Evaluation Metrics
At pruna, we focus on making your models smaller, faster, cheaper, and greener. So, we provide metrics to measure your model’s speed, memory usage, carbon emissions, energy spent, etc…
Speed Evaluation
total_time, latency, throughput
Memory Evaluation
disk, inference & training_memory…
Energy
Energy_consumed, co2_emissions
Pruna showcases objective metrics that capture quality, coherence, and originality. Our metrics catalog covers many existing quality metrics for image generation models.
Fidelity
FID, CMMD
Alignment
Clip Score
Diversity
PSNR, SSIM
Evaluation helps you understand how compression affects your model's quality and inference.
Not sure how far your model can be optimized? Do you think you reached the maximum performance?
Learn more about Evaluation Metrics with our blog articles
