Upload korscideberta-abstractcls.ipynb

korscideberta-abstractcls.ipynb

@@ -451,7 +451,7 @@
  "gc.collect()\n",
  "from transformers import TrainingArguments\n",
  "training_args = TrainingArguments(\n",
- " output_dir=\"korscideberta_abs1\",\n",
+ " output_dir=\"deberta_sent4455\",\n",
  " num_train_epochs=4, \n",
  " #learning_rate=5e-5,\n",
  " learning_rate=1.5e-5,\n",
@@ -674,6 +674,8 @@
  "source": [
  "## **Evaluate / Analyse the model**\n",
  "\n",
+ "학습된 파인튜닝 모델로 추론 및 평가하는 단계\n",
+ "\n",
  "Now that we have fine-tuned the model we need to be very careful about analyzing its performance. It's usually not enough to just look at basic metrics defining the quality of a model purely on a metric, such as *accuracy*.\n",
  "The better approach is to find a metric that best describes the actual use case of the model.\n",
  "\n",
@@ -787,7 +789,7 @@
  "source": [
  "#Huggingface에 업로드한 파인튜닝 모델 및 성능 확인\n",
  "from transformers import AutoModelForSequenceClassification\n",
- "model = AutoModelForSequenceClassification.from_pretrained(\"kkmkorea/deberta_sent4455\")\n",
+ "model = AutoModelForSequenceClassification.from_pretrained(\"kkmkorea/deberta_sent4455\") #위에서 학습한 파인튜닝 모델 경로&저장소\n",
  "#model = AutoModelForSequenceClassification.from_pretrained(\"patrickvonplaten/deberta_v3_amazon_reviews\")\n",
  "\n",
  "trainer = Trainer(\n",