Inference Performance Benchmarking on Standardized agricultural field image (test)
9.27Load Time (s)AgriChat (7B)
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| AgriChat (7B)GPU=NVIDIA RTX 3090 (24GB VRAM), Precision=4-bit, Batch size=1, Input image size=1328 KB, Repetitions=1002026.03 | 9.27 | 2.315 | 0.432 | 9.5 | 10.71 | 12.32 | |
| LLaVA-OneVision (7B)GPU=NVIDIA RTX 3090 (24GB VRAM), Precision=4-bit, Batch size=1, Input image size=1328 KB, Repetitions=1002026.03 | 9.98 | 1.546 | 0.647 | 12.94 | 9.99 | 10.99 | |
| Qwen-2.5 (VL-7B)GPU=NVIDIA RTX 3090 (24GB VRAM), Precision=4-bit, Batch size=1, Input image size=1328 KB, Repetitions=1002026.03 | 22.45 | 29.769 | 0.034 | 8.6 | 11.33 | 11.76 | |
| Llama-3.2 (Vision-11B)GPU=NVIDIA RTX 3090 (24GB VRAM), Precision=4-bit, Batch size=1, Input image size=1328 KB, Repetitions=1002026.03 | 29.05 | 9.938 | 0.101 | 25.86 | 7.18 | 8.16 |