update

examples/tutorials/dpo/ultrafeedback-dpo/step_2_train_dpo_model_single_gpu.py

@@ -4,7 +4,7 @@
 https://huggingface.co/docs/trl/v0.16.1/en/sft_trainer
 
 单卡 V00 32G 全参微调
-python3 step_2_train_sft_model_single_gpu.py
+python3 step_2_train_dpo_model_single_gpu.py
 
 """
 import argparse
@@ -27,8 +27,7 @@ else:
 from datasets import load_dataset
 import torch
 
-from modelscope import AutoModelForCausalLM
-from transformers import AutoTokenizer
+from modelscope import AutoModelForCausalLM, AutoTokenizer
 from trl import DPOConfig, DPOTrainer
 
 

examples/tutorials/dpo/ultrafeedback-dpo/step_6_push_to_modelscope.py

@@ -0,0 +1,49 @@
+#!/usr/bin/python3
+# -*- coding: utf-8 -*-
+# !/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+python step_6_push_to_modelscope.py \
+--model_dir /root/autodl-tmp/OpenMiniMind/temp/trainer_output/Qwen3-8B-sft-deepspeed/checkpoint-100 \
+--repo_id gyd2021/Qwen3-8B-sft-deepspeed \
+--token your_modelscope_token
+"""
+
+import os
+import argparse
+import logging
+
+# 设置日志格式
+logging.basicConfig(
+    level=logging.INFO,
+    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s'
+)
+logger = logging.getLogger(__name__)
+
+from modelscope.hub.api import HubApi
+
+
+def get_args():
+    parser = argparse.ArgumentParser()
+
+    parser.add_argument("--model_dir", type=str,)
+    parser.add_argument("--repo_id", type=str)
+    parser.add_argument("--token", type=str, default=None,)
+    parser.add_argument("--chunk_size", type=int, default=50)
+
+    return parser.parse_args()
+
+
+def main():
+    args = get_args()
+
+    api = HubApi(token=args.token)
+    api.upload_folder(
+        repo_id=args.repo_id,
+        folder_path=args.model_dir,
+    )
+    return
+
+
+if __name__ == "__main__":
+    main()