Downloads · 30 days
17
35% of all-time downloads
nimbus-sft-3b-lora
420yolomcswaggerpants/nimbus-sft-3b-lora
nimbus-sft-3b-lora is a machine learning model from 420yolomcswaggerpants. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for peft. The card lists the license as other.
A LoRA adapter fine-tuned on 4,000 synthetic Q&A pairs for Nimbus Coffee customer support. This is not a full model—it must be loaded onto Qwen/Qwen2.5-3B.
All-time downloads
48
Public
Repo size
18.8 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.json11.4 MB · 61%
At a glance
- Library
- peft
- License
- other
- Access
- Public
- Created
- Sep 4, 2026
- Updated
- Oct 8, 2026
- SHA
- ab638670
Base models
Model card
From the Hugging Face model README
Nimbus SFT 3B LoRA Adapter
A LoRA adapter fine-tuned on 4,000 synthetic Q&A pairs for Nimbus Coffee customer support. This is not a full model—it must be loaded onto Qwen/Qwen2.5-3B.
What This Is
- Base model: Qwen/Qwen2.5-3B
- Adapter: LoRA (Low-Rank Adaptation)
- Task: Customer support for Nimbus Coffee, a fictional Portland specialty coffee roaster
- Dataset: 4,000 synthetic instruction/response pairs generated by DeepSeek
Why LoRA?
LoRA adds small trainable adapter matrices instead of updating all model weights. This reduces memory and compute while preserving most of the base model's knowledge. It's the standard for fine-tuning large models on limited hardware.
Training Details
| Setting | Value |
|---|---|
| Epochs | 3 |
| Batch size | 1 |
| Gradient accumulation | 8 |
| Learning rate | 2e-4 |
| Weight decay | 0.01 |
| LR scheduler | Linear |
| Quantization | 4-bit NF4 |
| LoRA r | 8 |
| LoRA alpha | 32 |
| LoRA dropout | 0.1 |
| Target modules | q_proj, v_proj |
| Hardware | Google Colab T4 |
Usage
from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel
base_model = "Qwen/Qwen2.5-3B"
adapter = "420yolomcswaggerpants/nimbus-sft-3b-lora"
model = AutoModelForCausalLM.from_pretrained(base_model, torch_dtype="auto")
model = PeftModel.from_pretrained(model, adapter)
tokenizer = AutoTokenizer.from_pretrained(adapter)
prompt = "### Instruction:\nWhat coffees do you sell?\n\n### Response:\n"
inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=100)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
- Library
- peft
- License
- other
- Created
- Sep 4, 2026
- Updated
- Oct 8, 2026