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: 0.8086
+- Loss: 5.5149
 
 ## Model description
 
@@ -48,14 +48,14 @@ The following hyperparameters were used during training:
 
 | Training Loss | Epoch | Step | Validation Loss |
 |:-------------:|:-----:|:----:|:---------------:|
-| No log        | 1.0   | 2    | 0.8897          |
-| No log        | 2.0   | 4    | 0.8323          |
-| No log        | 3.0   | 6    | 0.8086          |
+| No log        | 1.0   | 2    | 7.6513          |
+| No log        | 2.0   | 4    | 6.0836          |
+| No log        | 3.0   | 6    | 5.5149          |
 
 
 ### Framework versions
 
 - Transformers 4.48.1
-- Pytorch 2.5.1+cu124
+- Pytorch 2.6.0+cu124
 - Datasets 3.2.0
 - Tokenizers 0.21.0

config.json

@@ -4,9 +4,9 @@
     "T5ForConditionalGeneration"
   ],
   "classifier_dropout": 0.0,
-  "d_ff": 2048,
+  "d_ff": 3072,
   "d_kv": 64,
-  "d_model": 512,
+  "d_model": 768,
   "decoder_start_token_id": 0,
   "dense_act_fn": "relu",
   "dropout_rate": 0.1,
@@ -18,9 +18,9 @@
   "layer_norm_epsilon": 1e-06,
   "model_type": "t5",
   "n_positions": 512,
-  "num_decoder_layers": 6,
-  "num_heads": 8,
-  "num_layers": 6,
+  "num_decoder_layers": 12,
+  "num_heads": 12,
+  "num_layers": 12,
   "output_past": true,
   "pad_token_id": 0,
   "relative_attention_max_distance": 128,

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:be3cae2d78dbf43116265e7183fdcde2475c43c261c46c223a2d294866a90cb8
-size 242041896
+oid sha256:0cd67950a31bd1d566516ed8a853137d022663df0d24cd6a37ac631cc021fafc
+size 891644712

tokenizer_config.json

@@ -928,12 +928,12 @@
     "<extra_id_98>",
     "<extra_id_99>"
   ],
-  "clean_up_tokenization_spaces": true,
+  "clean_up_tokenization_spaces": false,
   "eos_token": "</s>",
   "extra_ids": 100,
   "extra_special_tokens": {},
   "max_length": 128,
-  "model_max_length": 512,
+  "model_max_length": 1000000000000000019884624838656,
   "pad_to_multiple_of": null,
   "pad_token": "<pad>",
   "pad_token_type_id": 0,

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:028617996871b47c558b1d577b112de6829e3c2691c996f4c9f3738075a78b93
-size 5496
+oid sha256:71ebf4f712e783eac697e2e264d7886d225280dbca954c5fcfdba0db8245e236
+size 5560