chore: training args init

Author: ryanlinjui

train.ipynb

@@ -119,7 +119,7 @@
     "DATASETS_REPO_ID = \"ryanlinjui/menu-zh-TW\"              # set your dataset repo id for training\n",
     "PRETRAINED_MODEL_REPO_ID = \"naver-clova-ix/donut-base\"  # set your pretrained model repo id for fine-tuning\n",
     "TASK_PROMPT_NAME = \"<s_menu-text-detection>\"            # set your task prompt name for training\n",
-    "MAX_LENGTH = 1024                                       # set your max length for maximum output length, max to 1536 for donut-base\n",
+    "MAX_LENGTH = 768                                        # set your max length for maximum output length, max to 1536 for donut-base\n",
     "IMAGE_SIZE = [1280, 960]                                # set your image size for training\n",
     "\n",
     "register_heif_opener()\n",
@@ -185,7 +185,7 @@
     "\n",
     "HUGGINGFACE_MODEL_ID = \"ryanlinjui/donut-base-finetuned-menu\" # set your huggingface model repo id for saving / pushing to the hub\n",
     "EPOCHS = 100            # set your training epochs\n",
-    "TRAIN_BATCH_SIZE = 1    # set your training batch size\n",
+    "TRAIN_BATCH_SIZE = 8    # set your training batch size\n",
     "LEARNING_RATE = 3e-5    # set your learning rate\n",
     "WEIGHT_DECAY = 0.1      # set your weight decay\n",
     "\n",
@@ -231,17 +231,17 @@
     "    per_device_eval_batch_size=1,\n",
     "    output_dir=\"./.checkpoints\",\n",
     "    seed=42,\n",
-    "    warmup_steps=30,\n",
+    "    warmup_steps=300,\n",
     "    eval_strategy=\"steps\",\n",
-    "    eval_steps=200,\n",
+    "    eval_steps=1000,\n",
     "    fp16=(device == \"cuda\"),\n",
     "    predict_with_generate=True,\n",
     "    generation_max_length=MAX_LENGTH,\n",
     "    generation_num_beams=1,\n",
     "    logging_strategy=\"steps\",\n",
     "    logging_steps=50,\n",
     "    save_strategy=\"steps\",\n",
-    "    save_steps=200,\n",
+    "    save_steps=1000,\n",
     "    push_to_hub=True if HUGGINGFACE_MODEL_ID else False,\n",
     "    hub_model_id=HUGGINGFACE_MODEL_ID,\n",
     "    hub_strategy=\"every_save\",\n",
@@ -272,7 +272,7 @@
     "\n",
     "MODEL_REPO_ID = \"ryanlinjui/donut-base-finetuned-menu\"\n",
     "TASK_PROMPT_NAME = \"<s_menu-text-detection>\"\n",
-    "MAX_LENGTH = 1024\n",
+    "MAX_LENGTH = 768\n",
     "IMAGE_SIZE = [1280, 960]\n",
     "\n",
     "processor = DonutProcessor.from_pretrained(MODEL_REPO_ID)\n",
@@ -306,7 +306,7 @@
  ],
  "metadata": {
   "kernelspec": {
-   "display_name": ".venv",
+   "display_name": "menu-text-detection",
    "language": "python",
    "name": "python3"
   },