EQ3000
EQ3000
同样出现这个问题了。把modeling_webglm.py 代码中的 max_length=1024改成2048也不行,但是跑英文好像没问题
replace cocoeval.py line 506, 507 to: self.iouThrs = np.linspace(.5, 0.95, 10, endpoint=True) self.recThrs = np.linspace(.0, 1.00, 101, endpoint=True)
I am also finding this solution.!
I am also needing this THX!
Inference for image is OK, but **Inference for video is error.** I have the same question, also 4 bit is no use for me. /opt/conda/conda-bld/pytorch_1682343967769/work/aten/src/ATen/native/cuda/IndexKernel.cu:92: operator(): block: [51,0,0], thread: [96,0,0]...