Downloads · 30 days
0
qwen25-tax-3b-v2-adapters
dennisonb/qwen25-tax-3b-v2-adapters
qwen25-tax-3b-v2-adapters is a machine learning model from dennisonb. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
LoRA adapters for the v2 IRS Tax Code fine-tune of Qwen2.5-3B-Instruct.
Access
Public
Updated Mar 28, 2026
Repo size
692 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors958 MB · 100%
At a glance
- Access
- Public
- Created
- Mar 28, 2026
- Updated
- Mar 28, 2026
- SHA
- c2c04d2d
Model card
From the Hugging Face model README
qwen25-tax-3b-v2 — LoRA Adapters
LoRA adapters for the v2 IRS Tax Code fine-tune of Qwen2.5-3B-Instruct.
Training Pipeline
| Stage | Directory | LoRA Rank | Steps | Notes |
|---|---|---|---|---|
| SFT | sft/ | 32 | 1000 | Supervised fine-tune on IRC sections |
| DPO | dpo/ | 16 | 500 | Direct preference optimization; init from SFT |
| GRPO | grpo/ | 32 | 300 | Group relative policy optimization; init from DPO |
Usage (mlx-lm)
pip install mlx-lm
# Generate with GRPO adapter (best quality):
mlx_lm.generate \
--model Qwen/Qwen2.5-3B-Instruct \
--adapter-path grpo/ \
--prompt "What is the standard deduction for 2024?"
GGUF
See dennisonb/qwen25-tax-3b-v2-GGUF for a ready-to-use GGUF (q8_0) for Ollama / llama.cpp.
- Created
- Mar 28, 2026
- Updated
- Mar 28, 2026