Instructions to use Nwna/olmo3-7b-llm-engineering-lora with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Nwna/olmo3-7b-llm-engineering-lora with PEFT:

from peft import PeftModel
from transformers import AutoModelForCausalLM

base_model = AutoModelForCausalLM.from_pretrained("unsloth/olmo-3-7b-instruct-unsloth-bnb-4bit")
model = PeftModel.from_pretrained(base_model, "Nwna/olmo3-7b-llm-engineering-lora")

Transformers

How to use Nwna/olmo3-7b-llm-engineering-lora with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="Nwna/olmo3-7b-llm-engineering-lora")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("Nwna/olmo3-7b-llm-engineering-lora", dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use Nwna/olmo3-7b-llm-engineering-lora with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Nwna/olmo3-7b-llm-engineering-lora"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Nwna/olmo3-7b-llm-engineering-lora",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Nwna/olmo3-7b-llm-engineering-lora

SGLang

How to use Nwna/olmo3-7b-llm-engineering-lora with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "Nwna/olmo3-7b-llm-engineering-lora" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Nwna/olmo3-7b-llm-engineering-lora",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "Nwna/olmo3-7b-llm-engineering-lora" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Nwna/olmo3-7b-llm-engineering-lora",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Unsloth Studio new

How to use Nwna/olmo3-7b-llm-engineering-lora with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for Nwna/olmo3-7b-llm-engineering-lora to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for Nwna/olmo3-7b-llm-engineering-lora to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for Nwna/olmo3-7b-llm-engineering-lora to start chatting

Load model with FastModel

pip install unsloth
from unsloth import FastModel
model, tokenizer = FastModel.from_pretrained(
    model_name="Nwna/olmo3-7b-llm-engineering-lora",
    max_seq_length=2048,
)

Docker Model Runner
How to use Nwna/olmo3-7b-llm-engineering-lora with Docker Model Runner:
```
docker model run hf.co/Nwna/olmo3-7b-llm-engineering-lora
```

OLMo-3-7B 大模型工程化中文 LoRA Adapter

这是一个基于 allenai/Olmo-3-7B-Instruct / unsloth/olmo-3-7b-instruct-unsloth-bnb-4bit 训练得到的 LoRA adapter。训练目标是让模型更适合用中文回答大模型研发与工程化实践相关问题，例如 LoRA/PEFT、SFT 数据构造、RAG、向量检索、Agent、量化、部署与训练排障。

本仓库只包含 LoRA adapter，不包含完整基座模型权重。使用时需要先加载兼容的 OLMo-3-7B Instruct 基座模型，再加载本 adapter。

模型信息

模型类型：PEFT LoRA adapter
基座模型：unsloth/olmo-3-7b-instruct-unsloth-bnb-4bit
训练数据集：Nwna/llm-engineering-sft-8000
训练语言：中文为主
任务类型：中文技术问答、概念解释、方案对比、步骤指导、实践排错
训练方式：LoRA SFT，4bit 量化加载基座模型，只训练 adapter 参数

训练配置

主要训练参数如下：

base model = unsloth/olmo-3-7b-instruct-unsloth-bnb-4bit
dataset = Nwna/llm-engineering-sft-8000
num_train_epochs = 2.0
max_seq_length = 2048
per_device_train_batch_size = 4
gradient_accumulation_steps = 2
learning_rate = 1e-4
warmup_ratio = 0.05
lr_scheduler_type = cosine
optim = adamw_8bit
LoRA r = 16
LoRA alpha = 16
LoRA dropout = 0
target_modules = q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj

训练在 Modal H100 环境中完成，并使用 W&B 记录训练曲线和生成样例。

评估结果

本次训练完成 2 epoch，共 2000 step：

train_loss ≈ 1.1852

训练后使用固定 prompt 做了生成效果抽查，覆盖三类能力：

目标领域能力：LoRA、PEFT、4bit 量化、adapter 保存与部署、训练排障
相邻工程能力：RAG、Transformer、企业文档问答系统设计
通用能力保持：翻译、简单数学、邮件写作、常识解释

抽查结论：

LoRA/PEFT、RAG、Transformer 等目标领域回答整体可用。
翻译和简单数学回答正常。
个别通用常识问题仍可能出现事实错误，因此不建议把该 adapter 用作通用知识模型。

使用方式

示例代码：

from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel
import torch

base_model = "unsloth/olmo-3-7b-instruct-unsloth-bnb-4bit"
adapter = "Nwna/olmo3-7b-llm-engineering-lora"

tokenizer = AutoTokenizer.from_pretrained(adapter)
model = AutoModelForCausalLM.from_pretrained(
    base_model,
    torch_dtype=torch.bfloat16,
    device_map="auto",
)
model = PeftModel.from_pretrained(model, adapter)
model.eval()

messages = [
    {"role": "user", "content": "LoRA 和全参数微调有什么区别？为什么本作业选择 LoRA？"}
]
text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer([text], return_tensors="pt").to(model.device)

with torch.no_grad():
    outputs = model.generate(**inputs, max_new_tokens=512, do_sample=False)

print(tokenizer.decode(outputs[0], skip_special_tokens=True))

如果在 Unsloth 环境中使用，也可以先用 Unsloth 加载基座模型，再加载本 LoRA adapter。

适用场景

适合：

大模型研发课程作业演示
LoRA / PEFT 微调流程复现
中文大模型工程化知识问答实验
SFT 数据集与训练监控流程展示

不适合：

医疗、法律、金融等高风险决策
需要严格事实保证的生产问答系统
完全通用的知识问答模型
不经人工审核直接对外部署

局限性

数据集规模有限，覆盖的是大模型研发与工程化学习场景，不代表完整通用知识。
本仓库是 LoRA adapter，不能脱离兼容基座模型单独推理。
训练使用 4bit 量化加载基座模型，效果和全参数微调不可直接等同。
生成结果仍可能出现事实错误、表达重复或遗漏细节，使用前应结合人工评估。

关联资源

数据集：Nwna/llm-engineering-sft-8000
Adapter：Nwna/olmo3-7b-llm-engineering-lora

Downloads last month: 24

Nwna
/

olmo3-7b-llm-engineering-lora