Upload train_script.py with huggingface_hub

train_script.py

@@ -0,0 +1,83 @@
+# /// script
+# dependencies = ["trl>=0.12.0", "peft>=0.7.0", "trackio", "datasets"]
+# ///
+
+from datasets import load_dataset
+from peft import LoraConfig
+from trl import SFTTrainer, SFTConfig
+import trackio
+
+# 데이터셋 로드
+print("데이터셋 로드 중...")
+dataset = load_dataset("epinfomax/youtube-thumbnail-analysis", split="train")
+print(f"데이터셋 크기: {len(dataset)}개")
+
+# 학습/검증 분리
+dataset_split = dataset.train_test_split(test_size=0.1, seed=42)
+
+# LoRA 설정
+peft_config = LoraConfig(
+    r=16,
+    lora_alpha=32,
+    lora_dropout=0.05,
+    target_modules=["q_proj", "k_proj", "v_proj", "o_proj"],
+    task_type="CAUSAL_LM"
+)
+
+# 학습 설정
+training_args = SFTConfig(
+    output_dir="./outputs",
+
+    # Hub 저장 (필수!)
+    push_to_hub=True,
+    hub_model_id="epinfomax/youtube-thumbnail-trend-analyzer",
+    hub_strategy="every_save",
+
+    # 학습 파라미터
+    num_train_epochs=3,
+    per_device_train_batch_size=2,
+    gradient_accumulation_steps=4,
+    learning_rate=2e-4,
+    warmup_ratio=0.1,
+
+    # 평가
+    eval_strategy="steps",
+    eval_steps=20,
+
+    # 저장
+    save_strategy="steps",
+    save_steps=50,
+    save_total_limit=2,
+
+    # 최적화
+    gradient_checkpointing=True,
+    bf16=True,
+
+    # Trackio 모니터링
+    report_to="trackio",
+    run_name="youtube-thumbnail-trainer",
+
+    # 로깅
+    logging_steps=10,
+)
+
+# 트레이너 생성
+print("트레이너 초기화 중...")
+trainer = SFTTrainer(
+    model="Qwen/Qwen2.5-0.5B",
+    train_dataset=dataset_split["train"],
+    eval_dataset=dataset_split["test"],
+    peft_config=peft_config,
+    args=training_args,
+)
+
+# 학습 시작
+print("학습 시작!")
+trainer.train()
+
+# Hub에 최종 모델 저장
+print("모델 Hub에 저장 중...")
+trainer.push_to_hub()
+
+print("학습 완료!")
+print(f"모델 저장됨: https://huggingface.co/epinfomax/youtube-thumbnail-trend-analyzer")