Spaces:
Sleeping
Apply for community grant: Personal project (gpu)
Subject: Request for T4 GPU Grant for Interactive LLM Training Demo
Dear Hugging Face Team,
I’m building an educational Gradio-based LLM builder app that allows users to fine-tune small causal language models (like distilgpt2/gpt2) interactively — selecting architecture, starting training, and optionally pushing to the Hub. The goal is to make LLM training more accessible and understandable for students and developers.
While I’ve successfully prototyped this locally, deploying it on Hugging Face Spaces with ZeroGPU is unfortunately too limited for this use case — training even a small model like distilgpt2 requires sustained compute (5-15 mins), and ZeroGPU’s cold-start + timeout behavior interrupts the process before completion.
A persistent T4 GPU would allow this demo to:
- Run full training cycles reliably
- Show real-time logs and evaluation metrics
- Empower learners to experiment hands-on with fine-tuning
This is not for production or commercial use — purely educational. I’d be honored to receive a grant and will happily:
- Credit Hugging Face in the app
- Share the Space publicly for community learning
- Provide feedback or demo if helpful
Thank you for your incredible platform and support of open education 🙏
Warm regards,
Bc
Hello, unfortunately we don't provide grants for training.
Hi HF Team — quick clarification to avoid confusion:
I submitted two requests:
❌ First request: Framed as “training platform” → rightly declined (I understand — not what grants are for!)
✅ Second request: Reframed as “educational demo” — lightweight, 1–2 epoch fine-tuning for architecture/hyperparameter intuition — explicitly NOT production training.
This second request is my official, revised submission — aligned with Spaces like dreambooth-training and clip-prefix-training.
No need to consider the first — this is the one I’d love your feedback on 🙏
Thank you!
— Bc
@Keeby-smilyai we provide grants to only inference. the previous one was a mistake on my end, as I thought it was related to Segment Anything Model, meanwhile it was something else.
@merve Ok thank you! Could I please ask what type of GPUs huggingface grant in the community grant thingy? Thank you again! -regards Bc