Hi, team, thanks for your attention.
I try to evaluate the released model LMDrive-llava-v1.5-7b-v1.0, and observe that the evaluation process is too slow.
Here is the statistics of evaluation in one route scenario, is the evaluation speed as fast as expected ?
Evaluation running on one 3090 GPU.
========= Results of RouteScenario_51 (repetition 0) ------ SUCCESS =========
╒═════════════════════════════════╤═════════════════════╕
│ Start Time │ 2024-03-26 12:46:58 │
├─────────────────────────────────┼─────────────────────┤
│ End Time │ 2024-03-26 13:58:15 │
├─────────────────────────────────┼─────────────────────┤
│ Duration (System Time) │ 4276.92s │
├─────────────────────────────────┼─────────────────────┤
│ Duration (Game Time) │ 324.25s │
├─────────────────────────────────┼─────────────────────┤
│ Ratio (System Time / Game Time) │ 0.076 │
╘═════════════════════════════════╧═════════════════════╛
Hi, team, thanks for your attention. I try to evaluate the released model LMDrive-llava-v1.5-7b-v1.0, and observe that the evaluation process is too slow. Here is the statistics of evaluation in one route scenario, is the evaluation speed as fast as expected ? Evaluation running on one 3090 GPU.