End of training

README.md

@@ -21,16 +21,16 @@ model-index:
     metrics:
     - name: Precision
       type: precision
-      value: 0.9474485910129474
+      value: 0.9220389805097451
     - name: Recall
       type: recall
-      value: 0.9658385093167702
+      value: 0.9549689440993789
     - name: F1
       type: f1
-      value: 0.9565551710880431
+      value: 0.9382151029748284
     - name: Accuracy
       type: accuracy
-      value: 0.9613752122241087
+      value: 0.934634974533107
 ---
 
 <!-- This model card has been generated automatically according to the information the Trainer had access to. You
@@ -40,11 +40,11 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model is a fine-tuned version of [layoutlmv3](https://huggingface.co/layoutlmv3) on the mp-02/cord dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.2047
-- Precision: 0.9474
-- Recall: 0.9658
-- F1: 0.9566
-- Accuracy: 0.9614
+- Loss: 0.3103
+- Precision: 0.9220
+- Recall: 0.9550
+- F1: 0.9382
+- Accuracy: 0.9346
 
 ## Model description
 
@@ -64,23 +64,21 @@ More information needed
 
 The following hyperparameters were used during training:
 - learning_rate: 1e-05
-- train_batch_size: 10
-- eval_batch_size: 10
+- train_batch_size: 12
+- eval_batch_size: 12
 - seed: 42
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: linear
-- training_steps: 1500
+- training_steps: 1000
 
 ### Training results
 
-| Training Loss | Epoch  | Step | Validation Loss | Precision | Recall | F1     | Accuracy |
-|:-------------:|:------:|:----:|:---------------:|:---------:|:------:|:------:|:--------:|
-| No log        | 3.125  | 250  | 0.7037          | 0.8188    | 0.8665 | 0.8419 | 0.8587   |
-| 1.0839        | 6.25   | 500  | 0.3828          | 0.8926    | 0.9293 | 0.9106 | 0.9223   |
-| 1.0839        | 9.375  | 750  | 0.2811          | 0.9371    | 0.9596 | 0.9482 | 0.9469   |
-| 0.2469        | 12.5   | 1000 | 0.2295          | 0.9401    | 0.9620 | 0.9509 | 0.9529   |
-| 0.2469        | 15.625 | 1250 | 0.2106          | 0.9460    | 0.9658 | 0.9558 | 0.9601   |
-| 0.1263        | 18.75  | 1500 | 0.2047          | 0.9474    | 0.9658 | 0.9566 | 0.9614   |
+| Training Loss | Epoch   | Step | Validation Loss | Precision | Recall | F1     | Accuracy |
+|:-------------:|:-------:|:----:|:---------------:|:---------:|:------:|:------:|:--------:|
+| No log        | 3.7313  | 250  | 0.7079          | 0.8180    | 0.8657 | 0.8412 | 0.8476   |
+| 1.064         | 7.4627  | 500  | 0.4029          | 0.8791    | 0.9255 | 0.9017 | 0.9138   |
+| 1.064         | 11.1940 | 750  | 0.3349          | 0.9132    | 0.9480 | 0.9303 | 0.9308   |
+| 0.294         | 14.9254 | 1000 | 0.3103          | 0.9220    | 0.9550 | 0.9382 | 0.9346   |
 
 
 ### Framework versions