修改视觉模型的tokens_per_second没有生效
#13
by
kevinBusinessGenrator
- opened
"vision_config": {
"hidden_size": 1280,
"in_chans": 3,
"intermediate_size": 3456,
"model_type": "qwen2_5_vl",
"out_hidden_size": 8192,
"spatial_patch_size": 14,
"tokens_per_second": 2,
"torch_dtype": "bfloat16"
},
修改视觉模型的tokens_per_second没有生效，这个参数是做什么用的