# Training models

**URL:** <https://discuss.huggingface.co/t/training-models/154372>\
**Category:** Models\
**Created:** [May 10, 2025, 2:59am UTC](https://discuss.huggingface.co/t/training-models/154372 "2025-05-10T02:59:02Z")\
**Posts on this page:** 4\
**Page:** 1

<div class="post-metadata">

**Author:** ![Enderchef](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/enderchef/32/47235_2.png) [@Enderchef](https://discuss.huggingface.co/u/Enderchef)\
**Post date:** [May 10, 2025, 2:59am UTC](https://discuss.huggingface.co/t/training-models/154372/1 "2025-05-10T02:59:02Z")

</div>

I need support to train a model. I am trying to train the model _unsloth/Llama-4-Scout-17B-16E-Instruct-GGUF_ on the dataset _Enderchef/ICONN-1-BasicChat-Data-SuperLite_. I keep getting errors and was curious if somebody could either do the code for me or to the training for me.

---

<div class="post-metadata">

**Author:** ![John6666](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/john6666/32/27664_2.png) [@John6666](https://discuss.huggingface.co/u/John6666)\
**Post date:** [May 10, 2025, 5:17am UTC](https://discuss.huggingface.co/t/training-models/154372/2 "2025-05-10T05:17:50Z")

</div>

I don’t do coding or training for you, but…

GGUF is like corned beef, so it’s good to use, but not suitable for training. It’s better to train the pre-quantized model and then convert it to GGUF again.

> **[Common AI Model Formats](https://huggingface.co/blog/ngxson/common-ai-model-formats)**
>
> A Blog post by Xuan-Son Nguyen on Hugging Face

> **[Introduction - Hugging Face LLM Course](https://huggingface.co/learn/llm-course/chapter1/1)**
>
> We’re on a journey to advance and democratize artificial intelligence through open source and open science.

[https://unsloth.ai/blog/llama4](https://unsloth.ai/blog/llama4)

> **[Common AI Model Formats](https://huggingface.co/blog/ngxson/common-ai-model-formats)**
>
> A Blog post by Xuan-Son Nguyen on Hugging Face

---

<div class="post-metadata">

**Author:** ![Enderchef](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/enderchef/32/47235_2.png) [@Enderchef](https://discuss.huggingface.co/u/Enderchef)\
**Post date:** [May 10, 2025, 2:31pm UTC](https://discuss.huggingface.co/t/training-models/154372/3 "2025-05-10T14:31:15Z")

</div>

Thanks for letting me know.

---

<div class="post-metadata">

**Author:** ![Pimpcat-AU](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/pimpcat-au/32/48989_2.png) [@Pimpcat-AU](https://discuss.huggingface.co/u/Pimpcat-AU)\
**Post date:** [June 10, 2025, 8:22pm UTC](https://discuss.huggingface.co/t/training-models/154372/4 "2025-06-10T20:22:31Z")

</div>

Direct answer and script:

```
You cannot train directly on a GGUF file.

Convert GGUF to a supported PyTorch or Hugging Face format, train, then quantize and convert back to GGUF.

```

Example:

1. Download or export model in HuggingFace/transformers format, not GGUF.
2. Train/fine-tune with your chosen framework (e.g. unsloth, transformers, etc).
3. Convert the resulting checkpoint to GGUF after training.

For Unsloth, see: [Llama 4 - Finetune & Run with Unsloth](https://unsloth.ai/blog/llama4)

Solution provided by Triskel Data Deterministic AI.