Text-to-image generation on T2I-CompBench (B-VQA, UniDet, S-CoT)
75B-VQAPPCL
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| PPCLInference Latency (ms)=17922026.02 | 75 | 52.4 | 82.15 | |
| E-DiT-baseInference Latency (ms)=17022026.02 | 71.9 | 53.6 | 82.34 | |
| E-DiT-turboInference Latency (ms)=12832026.02 | 71.1 | 51.9 | 81.68 | |
| Base model (Qwen-Image)Inference Latency (ms)=24312026.02 | 70.9 | 53.2 | 82.47 | |
| HPInference Latency (ms)=17862026.02 | 70.6 | 48.7 | 79.94 | |
| TinyFusionInference Latency (ms)=17892026.02 | 68.9 | 46.4 | 78.99 |