⚙️ Optimizer
🌍 Compute Footprint
⚙️ Optimizer
🌍 Compute Footprint
Model Configuration
Preset Model
Custom Model ID (overrides dropdown)
Leave blank to use dropdown selection
Device
Max Perplexity Increase Tolerance (%)
↺
0
5
Calibration
Calibration Samples (1–32)
↺
1
32
Sequence Length
Calibration Dataset
Allowed Precisions
FP16
BF16
INT8 (CUDA only)
⚡ Run Optimization
Optimization Log
Real-Time Logs
Results
Base TPS
Optimized TPS
Speedup ×
Base Memory (MB)
Optimized Memory (MB)
Memory Saved %
Base Perplexity
Optimized Perplexity
PPL Δ %
⬇️ Download Optimized Model
Optimized Model (ZIP — load with HuggingFace)