尝试使用最远点采样算子:
if name == 'main':
op1 = DFPS_Sampler()
features = paddle.cast(features, dtype=paddle.float32)
points = paddle.randn([2,16,3],dtype=paddle.float64)
features = paddle.randn([2,16,3],dtype=paddle.float64)
npoint = 4
idx = op1(points ,features, npoint)
出现如下错误:
File "/home/hjl/.cache/paddle_extensions/furthest_point_sample_ops/furthest_point_sample_ops_140434522163008.py", line 46, in furthest_point_sampling_wrapper
_dygraph_tracer().trace_op(type="furthest_point_sampling_wrapper", inputs=ins, outputs=outs, attrs=attrs)
File "/home/hjl/anaconda3/envs/paconv/lib/python3.8/site-packages/paddle/fluid/dygraph/tracer.py", line 43, in trace_op
self.trace(type, inputs, outputs, attrs,
ValueError: vector::reserve