Model save

README.md

@@ -1,9 +1,6 @@
 ---
-license: apache-2.0
-base_model: google/vit-base-patch16-224-in21k
+base_model: microsoft/dit-base-finetuned-rvlcdip
 tags:
-- image-classification
-- vision
 - generated_from_trainer
 metrics:
 - f1
@@ -17,10 +14,10 @@ should probably proofread and complete it, then remove this comment. -->
 
 # donut-base-beans
 
-This model is a fine-tuned version of [google/vit-base-patch16-224-in21k](https://huggingface.co/google/vit-base-patch16-224-in21k) on an unknown dataset.
+This model is a fine-tuned version of [microsoft/dit-base-finetuned-rvlcdip](https://huggingface.co/microsoft/dit-base-finetuned-rvlcdip) on an unknown dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.0410
-- F1: 0.9461
+- F1: 0.9705
+- Loss: 0.0391
 
 ## Model description
 
@@ -51,11 +48,11 @@ The following hyperparameters were used during training:
 
 | Training Loss | Epoch | Step | F1     | Validation Loss |
 |:-------------:|:-----:|:----:|:------:|:---------------:|
-| 0.0663        | 1.0   | 1981 | 0.9705 | 0.0594          |
-| 0.0747        | 2.0   | 3962 | 0.9705 | 0.0410          |
-| 0.0676        | 3.0   | 5943 | 0.9705 | 0.0463          |
-| 0.0706        | 4.0   | 7924 | 0.9705 | 0.0445          |
-| 0.0595        | 5.0   | 9905 | 0.0460 | 0.9429          |
+| 0.0594        | 1.0   | 1981 | 0.9705 | 0.0440          |
+| 0.0678        | 2.0   | 3962 | 0.9705 | 0.0447          |
+| 0.0615        | 3.0   | 5943 | 0.9705 | 0.0416          |
+| 0.0685        | 4.0   | 7924 | 0.9705 | 0.0385          |
+| 0.0582        | 5.0   | 9905 | 0.9705 | 0.0391          |
 
 
 ### Framework versions

all_results.json

@@ -1,13 +1,13 @@
 {
     "epoch": 5.0,
-    "eval_f1": 0.9460782887783499,
-    "eval_loss": 0.04101279005408287,
-    "eval_runtime": 559.6733,
-    "eval_samples_per_second": 368.54,
-    "eval_steps_per_second": 2.88,
-    "total_flos": 9.820471825285631e+19,
-    "train_loss": 0.011631221487809769,
-    "train_runtime": 2940.6816,
-    "train_samples_per_second": 430.955,
-    "train_steps_per_second": 3.368
+    "eval_f1": 0.9557601497125016,
+    "eval_loss": 0.038467586040496826,
+    "eval_runtime": 795.0811,
+    "eval_samples_per_second": 259.423,
+    "eval_steps_per_second": 2.027,
+    "total_flos": 9.82152667464321e+19,
+    "train_loss": 0.0,
+    "train_runtime": 0.074,
+    "train_samples_per_second": 17128131.172,
+    "train_steps_per_second": 133870.543
 }

config.json

@@ -1,10 +1,15 @@
 {
-  "_name_or_path": "google/vit-base-patch16-224-in21k",
+  "_name_or_path": "microsoft/dit-base-finetuned-rvlcdip",
+  "add_fpn": false,
   "architectures": [
-    "ViTForImageClassification"
+    "BeitForImageClassification"
   ],
   "attention_probs_dropout_prob": 0.0,
-  "encoder_stride": 16,
+  "auxiliary_channels": 256,
+  "auxiliary_concat_input": false,
+  "auxiliary_loss_weight": 0.4,
+  "auxiliary_num_convs": 1,
+  "drop_path_rate": 0.1,
   "finetuning_task": "image-classification",
   "hidden_act": "gelu",
   "hidden_dropout_prob": 0.0,
@@ -19,13 +24,54 @@
     "Value(dtype='int64', id=None)": "0"
   },
   "layer_norm_eps": 1e-12,
-  "model_type": "vit",
+  "layer_scale_init_value": 0.1,
+  "model_type": "beit",
   "num_attention_heads": 12,
   "num_channels": 3,
   "num_hidden_layers": 12,
+  "out_features": [
+    "stage3",
+    "stage5",
+    "stage7",
+    "stage11"
+  ],
+  "out_indices": [
+    3,
+    5,
+    7,
+    11
+  ],
   "patch_size": 16,
-  "problem_type": "regression",
-  "qkv_bias": true,
+  "pool_scales": [
+    1,
+    2,
+    3,
+    6
+  ],
+  "reshape_hidden_states": true,
+  "semantic_loss_ignore_index": 255,
+  "stage_names": [
+    "stem",
+    "stage1",
+    "stage2",
+    "stage3",
+    "stage4",
+    "stage5",
+    "stage6",
+    "stage7",
+    "stage8",
+    "stage9",
+    "stage10",
+    "stage11",
+    "stage12"
+  ],
   "torch_dtype": "float32",
-  "transformers_version": "4.43.0.dev0"
+  "transformers_version": "4.43.0.dev0",
+  "use_absolute_position_embeddings": true,
+  "use_auxiliary_head": true,
+  "use_mask_token": false,
+  "use_mean_pooling": true,
+  "use_relative_position_bias": false,
+  "use_shared_relative_position_bias": false,
+  "vocab_size": 8192
 }

eval_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 5.0,
-    "eval_f1": 0.9460782887783499,
-    "eval_loss": 0.04101279005408287,
-    "eval_runtime": 559.6733,
-    "eval_samples_per_second": 368.54,
-    "eval_steps_per_second": 2.88
+    "eval_f1": 0.9557601497125016,
+    "eval_loss": 0.038467586040496826,
+    "eval_runtime": 795.0811,
+    "eval_samples_per_second": 259.423,
+    "eval_steps_per_second": 2.027
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:572303f761b7a7387078fed0e6d4bb406e84528aa38cd5998d45fe55f1830673
-size 343220892
+oid sha256:e5e6ad99f326c79142618188e2f2311b61803f4a73856b3c1a458bcde5366aa2
+size 343258940

preprocessor_config.json

@@ -1,5 +1,11 @@
 {
+  "crop_size": {
+    "height": 224,
+    "width": 224
+  },
+  "do_center_crop": false,
   "do_normalize": true,
+  "do_reduce_labels": false,
   "do_rescale": true,
   "do_resize": true,
   "image_mean": [
@@ -7,7 +13,7 @@
     0.5,
     0.5
   ],
-  "image_processor_type": "ViTImageProcessor",
+  "image_processor_type": "BeitImageProcessor",
   "image_std": [
     0.5,
     0.5,

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 5.0,
-    "total_flos": 9.820471825285631e+19,
-    "train_loss": 0.011631221487809769,
-    "train_runtime": 2940.6816,
-    "train_samples_per_second": 430.955,
-    "train_steps_per_second": 3.368
+    "total_flos": 9.82152667464321e+19,
+    "train_loss": 0.0,
+    "train_runtime": 0.074,
+    "train_samples_per_second": 17128131.172,
+    "train_steps_per_second": 133870.543
 }

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e9d8436778df1e6f25ce1b6ddeaa31273eac5f992183149befb2bb90f6aa8b36
-size 5240
+oid sha256:fa4d5febffce511b6da18d708f0e85b9ddaebce0c9548a11dc40ef9e37284f67
+size 5368