---
library_name: transformers
tags:
- grok
- unsloth
license: other
license_name: grok-2
license_link: https://huggingface.co/xai-org/grok-2/blob/main/LICENSE
base_model:
- xai-org/grok-2
---
<div>
  <p style="margin-bottom: 0; margin-top: 0;">
    <strong>Learn how to run Grok 2 correctly - <a href="https://docs.unsloth.ai/basics/grok-2">Read our Guide</a>.</strong>
  </p>
<p style="margin-top: 0;margin-bottom: 0;">
    <em><a href="https://docs.unsloth.ai/basics/unsloth-dynamic-v2.0-gguf">Unsloth Dynamic 2.0</a> achieves superior accuracy & outperforms other leading quants.</em>
  </p>
  <div style="display: flex; gap: 5px; align-items: center; ">
    <a href="https://github.com/unslothai/unsloth/">
      <img src="https://github.com/unslothai/unsloth/raw/main/images/unsloth%20new%20logo.png" width="133">
    </a>
    <a href="https://discord.gg/unsloth">
      <img src="https://github.com/unslothai/unsloth/raw/main/images/Discord%20button.png" width="173">
    </a>
    <a href="https://docs.unsloth.ai/basics/grok-2">
      <img src="https://raw.githubusercontent.com/unslothai/unsloth/refs/heads/main/images/documentation%20green%20button.png" width="143">
    </a>
  </div>
<h1 style="margin-top: 0rem;">Grok 2 Usage Guidelines</h1>
</div>

- Use `--jinja` for `llama.cpp`. You must use [PR 15539](https://github.com/ggml-org/llama.cpp/pull/15539). For example use the code below: <br>
- `git clone https://github.com/ggml-org/llama.cpp`
- `cd llama.cpp && git fetch origin pull/15539/head:MASTER && git checkout MASTER && cd ..`

Utilizes Alvaro's Grok-2 HF compatible tokenizer as provided [here](https://huggingface.co/alvarobartt/grok-2-tokenizer)

# Grok 2

This repository contains the weights of Grok 2, a model trained and used at xAI in 2024.

## Usage: Serving with SGLang

- Download the weights. You can replace `/local/grok-2` with any other folder name you prefer.

  ```
  hf download xai-org/grok-2 --local-dir /local/grok-2
  ```

  You might encounter some errors during the download. Please retry until the download is successful.  
  If the download succeeds, the folder should contain **42 files** and be approximately 500 GB.

- Launch a server.

  Install the latest SGLang inference engine (>= v0.5.1) from https://github.com/sgl-project/sglang/

  Use the command below to launch an inference server. This checkpoint is TP=8, so you will need 8 GPUs (each with > 40GB of memory).
  ```
  python3 -m sglang.launch_server --model /local/grok-2 --tokenizer-path /local/grok-2/tokenizer.tok.json --tp 8 --quantization fp8 --attention-backend triton
  ```

- Send a request.

  This is a post-trained model, so please use the correct [chat template](https://github.com/sgl-project/sglang/blob/97a38ee85ba62e268bde6388f1bf8edfe2ca9d76/python/sglang/srt/tokenizer/tiktoken_tokenizer.py#L106).

  ```
  python3 -m sglang.test.send_one --prompt "Human: What is your name?<|separator|>\n\nAssistant:"
  ```

  You should be able to see the model output its name, Grok.

  Learn more about other ways to send requests [here](https://docs.sglang.ai/basic_usage/send_request.html).

## License

The weights are licensed under the [Grok 2 Community License Agreement](https://huggingface.co/xai-org/grok-2/blob/main/LICENSE).