Fine-tuned Construction Receipt Model

README.md

@@ -16,7 +16,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model is a fine-tuned version of [DanSarm/receipt-core-model](https://huggingface.co/DanSarm/receipt-core-model) on an unknown dataset.
 It achieves the following results on the evaluation set:
-- Loss: 2.8093
+- Loss: 1.2568
 
 ## Model description
 
@@ -35,22 +35,31 @@ More information needed
 ### Training hyperparameters
 
 The following hyperparameters were used during training:
-- learning_rate: 5e-05
-- train_batch_size: 8
-- eval_batch_size: 8
+- learning_rate: 0.0003
+- train_batch_size: 16
+- eval_batch_size: 16
 - seed: 42
+- gradient_accumulation_steps: 2
+- total_train_batch_size: 32
 - optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
 - lr_scheduler_type: linear
-- num_epochs: 3
+- num_epochs: 20
 - mixed_precision_training: Native AMP
 
 ### Training results
 
 | Training Loss | Epoch | Step | Validation Loss |
 |:-------------:|:-----:|:----:|:---------------:|
-| No log        | 1.0   | 5    | 3.4726          |
-| No log        | 2.0   | 10   | 2.9864          |
-| No log        | 3.0   | 15   | 2.8093          |
+| 2.911         | 1.0   | 2    | 2.5178          |
+| 2.0298        | 2.0   | 4    | 2.0128          |
+| 1.7239        | 3.0   | 6    | 1.7704          |
+| 1.5708        | 4.0   | 8    | 1.6107          |
+| 1.4319        | 5.0   | 10   | 1.5010          |
+| 1.3463        | 6.0   | 12   | 1.4165          |
+| 1.262         | 7.0   | 14   | 1.3495          |
+| 1.2137        | 8.0   | 16   | 1.3015          |
+| 1.1761        | 9.0   | 18   | 1.2705          |
+| 1.1369        | 10.0  | 20   | 1.2568          |
 
 
 ### Framework versions

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3195bbd14f198686395ba9de9dfb4aa4f7f160f1c8f12561c9efe358f5859ae4
+oid sha256:2a419ec4234254612766ea7d422cebfc2c63ebe3c27896ee858ecacfe8ab7aca
 size 242041896

tokenizer.json

@@ -7,9 +7,7 @@
     "stride": 0
   },
   "padding": {
-    "strategy": {
-      "Fixed": 128
-    },
+    "strategy": "BatchLongest",
     "direction": "Right",
     "pad_to_multiple_of": null,
     "pad_id": 0,

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a204ff7178196e8401fcfa3e04ecea47f012bafeab99ba549a23e4782bb52d0f
-size 5496
+oid sha256:a0d80b0af9f788d93aaf056a86ceafc9ba182f5cf4fdd2d3d18137a795c34e08
+size 5432