---
base_model: meta-llama/Llama-2-7b-hf
language:
- en
license: llama2
pipeline_tag: text-generation
library_name: transformers
tags:
- kronq
- quantization
- group-quantization
- fake-quant
- fp16
---

# Llama-2-7b — KronQ W3A16 g128 (fake-quant fp16)

**Paper:** [arXiv:2607.07964](https://arxiv.org/abs/2607.07964) · **Code:** [GitHub](https://github.com/Intelligent-Computing-Lab-Panda/KronQ)

> ⚠️ **Fake-quant fp16 checkpoint.** 3-bit group-128 weights stored in **fp16** (KronQ does not pack int3) — **same size as bf16**, for PPL/accuracy reproduction. For deployable low-bit see the W4A16-g128 / W2A16-g128 (packed) repos.

[Llama-2-7b](https://huggingface.co/meta-llama/Llama-2-7b-hf) quantized to 3-bit weights (group 128) with **KronQ**, exported as a standard fp16 model.

## Results (WikiText-2, seqlen 2048)

**Perplexity:** **5.773**

**Zero-shot accuracy:**

| PIQA | ARC-E | ARC-C | HellaSwag | WinoGrande | BoolQ | OBQA | Average |
|---|---|---|---|---|---|---|---|
| 77.48 | 72.01 | 43.43 | 73.32 | 67.40 | 75.11 | 40.80 | **64.22** |

(lm-evaluation-harness 0-shot.)

## Usage

Loads as a **standard fp16 model** (no KronQ code):

```python
from transformers import AutoModelForCausalLM
m = AutoModelForCausalLM.from_pretrained("donghyunli/Llama-2-7b-KronQ-W3A16-g128-fake", torch_dtype="float16", device_map="auto")
```

## Recipe

Group-128 asymmetric W3, weight-only, `--alpha 0.25`, `--act_order`, BiIP, raw H_G.

## License

Derivative of Llama-2-7b — [llama2 license](https://ai.meta.com/llama/license/).