GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
mikuhhn1239
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Run this model inference with full control and performance in your environment.
Model Details
Model Provider
mikuhhn1239
Model Tree
Base
mikuhhn1239/qwen3-8b-novel-base-sft
Adapter
this model
Input Modalities
Text
Output Modalities
Text
Supported Functionality
Dedicated Endpoints
Container
[1] "..." [2] "..." ...{"labels": [{"unit_id": "N", "type": "dialogue|narration|thought|action|scene_description"}]}| 类型 | 含义 |
|---|---|
dialogue | 对话 |
narration | 叙述 |
thought | 心理 |
action | 动作 |
scene_description | 场景描写 |
markdown
输入:[1] "你怎么来了？"[2] 她愣了一下。[3] 其实我也不知道自己为什么会来。输出:{"labels": [{"unit_id": "1", "type": "dialogue"},{"unit_id": "2", "type": "action"},{"unit_id": "3", "type": "thought"}]}
python
from transformers import AutoModelForCausalLMfrom peft import PeftModelbase = AutoModelForCausalLM.from_pretrained("mikuhhn1239/qwen3-8b-novel-base-sft",torch_dtype="auto", device_map="auto",)model = PeftModel.from_pretrained(base, "mikuhhn1239/qwen3-8b-narrative-parsing-lora")
markdown
基座: Qwen3-8B-Novel-Base-SFT (Stage1 全参 SFT, 72K 小说续写数据)方法: LoRA (r=64, α=128, dropout=0.05)数据: 616 条 (577 train / 39 val / 39 test)框架: transformers Trainer + PEFT优化器: AdamW (adamw_torch_fused), cosine schedule, warmup=5%epoch: 5 | LR: 1e-4 | batch: 1×16(accum) | bf16 | max_length: 4096
| 版本 | 数据量 | epochs | 硬件 | JSON解析 | 类型准确率 | 说明 |
|---|---|---|---|---|---|---|
| 零基座 | — | — | — | 0% | 0% | Qwen3-8B 原始完全不会 |
| +Stage1 | — | — | — | 0% | 0% | 读完 669 本也不会 |
| 类型 | v3.2 | v4 | 变化 |
|---|---|---|---|
| narration | — | 82% | 最强 |
| dialogue | — | 70% | |
| thought | — | 62% | |
| action | — | 58% | |
| scene_description |
""→「」 + max_new_tokens 256→1024| v1 | 56 | 3 | 单卡 | 57.1% | 25.0% | 端到端（切分+分类） |
| v2 | 310 | 3 | 单卡 | 2.6% | 63.6% | 只分类，引号冲突 |
| v3.1 | 310 | 5 | 单卡 | 2.6% | 63.6% | tokens=256 截断 JSON |
| v3.2 | 577 | 5 | 单卡 | 100% | 69.5% | 引号修复 + 扩标 + tokens→1024 |
| v4 ⭐ | 577 | 5 | 8 卡 DDP | 100% | 72.8% | 8 卡重训，+3.3pp，各类型全面提升 |
| — |
| 54% |
| 最弱 |
| 总体 | 69.5% | 72.8% | +3.3pp |