Hugging Face's logo Hugging Face
  • Models
  • Datasets
  • Spaces
  • Buckets new
  • Docs
  • Enterprise
  • Pricing
    • Website
      • Tasks
      • HuggingChat
      • Collections
      • Languages
      • Organizations
    • Community
      • Blog
      • Posts
      • Daily Papers
      • Learn
      • Discord
      • Forum
      • GitHub
    • Solutions
      • Team & Enterprise
      • Hugging Face PRO
      • Enterprise Support
      • Inference Providers
      • Inference Endpoints
      • Storage Buckets

  • Log In
  • Sign Up

Spaces:
Reza2kn
/
uncgpt-69-hybrid-demo
Running

App Files Files Community
uncgpt-69-hybrid-demo
821 kB
Ctrl+K
Ctrl+K
  • 2 contributors
History: 25 commits
Reza Sayar
Fast path: fix off-by-one cache position (was reading uninitialised cache slot at promptLen β†’ attention contamination); fuse seed argmax + capture into prefill encoder (one fewer sync per generation)
813d4ba 18 days ago
  • assets
    Fast path: fix off-by-one cache position (was reading uninitialised cache slot at promptLen β†’ attention contamination); fuse seed argmax + capture into prefill encoder (one fewer sync per generation) 18 days ago
  • .gitattributes
    1.52 kB
    initial commit 30 days ago
  • README.md
    1.72 kB
    WebGPU + chat: real in-browser chat (no ZeroGPU), model picker, streaming tokens, download progress bar, conversation history. SP via sentencepiece-js matches Python tokenization exactly. v1 entry placeholder until re-training finishes. 18 days ago
  • index.html
    5.32 kB
    Fast path: fix off-by-one cache position (was reading uninitialised cache slot at promptLen β†’ attention contamination); fuse seed argmax + capture into prefill encoder (one fewer sync per generation) 18 days ago