1311        | 4.44  | 1000 | 0.1716          | 0.7690    | 0.8591 | 0.8116 | 0.9596   |
-| 0.0882        | 6.67  | 1500 | 0.1785          | 0.7616    | 0.8714 | 0.8128 | 0.9613   |
-| 0.062         | 8.89  | 2000 | 0.1536          | 0.8212    | 0.8928 | 0.8555 | 0.9669   |
-| 0.0457        | 11.11 | 2500 | 0.1783          | 0.8204    | 0.8673 | 0.8432 | 0.9645   |
-| 0.0353        | 13.33 | 3000 | 0.1829          | 0.8259    | 0.8809 | 0.8525 | 0.9655   |
-| 0.0289        | 15.56 | 3500 | 0.1865          | 0.8283    | 0.8837 | 0.8551 | 0.9664   |
+| Training Loss | Epoch | Step  | Validation Loss | Precision | Recall | F1     | Accuracy |
+|:-------------:|:-----:|:-----:|:---------------:|:---------:|:------:|:------:|:--------:|
+| 0.1989        | 1.0   | 7193  | 0.1998          | 0.7372    | 0.8249 | 0.7786 | 0.9575   |
+| 0.1398        | 2.0   | 14386 | 0.1890          | 0.8257    | 0.8827 | 0.8533 | 0.9662   |
+| 0.0926        | 3.0   | 21579 | 0.1777          | 0.8366    | 0.8947 | 0.8647 | 0.9698   |
 
 
 ### Framework versions