---
license: llama2
tags:
  - arxiv:2607.07964
  - kronq
  - quantization
  - hessian
  - fisher
task_categories:
  - text-generation
---

# Llama-2-7b — KronQ H_G (output-side gradient covariance)

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

Pre-computed **H_G** for [Llama-2-7b](https://huggingface.co/meta-llama/Llama-2-7b-hf), the **output-side** curvature factor used by **KronQ** under the K-FAC factorization `H ≈ H_X ⊗ H_G`. H_G is the per-sublayer sampled-Fisher gradient covariance (labels drawn from the model distribution) (`E[g gᵀ]` over the layer output), distinct from the standard input-side Hessian `H_X` (which GPTQ/GPTAQ build online during calibration).

Publishing this lets you **reproduce KronQ quantization without the offline Fisher precompute step**.

## Contents (32 layers × 7 sublayers, ~39 GB)

```
layer_<i>/self_attn_{q,k,v,o}_proj_G.pt
layer_<i>/mlp_{gate,up,down}_proj_G.pt
metadata.pt
```

Each `*_G.pt` is the `out_features × out_features` gradient covariance for that sublayer.

## Usage

Point KronQ's `--grad_dir` at the downloaded folder — it skips `precompute_gradients.py`:

```bash
python main.py --model meta-llama/Llama-2-7b-hf \
    --w_bits 4 --w_groupsize -1 --w_clip --w_asym --a_bits 16 --act_order \
    --bi_calibration --use_gptaq --incoh_rotate --incoh_kernel had --incoh_mode full \
    --alpha 0.25 --grad_dir <downloaded_HG_dir>
```

This is **raw (unrotated) H_G**, the form used by the weight-only recipe (H_G cancels in the OBS update, so raw and rotated give identical weights for per-channel weight-only).

## License

Derived from Llama-2-7b — subject to the [Llama 2 Community License](https://ai.meta.com/llama/license/).