GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
DennisHuang648
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
DennisHuang648
Model Tree
openbmb/MiniCPM5-1B
Input Modalities
Output Modalities
Supported Functionality
本 LoRA 的微调数据来源为 neko30k 数据集：
| 项目 | 说明 |
|---|---|
| 数据集名称 | neko30k（NekoQA-30K） |
| Hugging Face | liumindmind/NekoQA-30K |
| 样本量 | 30,834 条猫娘 QA 对话 |
| 类别 | ACG / 心理疗愈 / 创意写作 / 安全 / 数学 / 代码 / 职场 等 12 类 |
python
from datasets import load_datasetds = load_dataset("liumindmind/NekoQA-30K")tie_word_embeddings=False + GGUF special-token 修复）| 指标 | v2 (本仓库) | v1 |
|---|---|---|
| train/loss | 2.14 | ~2.07 |
| eval/loss | 2.18 | ~2.14 |
python
import torchfrom peft import PeftModelfrom transformers import AutoModelForCausalLM, AutoTokenizer BASE = "openbmb/MiniCPM5-1B"ADAPTER = "DennisHuang648/MiniCPM5-1B-NekoQA-v2-LoRA" tok = AutoTokenizer.from_pretrained(BASE, trust_remote_code=True)base = AutoModelForCausalLM.from_pretrained( BASE, trust_remote_code=True, torch_dtype=torch.bfloat16, attn_implementation="sdpa", device_map="auto",)model = PeftModel.from_pretrained(base, ADAPTER).eval() SYSTEM = ( "你是一只可爱的猫娘，名字叫宝宝。请用毛茸茸、撒娇、带「喵」「的说」" "「呜哇」等语气词的口吻，配合 (动作) 描述回应主人。")msgs = [ {"role": "system", "content": SYSTEM}, {"role": "user", "content": "我今天好累啊"},]text = tok.apply_chat_template( msgs, tokenize=False, add_generation_prompt=True, enable_thinking=False,)ids = tok(text, return_tensors="pt").to(model.device)ids.pop("token_type_ids", None)out = model.generate(**ids, max_new_tokens=200, do_sample=False, pad_token_id=tok.pad_token_id)print(tok.decode(out[0, ids.input_ids.shape[1]:], skip_special_tokens=True))llama.cpp / MiniCPM Desk Pet 请使用 GGUF 版本：
DennisHuang648/MiniCPM5-1B-NekoQA-v2-LoRA-GGUF
| 文件 | 说明 |
|---|---|
adapter_model.safetensors | PEFT LoRA 权重 (~22 MB) |
adapter_config.json | PEFT 配置 |
train_meta.json | 训练超参 |
capability_loss.jsonl | 24-prompt 能力回归测试结果 |
USAGE.md | 详细用法 |
openbmb/MiniCPM5-1Bfrom datasets import load_datasetds = load_dataset("liumindmind/NekoQA-30K")import torchfrom peft import PeftModelfrom transformers import AutoModelForCausalLM, AutoTokenizer BASE = "openbmb/MiniCPM5-1B"ADAPTER = "DennisHuang648/MiniCPM5-1B-NekoQA-v2-LoRA" tok = AutoTokenizer.from_pretrained(BASE, trust_remote_code=True)base = AutoModelForCausalLM.from_pretrained( BASE, trust_remote_code=True, torch_dtype=torch.bfloat16, attn_implementation="sdpa", device_map="auto",)model = PeftModel.from_pretrained(base, ADAPTER).eval() SYSTEM = ( "你是一只可爱的猫娘，名字叫宝宝。请用毛茸茸、撒娇、带「喵」「的说」" "「呜哇」等语气词的口吻，配合 (动作) 描述回应主人。")msgs = [ {"role": "system", "content": SYSTEM}, {"role": "user", "content": "我今天好累啊"},]text = tok.apply_chat_template( msgs, tokenize=False, add_generation_prompt=True, enable_thinking=False,)ids = tok(text, return_tensors="pt").to(model.device)ids.pop("token_type_ids", None)out = model.generate(**ids, max_new_tokens=200, do_sample=False, pad_token_id=tok.pad_token_id)print(tok.decode(out[0, ids.input_ids.shape[1]:], skip_special_tokens=True))