# Which Transformers/Libraries Should I use?

**URL:** <https://discuss.huggingface.co/t/which-transformers-libraries-should-i-use/131917>\
**Category:** Beginners\
**Created:** [December 17, 2024, 3:21pm UTC](https://discuss.huggingface.co/t/which-transformers-libraries-should-i-use/131917 "2024-12-17T15:21:56Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![peteletkemanca](https://avatars.discourse-cdn.com/v4/letter/p/ee59a6/32.png) [@peteletkemanca](https://discuss.huggingface.co/u/peteletkemanca)\
**Post date:** [December 17, 2024, 3:21pm UTC](https://discuss.huggingface.co/t/which-transformers-libraries-should-i-use/131917/1 "2024-12-17T15:21:56Z")

</div>

I am able to use [LM Studio](https://lmstudio.ai/) to prompt against Hugging Face models which I’ve downloaded with prompts like

```auto
Rewrite the following email make it sound more professional:
....

```

And

```auto
What does the follow Python code do
....

```

And

```auto
Rewrite the following Python code as Java code
....

```

Which HF transformers/libraries should I be using to create something like this with HF and Python?

---

<div class="post-metadata">

**Author:** ![Alanturner2](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/alanturner2/32/37542_2.png) [@Alanturner2](https://discuss.huggingface.co/u/Alanturner2)\
**Post date:** [December 17, 2024, 3:43pm UTC](https://discuss.huggingface.co/t/which-transformers-libraries-should-i-use/131917/2 "2024-12-17T15:43:04Z")

</div>

- **LLaMA (e.g., CodeLlama)**: Great for general-purpose text and code generation. Available variants specialized for code.  
Model: `codellama/CodeLlama-7b-hf`
- **Mistral** : Lightweight yet powerful, suitable for local use.  
Model: `mistralai/Mistral-7B-v0.1`
- **Falcon** : Optimized for performance and open-weight inference.  
Model: `tiiuae/falcon-7b-instruct`
- **GPT-NeoX / GPT-J** : Smaller GPT-like models optimized for general NLP tasks.  
Model: `EleutherAI/gpt-neox-20b` or `EleutherAI/gpt-j-6B`

**But**

I recommend you using chatgpt. Chatgpt has versatile ability to compile or generate python code.  
**BUT**  
You should consider that LLM is just model that was made by human. **It worse than you!!!**

---

<div class="post-metadata">

**Author:** ![peteletkemanca](https://avatars.discourse-cdn.com/v4/letter/p/ee59a6/32.png) [@peteletkemanca](https://discuss.huggingface.co/u/peteletkemanca)\
**Post date:** [December 17, 2024, 8:49pm UTC](https://discuss.huggingface.co/t/which-transformers-libraries-should-i-use/131917/3 "2024-12-17T20:49:22Z")

</div>

So those appear to be models, not exactly what I’m looking for.  
In LM Studio, you can load any GGUF model that your workstation has enough resources for.  
And you can have LM Studio prompt the LLM for things like “rewrite this code” or “summarize this email” etc.  
I think that if/when you were to use the HF Pipeline you only get one feature at a time e.g. classification or translation or summarization.  
Given that LM Studio _appears_ to be able to do more than one (but I could be wrong) using a single GUFF I’m wondering how I can do that?  
I want to create something that uses Python and the Electron JS framework.  
Maybe this is a matter of creating the correct prompts as the LM uses the same LLM for every single task.