Spaces:
Sleeping
Sleeping
Commit History
-
Align with OFPBadWord pattern: FastAPI app, clean routes, no ZeroGPU
BolyosCsaba +1: Claude Sonnet 4.6- Create own FastAPI() app, register /manifest and /ofp directly on it (no middleware hacks needed — FastAPI owns the routes before Gradio mounts) - Mount Gradio inside FastAPI with gr.mount_gradio_app, serve with uvicorn - Remove ZeroGPU (@spaces.GPU, import spaces): GPU acquisition latency often exceeds inference time for Qwen3-0.6B with thinking disabled - llm_client: device_map="auto" (GPU if available, CPU fallback) - Clean up endpoint paths to /manifest and /ofp (matching OFPBadWord) - Remove 'spaces' from requirements.txt Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>14c3b4e -
Optimize Qwen3 for fast social chat: disable thinking mode, cap tokens
BolyosCsaba +1: Claude Sonnet 4.6- Disable Qwen3's <think>...</think> reasoning block (enable_thinking=False) which was the primary cause of timeouts — the model was spending most of its time on chain-of-thought before generating a single word of response - Reduce max_tokens from 16384 → 300 (conversational replies don't need 16k) - Update system prompt to reinforce short, natural 1–3 sentence replies - Add CLAUDE.md with architecture and routing documentation Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>5f3a4a0
-
er gpu changes
BolyosCsaba86185d3 -
added smaller qwne model
BolyosCsaba8c2cdf0 -
decoration try for zero GPU
BolyosCsaba40b9f99 -
fix
BolyosCsabaab20dfe -
fix
BolyosCsabaec5c857 -
GPU enabled
BolyosCsaba31595f1 -
qwen 3 through transformers
BolyosCsabad905d29
-
UPdate your project to use Qwen3-4B-Instruct-2507 and remove all other model references
BolyosCsaba60c3b5b
-
initial commit
BolyosCsabae214abd