# Pro Space GPU not detected: 'Can't initialize NVML' error

**URL:** <https://discuss.huggingface.co/t/pro-space-gpu-not-detected-cant-initialize-nvml-error/174696>\
**Category:** Spaces\
**Created:** [March 27, 2026, 10:31am UTC](https://discuss.huggingface.co/t/pro-space-gpu-not-detected-cant-initialize-nvml-error/174696 "2026-03-27T10:31:42Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![IArjunS](https://avatars.discourse-cdn.com/v4/letter/i/e99b99/32.png) [@IArjunS](https://discuss.huggingface.co/u/IArjunS)\
**Post date:** [March 27, 2026, 10:31am UTC](https://discuss.huggingface.co/t/pro-space-gpu-not-detected-cant-initialize-nvml-error/174696/1 "2026-03-27T10:31:43Z")

</div>

Hello,

I am new user of hugging face hub with user name IArjunS, and trying to host and run inference from a fine tuned ML model , I am not able to launch the app  
because of unavailibility of GPU i think. Can someone look into this error and provide help

Space URL: [Gemma 2b Finetuned Quotes Generator - a Hugging Face Space by IArjunS](https://huggingface.co/spaces/IArjunS/gemma-2b-finetuned-quotes-generator)  
My account type - (Pro subscription).

UserWarning: Can’t initialize NVML  
Loading model ‘IArjunS/gemma-2b-finetuned-quotes’ on device: cpu…  
Stopping Node.js server…  
My app.py has the DEVICE = “cuda” if torch.cuda.is\_available() else “cpu” logic, and in settings Free GPU seems to be running automatically but error log suggests otherwise.  
I think the GPU is not being properly exposed to my container.

Thank you  
Arjun

---

<div class="post-metadata">

**Author:** ![IArjunS](https://avatars.discourse-cdn.com/v4/letter/i/e99b99/32.png) [@IArjunS](https://discuss.huggingface.co/u/IArjunS)\
**Post date:** [March 27, 2026, 12:00pm UTC](https://discuss.huggingface.co/t/pro-space-gpu-not-detected-cant-initialize-nvml-error/174696/2 "2026-03-27T12:00:52Z")

</div>

I could solve the problem, i had not added @spaces.GPU in app.py .