Upload folder using huggingface_hub

README.md

@@ -0,0 +1,56 @@
+---
+license: cc-by-nc-4.0
+base_model: kaonai/kaon-c-mistral-12b-v0.1-dpo-v0.1
+tags:
+- peft
+- lora
+- dpo
+- role-playing
+- chinese
+library_name: peft
+pipeline_tag: text-generation
+---
+
+# Kaon-W: GPT-4 Optimized DPO LoRA
+
+基于  进行GPT-4数据优化的DPO LoRA适配器。
+
+## 模型信息
+
+- **基础模型**: kaonai/kaon-c-mistral-12b-v0.1-dpo-v0.1
+- **训练方法**: DPO (Direct Preference Optimization)
+- **适配器类型**: LoRA (Low-Rank Adaptation)
+- **数据优化**: GPT-4全面优化的对话数据
+- **训练硬件**: 8x H200 GPU
+- **训练时长**: 5.3小时 (968步, 2轮)
+
+## 训练配置
+
+- **LoRA Rank**: 64
+- **LoRA Alpha**: 64
+- **学习率**: 2e-6
+- **Beta**: 0.12
+- **批次大小**: 2 (per device)
+- **梯度累积**: 4步
+
+## 使用方法
+
+
+
+## 训练数据
+
+使用GPT-4优化的DPO数据集:
+- 主数据集: 3,500条优化样本
+- NSFW修复: 150条优化样本  
+- 防重复: 297条优化样本
+- 总计: 3,947条高质量训练样本
+
+## 性能指标
+
+- **最终准确率**: 96.15%
+- **奖励边距**: 20.14+
+- **训练损失**: 1.27 (完美收敛)
+
+## 许可证
+
+本模型采用 CC-BY-NC-4.0 许可证，仅供非商业用途。

adapter_config.json

@@ -0,0 +1,34 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "/mnt/local/zeyu/models/kaon-c-mistral-12b-v0.1-dpo-v0.1",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": [],
+  "peft_type": "LORA",
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "k_proj",
+    "down_proj",
+    "o_proj",
+    "up_proj",
+    "q_proj",
+    "v_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": false
+}

adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6118dc0c1fe4fa1f135519b2fcab1c4612f6c256a9554f94c30050f74c6a7ab3
+size 912336848

upload_script.sh

@@ -0,0 +1,56 @@
+#!/bin/bash
+
+# HuggingFace模型上传脚本
+# ================================
+
+echo "🚀 HuggingFace模型上传脚本"
+echo "=============================="
+
+# 检查是否已登录
+echo "📋 检查HuggingFace登录状态..."
+if ! ~/.local/bin/hf whoami &>/dev/null; then
+    echo "❌ 未登录HuggingFace，请先登录:"
+    echo "   ~/.local/bin/hf login"
+    echo "   然后输入你的Access Token"
+    exit 1
+fi
+
+echo "✅ 已登录HuggingFace"
+
+# 获取用户输入
+read -p "请输入模型名称 (例如: 你的用户名/kaon-w): " REPO_NAME
+
+if [ -z "$REPO_NAME" ]; then
+    echo "❌ 模型名称不能为空"
+    exit 1
+fi
+
+echo "📂 准备上传模型: $REPO_NAME"
+echo "📁 文件列表:"
+ls -la /tmp/kaon-w-upload/
+
+echo ""
+read -p "确认上传? (y/N): " CONFIRM
+
+if [[ $CONFIRM != [yY] ]]; then
+    echo "❌ 取消上传"
+    exit 0
+fi
+
+echo "⬆️ 开始上传..."
+cd /tmp/kaon-w-upload/
+
+# 上传模型
+~/.local/bin/hf upload $REPO_NAME . --repo-type model
+
+if [ $? -eq 0 ]; then
+    echo ""
+    echo "🎉 上传成功！"
+    echo "🔗 模型地址: https://huggingface.co/$REPO_NAME"
+    echo ""
+    echo "📋 使用方法:"
+    echo "from peft import PeftModel"
+    echo "model = PeftModel.from_pretrained(base_model, '$REPO_NAME')"
+else
+    echo "❌ 上传失败，请检查网络连接和权限"
+fi