muradinfosec/deepseek-v2-chat

DeepSeek-V2-Chat on Hugging Face Spaces

Runs DeepSeek-V2-Chat (236B MoE) with INT8 quantization on 4× A100 80GB using vLLM + Gradio.

Setup

1. 1.Fork this Space
2. 2.Settings → Hardware → select Nvidia 4x A100 ($10/hr)
3. 3.Settings → Secrets → add HF_TOKEN = your Hugging Face token
4. 4.Space builds automatically — first boot takes ~30 min (model download)

Usage

Open the Space URL and start chatting. Adjust temperature, max tokens, and system prompt in the sidebar.