Training in progress, epoch 1

command.txt

@@ -1,7 +1,7 @@
 python run_image_classification.py \
     --train_dir /trunk/shared/eebo_data/images_cropped/ \
     --validation_dir /trunk/shared/eebo_data/images_cropped/ \
-    --output_dir ./default_model/ \
+    --output_dir ./microsoft_dit/ \
     --remove_unused_columns False \
     --label_column_name labels \
     --do_train \
@@ -9,9 +9,9 @@ python run_image_classification.py \
     --push_to_hub \
     --push_to_hub_model_id donut-base-beans \
     --learning_rate 3e-5 \
-    --num_train_epochs 10 \
-    --per_device_train_batch_size 64 \
-    --per_device_eval_batch_size 64 \
+    --num_train_epochs 5 \
+    --per_device_train_batch_size 32 \
+    --per_device_eval_batch_size 32 \
     --logging_strategy steps \
     --logging_steps 10 \
     --eval_strategy epoch \
@@ -22,4 +22,4 @@ python run_image_classification.py \
     --ignore_mismatched_sizes \
     --image_column_name image_url \
     --label_column_name label \
-    --resume_from_checkpoint /trunk2/yufei/summer24/transformers/examples/pytorch/image-classification/default_model/checkpoint-67330
+    --model_name_or_path microsoft/dit-base-finetuned-rvlcdip

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "microsoft/beit-base-patch16-224-pt22k-ft22k",
+  "_name_or_path": "microsoft/dit-base-finetuned-rvlcdip",
   "add_fpn": false,
   "architectures": [
     "BeitForImageClassification"
@@ -30,10 +30,16 @@
   "num_channels": 3,
   "num_hidden_layers": 12,
   "out_features": [
-    "stage12"
+    "stage3",
+    "stage5",
+    "stage7",
+    "stage11"
   ],
   "out_indices": [
-    12
+    3,
+    5,
+    7,
+    11
   ],
   "patch_size": 16,
   "pool_scales": [
@@ -62,11 +68,11 @@
   ],
   "torch_dtype": "float32",
   "transformers_version": "4.43.0.dev0",
-  "use_absolute_position_embeddings": false,
+  "use_absolute_position_embeddings": true,
   "use_auxiliary_head": true,
   "use_mask_token": false,
   "use_mean_pooling": true,
-  "use_relative_position_bias": true,
+  "use_relative_position_bias": false,
   "use_shared_relative_position_bias": false,
   "vocab_size": 8192
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:87a3e53f87c26251c9485be8d2e6b4896fb3ca5fbc1fe742353ffe4c57e091e8
-size 343077252
+oid sha256:b5ab9b8b1e14e6152ce977b120fdb25da84fd41b3cc0f9e81bfe849771b48f31
+size 343258940

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a0a2d72729478c9c56a6673fbc1a93a59085bf2829c446ff056e0e39458bcde9
-size 5304
+oid sha256:a8ccf78e753d42e33d8326a19b03c04ad80293fadf790fa06fff55f386aebcd4
+size 5240