Spaces:
Running on Zero
Running on Zero
Commit History
-
Add direct-load BF16 repeat comparison
Verified signature647387b -
Add bounded larger-family CUDA quantizer controls
Verified signaturebcba9be -
Load 8B directly inside allocated ZeroGPU call
Verified signature7740aa9
-
Switch larger-family control to root-loaded Qwen3 8B
Verified signaturebd2dad5 -
Disable unused KV cache for embedding controls
Verified signature97e5e64 -
Expose larger-family benchmark traceback
Verified signature2fad4af -
Isolate GTE startup packing from completed 0.6B model
Verified signature234931c -
Move GTE 1.5B CUDA placement to ZeroGPU startup
Verified signature88534d9 -
Add mounted 1.5B and 8B BF16 cross-runtime controls
Verified signature7973967 -
Lazy-load CUDA quantizer configuration inside GPU call
Verified signature22338c1 -
Add bounded bitsandbytes INT8 and NF4 controls
Verified signature1a6347c -
Add bounded ZeroGPU BF16 CUDA control
Verified signaturebc3f111